Ritter 17

Information Systems 67 (2017) 36–57
Contents lists available at ScienceDirect
Information Systems
journal homepage: www.elsevier.com/locate/is
Patterns for emerging application integration scenarios: A survey

Daniel Ritter a,b,∗, Norman May a, Stefanie Rinderle-Ma b
a
SAP SE, Germany
b
University of Vienna, Faculty of Computer Science, Austria
a r t i c l e i n f o a b s t r a c t
Article history: The discipline of enterprise application integration (EAI) enables the decoupled communication between
Received 11 February 2017 (business) applications, and thus became a cornerstone of today’s IT architectures. In 2004, the book by
Revised 13 March 2017
Hohpe and Woolf on Enterprise Integration Patterns (EIP) provided a fundamental collection of messag-
Accepted 15 March 2017
ing patterns, denoting the building blocks of many EAI system implementations. Since then, multiple new
Available online 18 March 2017
trends and a broad range of new application scenarios have emerged, e. g., cloud and mobile computing,
Keywords: multimedia streams. These developments ultimately lead to conceptual changes and challenges such as
Cloud integration larger data volumes (i. e., message sizes), a growing number of messages (i. e., velocity) and commu-
Device integration nication partners, and even more diverse message formats (i. e., variety). However, the research since
Enterprise application integration 2004 focused on isolated EAI solutions, and thus a broader and integrated analysis of solutions and new
Enterprise integration patterns patterns is missing. In this survey, we summarize new trends and application scenarios which serve as
Hybrid integration
a frame to structure our survey of academic research on EIP, existing systems for EAI and also to clas-
sify integration patterns from these sources. We evaluate recently developed integration solutions and
patterns in the context of real-world integration scenarios. Finally, we derive and summarize remaining
challenges and open research questions.
© 2017 Published by Elsevier Ltd.
1. Introduction quite common for EAI in on-premise setups. Most of the current
research focuses on RPC-style Service-oriented Architecture (SOA).
Enterprise Application Integration (EAI) addresses the re-
quirement to integrate independent applications which need to
communicate with each other [1,2]. Hence, some middleware 1.1. New challenges for enterprise application integration
is employed to abstract from the details of communication and
orchestration of applications. For the purposes of integration, a set In this paper we identify further new IT trends and application
of core Enterprise Integration Patterns (EIP) were documented in scenarios which emerged after the seminal book on EIP by Hophe
[3], which describe recurring scenarios and solutions to realize EAI and Woolf [3]. Some of these changes, e. g., Cloud and Mobile
using messaging. Computing, IoT, Microservices, and API Management, were even
Originally, EAI focused on the integration of applications within recently acknowledged by the EIP authors [5].
a single organization. However, as hosting (parts of) applications One major source for identifying new trends is the yearly pub-
in the cloud becomes increasingly popular, EAI also needs to ad- lished “Emerging Technologies Hype Cycle” report between 2005
dress scenarios where applications that are hosted in the cloud or and 2017 by Gartner [6]. We focused on the most relevant trends
on-premise (i. e., within company networks) need to be integrated. for application integration today, i. e., we excluded trends like ma-
We refer to such scenarios as hybrid applications, following For- chine learning and analytics in the analysis presented in this paper.
rester [4]. Especially hybrid applications require a stronger decou- The results are depicted in Fig. 1. Both our literature review in
pling to integrate on-premise with cloud applications, and conse- Section 2 and our system review in Section 3 are consistent with
quently, hybrid applications prefer to use (asynchronous) message- the trends identified by the Gartner reports because both academic
based communication patterns, while RPC-style integration is still research as well as concrete systems address these trends.
Broadly speaking, the early years (20 05–20 07) are dominated
by Service-oriented Architecture (SOA) and Event-driven Architec-
∗
Corresponding author. ture (EDA) styles. But also related technologies like Microservices
E-mail addresses: [email protected] (D. Ritter), [email protected] (N. are mentioned by Gartner in 2017 [6], and API Management by
May), [email protected] (S. Rinderle-Ma). Forrester for 2016–2018 [7].
http://dx.doi.org/10.1016/j.is.2017.03.003
0306-4379/© 2017 Published by Elsevier Ltd.
D. Ritter et al. / Information Systems 67 (2017) 36–57 37
Fig. 1. IT trends since 2005 in the context of application integration.
The Cloud Computing trend became prominent in 2007 and When the applications in an EAI scenario are partly hosted by
subsequently led to trends like Hybrid Computing, i. e., multi-cloud cloud providers, monitoring becomes more challenging because
and on-premise applications, from 2013 to 2015 and the move the interfaces available for monitoring may be limited. We also
from B2B to cloud-based business networks. Early developments review these and other non-functional aspects in our literature
in the Internet of Things (IoT) became influential even before review in Section 2, our system review in Section 3 and in the
20 06–20 08 even though devices were not yet affordable and wide context of real-world integration scenarios in Section 6.
spread. However, with the advent of Mobile Computing in 2010,
mobile and IoT devices and applications (since 2012) started to 1.2. Research method
play a role for application integration. As countless devices and
applications generated an increasing amount of data, Big Data This survey relies on the design science methodology [8] as a
(from 2011) became influential and a challenge not only for inte- rigid method to collect and evaluate the new trends mentioned
gration systems. Finally, humans increasingly organized themselves above, to summarize research which adds new patterns to the
in social media with its momentum from 2008 to 2012, which original EIP, and to evaluate these new patterns in the context of
evolves to personal computing, supported by wearable and mobile real world application scenarios. Fig. 3 depicts the research method
devices and applications. applied in this paper, and we use it to structure this paper.
In Fig. 2 we associate the trends mentioned in Fig. 1 with Our fundamental theory and motivation for this paper is: The
aspects of application integration. While some of the nodes repre- original EIP from 2004 do not completely cover new trends in 2016
sent the trends (i. e., without application and integration system), and beyond. From this we derive hypothesis (H1), i. e., the exist-
the edges denote required interaction and (transitive) communica- ing EIP do not suffice for all application scenarios after 2004. This
tion, which also gives hints on existing as well as new integration hypothesis is tested based on two observation artifacts, i. e., a
scenarios for the different combinations. Node spanning trends are systematic literature review in Section 2 and a systematic system
denoted by “dashed-line” nodes. review in Section 3. Based on the literature review we analyze
It is noteworthy that for hybrid application integration for both whether new trends and application scenarios can be seen after
on-premise to cloud as well as cloud to cloud communication 2004 and which solutions are provided. The system review aims at
becomes relevant, e. g., for migration of on-premise applications analyzing available systems regarding their support for integration.
to the cloud. This raises technical issues like security but also Interpreting the literature and system reviews then leads us to the
robustness in the face of errors or unavailable communication tentative hypothesis (H2) that current system implementations sup-
partners. Furthermore, cloud, on-premise and mobile applications port patterns beyond EIP which results in a strong demand for sys-
generate communication traffic of an ever increasing scale with tematic description. In order to address the detected gaps we pro-
respect to the amount of data but also the number of commu- pose new pattern categories and patterns. In hypothesis (H3) we
nication partners. In this cloud setup organizations replace the argue that some trends are handled in an (yet) immature and ad-hoc
bilateral RPC-style communication by asynchronous, message- fashion, and thus require a structuring in form of patterns. These
based interactions which are mediated by integration systems. artifacts are then evaluated based on a quantitative analysis of sev-
38 D. Ritter et al. / Information Systems 67 (2017) 36–57
Fig. 2. IT trends since 2005 an their relationship to application integration.
Fig. 3. Design science methodology used in this paper.
eral real-world integration scenarios following the hypothesis (H4) analysis of the most influential system implementations of this
solutions not in EIP can be found in real-world integration scenarios domain (→ H2).
for the trends. Finally, resulting research directions are described. • An extended pattern template plus an example based on
descriptions of cross-concern technical qualities (e. g., (stateful)
1.3. Contributions and paper outline conversation, streaming, security) for a comprehensive coverage
of new requirements (→ H3).
In this paper we make the following contributions: • The evaluation of the found patterns as part of integration sce-
narios in a well-established cloud integration system in form
• A systematic literature review of the trends, e. g., cloud and of a quantitative analysis based on new monitoring patterns
hybrid application integration approaches (→ H1), and an (→ H4).
The paper is structured as follows. In Section 2 the literature [3]) plays a dominant role, particularly in the years 2005–2013.
review of approaches that are closely related to application in- During this period, some topics such as mashups, cloud, and EDA
tegration and integration patterns is conducted by setting them were occasionally mentioned. In the last years, i.e., 2014–2016 the
into context to the new trends since 2004. Section 3 describes picture seems to change, turning away from the strong focus on
the system review with focus on non-functional aspects (NFA) – SOA towards topics such as cloud, hybrid, and IoT.
related to the trends – identified during the literature review, thus From Fig. 4 it can be concluded that some of the trends oc-
leading to a list of potentially missing functionality. In Section 4, curred in the literature after 2004 with a dominant occurrence
we discuss how to capture the functionality as patterns, similar to of SOA. Apparently, since 2014 SOA loses significance, and other
the original EIP, and discuss two patterns in more detail Section 5. trends such as IoT and cloud seem to gain more attention. From
Then we analyze and discuss real-world integration scenarios the dominance of SOA we also conclude that a more fine-grained
related to the trends and their usage of the new patterns in analysis of the mentioned topics is meaningful. Hence, in the
Section 6. Section 7 concludes the paper and states open issues following, summaries for the analyzed approaches are provided,
which are not addressed in existing work. ordered by the topics and areas they work on. Subsequently, the
list of trends will be complemented with Non Functional Aspects
2. Literature review or requirements (NFA) mentioned by literature that constitute fur-
ther important topics for EAI since 2005. Moreover, if approaches
In this section we conduct a literature review in order to provide solutions with respect to the different topics, the type of
answer the hypothesis H1: existing integration foundations in form solution will be collected.
of patterns do not suffice for all application scenarios as set out
in Fig. 3. The hypothesis raises two questions to be investigated 2.2. Literature summaries
in the literature review, i. e., a) are there any topics after 2004
not yet covered by the original EIP? and if yes b) do existing This section summarizes the approaches identified in the
approaches provide solutions to these topics?. literature search. We organize the summaries chronologically by
The literature review is based on the guidelines described following the timeline from Fig. 1. In the context of EAI, no work
in [9]. The primary selection of references was conducted using was found on the internet of things, social/personal computing,
google scholar (scholar.google.com) on 2016-10-4. The search microservices, and API management, which can be seen as suc-
string was cessor of the Service-oriented Architecture trend. In addition to
allintitle: integration patterns excluding patents the trends, for each approach we try to derive additional NFA as
and citations. As a general baseline, only papers after 2004 are well as the proposed solutions. The harvested NFA and solutions
considered as the main theory behind this study is that the EIP are summarized at the end of the section in Table 1 and yield the
from 2004 do not cover trends in 2016. Hence the time range was input for the further analysis.
set to 2005 – 2016. Overall this resulted in 525 hits. On these hits,
the following selection criteria were applied: 2.2.1. Service-oriented and event-driven architectures
According to the timeline, the first EAI solutions after 2005
• relation to computer science, enterprise application integration, were provided by Service-oriented and Event-driven Architec-
service integration, data integration, system integration tures representing mostly RPC-style solutions (i. e., a post shared
• availability of the document database and file sharing integration style, compared to “messag-
• published in English ing” like EIP, according to [3]).
• published (excluding Master theses)
Service-oriented Architecture. Hentrich and Zdun present patterns
Altogether, 52 papers were selected as relevant (the primary that address data integration issues such as incompatible data
literature list can be found here: http://cs.univie.ac.at/wst/research/ definitions, inconsistent data across the enterprise, data redun-
projects/project/infproj/1085/). These 52 papers were further ana- dancy, and update anomalies [10]. It is described how to integrate
lyzed whether they contribute as observations to the hypotheses. the application-specific business object models of various external
This resulted in removing 23 papers from the primary literature systems into a consistent process-driven and service-oriented ar-
list (for example, papers were excluded that focus on data integra- chitecture. In summary, the proposed solution combines SOA with
tion). Then a vertical search was conducted in forward and back- patterns, e. g., refactoring patterns. In [11], the authors propose
ward direction, resulting in 43 papers, including one paper that a pattern language for design issues of business process-driven
was added based on expert knowledge. After analyzing these pa- service orchestrations. The patterns illustrate how these types
pers, 34 were included in the secondary literature list. Overall, this of service invocation need to be reflected in process models in
results in 63 papers for the secondary literature list (to be found order to integrate processes with services. Implications regarding
at http://cs.univie.ac.at/wst/research/projects/project/infproj/1085/). the functional architecture are also captured by the patterns.
Specifically, the patterns reflect solutions for general business
2.1. Processing of selected literature – topics and trends requirements that can be found in SOA engagements. Overall,
the paper proposes a solution, more precisely, a pattern language
At first, all papers from the secondary literature list were an- covering, for example, Synchronous Service Activity, Fire Event
alyzed with respect to the topics they are mentioning. Comparing Activity, and Asynchronous Sub-process Service.
the harvested topics with the trends identified in the introduction In subsequent work [12] the authors present solutions to
gives an answer to question a) are there any topics after 2004 not Process-driven SOA patterns in the sense of a process integration
yet covered by the EIP?. In this first step it is sufficient that a topic architecture featuring patterns at Macro Flow (business process)
is mentioned. It was not necessary that a solution was provided. and Micro Flow level (transaction or human), as well as Integration
As the collected topics are very fine granular and spread widely, Adapter, Configurable Dispatcher, and Integration Adapter Repos-
they were first grouped according to the trends mentioned the itory. These patterns correspond to the ones proposed in [10].
introduction. Furthermore long-running business processes are distinguished
Fig. 4 depicts the distribution of topic mentions along the trend from short-running technical processes. Zdun et al. present a sur-
over time. It can be seen that SOA (i. e., RPC-style integration vey of technology-independent patterns that are relevant for SOA
Fig. 4. Distribution of topics mentioned in literature over time.
Table 1
Solutions for trends and non-functional aspects (parentheses mean partial solution).
Trends Patterns [3] Formalization [5,66] Modeling [46,47,66]
Service-oriented Architecture [10–16,18,32,37,38], security [30,31] [16], adapters [21,22,35,36,67], control flow [33], [17]
interact.[34]
Internet of Things
Event-driven Architecture [42,59]
Cloud computing [44,45], (migration [43])
B2B/ Business Network (by example [46,47])
Social/ Personal Computing
Mobile Computing SOA device patterns [25]
Big Data
Hybrid Computing (migration [43,48])
API Management (for SOA [15])
Mashups [26,27], SOA migration [29]
NFA with evidence
Asynch [3] EIP [3], strategies [56] [57,58]
Security [15,31,45,50,68,69] (for SOA [30,31])
Media [6]
Synch / Streaming [5]
Conversations [5], [6] [55], (for SOA [37]) (for SOA [19,34])
Error Handling [5,45] (EIP [3,70])
Monitoring [45,61,62] ([3])
and argue towards formalized pattern-based reference architecture and describes preliminary steps towards exploiting Enterprise In-
model to describe SOA concepts [13]. Finally, Zdun describes a fed- tegration Patterns to deal with a form of choreography adaptation.
eration model to control remote objects and proposes a solution Concretely, an adapter generator and prototype using spring inte-
based on patterns, e. g., broker and software patterns [14]. gration is presented. Example patterns comprise Message Routing
Autili et al. discuss challenges posed by the heterogeneity of Patterns, namely Message Filter, Aggregator, Splitter, and Rese-
Future Internet services [15]. Modern service-oriented applications quencer. Overall, this work bridges SOA to EAI using EIP and pro-
automatically compose and dynamically coordinate software ser- tocol adapters for services. Moreover, it is planned to integrate EIP
vices through service choreographies described based on BPMN 2.0 with security patterns and message transformation as future work.
Choreography Diagrams. The authors state that currently composi- In Gacitua-Decar and Pahl an ontology-based approach to cap-
tion and adaptation is often a manual task, Hence, they advocate ture architecture and process patterns is presented [16]. Ontology
towards the automatic synthesis of choreography-based systems techniques for pattern definition, extension and composition are
developed and their applicability in business process-driven ap- leveraged in order to compose reusable mashup components. The
plication integration is demonstrated. The proposed solution is an authors also present a service oriented architecture that addresses
architecture framework for SOA-based EAI as well as an ontology- reusability and integration needs for building enterprise mashup
based notion of patterns to link business processes and service applications. The proposed solutions focus on SOA and mashups,
architectures. This could be seen as a formalization approach. A but no solution to EIP and new trends is provided. The work by
SOA service integration framework with a pattern-based modeling Braga et al. addresses issues of complexity of service composi-
approach is presented by Heller and Allgaier [17]. It features con- tions with adequate abstraction to give end users easy-to-use
trolled extensibility of enterprise systems for unforeseen service development environments [28]. Abstract formalisms must be
integration and can be estimated as similar to related B2B Integra- equipped with suitable runtime environments capable of deriv-
tion and Enterprise Application Integration. The framework lever- ing executable service invocation strategies. The solution tends
ages structural or behavioral interface mediation techniques. The towards mashups and modeling as users declaratively compose
modeling approach with adaptation patterns and runtime support services in a drag-and-drop fashion while low-level implementa-
is demonstrated with a UI integration prototype in the automotive tion details are hidden. However, the solution could not be clearly
domain. Overall, this work suggests pattern-based modeling as identified and is hence not included in Table 1. Finally, Cetin et al.
solution. Kaneshima and Braga analyse whether EAI can be con- chart a road map for migration of legacy software to pervasive
ducted by web services and SOA or DB sharing [18]. Both solutions service-oriented computing [29]. Integration takes place even at
are being adopted by organizations, although they present advan- the presentation layer. No solution is provided for EIP and trends,
tages and disadvantages that should be analysed. This work doc- however, mashups are used as migration strategy to SOA based for
uments these problems and solutions in the form of patterns like the Web 2.0 integration challenge.
access via Shared Database, direct RPC-style integration via web
services, Intermediate Duplication with access via DB or web ser- SOA Security. Qu et al. present six bilateral patterns (Binding,
vices. Hence, the proposed solution is based on SOA and patterns. On-demand, Tailor, Composite, Contract and Migration) and four
Umapathy and Purao transform EIP to web service implemen- multilateral patterns (Separated, Shared, Mediated and Enhanced)
tations using a transformation model called ceipXML [19]. The as a solution for integrating new services with Grid security
proposed solution comprises conversation models that may be services [30]. For each pattern, the authors discuss its intent,
used to implement interactions among Web services as well as applicability, participants and consequences. Shah and Patel anal-
a methodology that generates the design elements in the form yse the security requirements for global SOA [31]. For security
of conversation policies for Web services. Current integration ap- concerns, dynamic configuration of handlers, sequence, and iden-
proaches do not support the end user development requirements tification of handlers is proposed as solution. Fisher et al. provide
for infrequent, situational or ad-hoc integration and collaboration practical implementations in Java and .NET for interoperable, syn-
as stated by Zheng et al. [20]. The work differentiates between chronous, and asynchronous integration [32]. Hence, the proposed
UI, component, business logic, resource and data integration. solution consists of implementation details for SOA, WS security
Gierds et al. define an approach for behavioral adapters based on examples, and best practices such as a secure object handler
domain-specific transformation rules that reflect the elementary adding custom interceptor logic for, e. g., adding digital signatures.
operations that adapters can perform; synthesize complex adapters
that adhere to these rules [21]. The proposed solution comprises SOA and Business Processes. Ouyang et al. formalize process con-
a formalization, specification of the elementary activities to model trol flows into BPEL processes by an intermediate translation to
domain knowledge, separating data from control, and a reduction Petri nets [33]. From the same group, Wang et al. construct and
from adapter synthesis to controller synthesis. An adapter is gen- interface adaptation machine that sits between pairs of services
erated to reconcile mismatches (e. g., incompatible protocols) in and manipulates the exchanged messages according to a repository
Sequel et al. [22]. The proposed solution is constituted by a survey of mapping rules. For both approaches, the proposed solution is a
of protocol adapter generation (e.g., semi-automated protocol formalization. Lohmann et al. analyze the interaction between WS-
adapter generation). Gudivada and Nandigam deal with EAI using BPEL processes using Petri nets [34]. Again the proposed solution is
extensible Web services [23]. A solution is not directly proposed, a formalization. With a similar goal, Kumar and Shan aim at sim-
but rather a practical implementation. Deng et al. combines SOA plifying the pattern compatibility based on a matrix and rules that
and Web service technology to simplify EAI by studying the enable the simplification of checking compatibility between two or
service-oriented software analyzing and development characteris- more processes because these prerequisite rules can be applied to
tics [24]. The approach distinguishes between vertical integration each pattern separately [35]. The proposed solution is an algorithm
within an enterprise while B2B emphasize on the horizontal inte- and can hence be subsumed as formalization. Mismatch patterns
gration. Again the paper presents a more practical implementation. that capture the possible differences between two service (busi-
ness) protocols to adapt and automatically generate BPEL adapters
SOA and Mobile Computing. Mauro et al. [25] target design prob- are presented by Jiang et al. [36]. They introduce several depen-
lems of SOA for mobile devices with Service Oriented Device dencies such as transformation dependency (incl. message split),
Architecture (SODA). For this SOA design patterns like Enterprise synchronization dependency, choice dependency (choice among
Inventory are analyzed with respect to their applicability to two or more messages), and priority dependency. The proposed so-
SODA, and new pattern candidates like Service Virtualization are lution is the formalization of mismatches. Barros et al. propose SOA
identified. From these candidates new (device) patterns including process interaction patterns including Send, Receive, Send/Receive,
Auto-Publishing, Dynamical Adapter, Server Adapter, Integrated and Racing Incoming Messages [37]. Patterns for synchronization
Adapter, External Adapter are proposed as solution. problems in the area of process-driven architectures, e. g., Wait-
ing Activity or Timeout Handler, are introduced by Köllmann and
SOA and Mashups. Liu et al. combine several common architecture Hentrich [38]. Vernadat looks at architectures and methods to
integration patterns, namely Pipes and Filters, Data Federation, build interoperable enterprise systems, advocating a mixed service
and Model-View-Controller to compose enterprise mashups [26]. and process orientation and the classification of integration levels,
Moreover, these patterns are customized for specific mashup physical system, application, business integration, and enumerates
needs. In [27] enterprise architecture integration patterns (e. g., SOA concepts [39]. No specific solution is proposed. Grossmann
Pipes and Filter, Data Federation, Model-View-Controller) are et al. derive integration configurations from underlying business
processes, e. g., activities [40]. Future work names exception han- Hybrid Applications. A major challenge in hybrid applications is
dling as challenge, however no solutions are provided. the decision where to host parts of the application. In this regard,
Mansor recommends to bear in mind the patterns in the envi-
Event-driven Architecture (EDA) and SOA. Taylor et al. address the sioned process [48]. The work addresses technical challenges when
SOA - EDA connection as service network and provide a reference implementing a hybrid architecture. The proposed solution refers
EDA manual [41]. As no solution is provided, the approach is not to architectural patterns. A holistic approach for the development
included in Table 1. A theoretical framework for modeling events of a service-oriented enterprise architecture with custom and
and semantics of event processing is provided by Patri et al. [42]. standard software packages is presented by Buckow et al. [49]. The
The formal approach enables to model real-world entities and their system architecture to be developed is often based on integration
interrelationships and specifies the process of moving from data patterns for the physical integration of systems. No solution is
streams to event detection to event-based goal planning. More- provided in the context of this work.
over, the model links event detection to states, actions, and roles
enabling event notification, filtering, context awareness, and esca- 2.2.3. Internet of things and big data
lation. The proposed solution consists of events and formalization. With affordable and widespread mobile sensors and devices
comes the Internet of Things and together with the immense
amount of data from cloud and mobile computing comes Big data.
2.2.2. Cloud computing, business networks, and hybrid applications
The successor of grid and cluster computing is cloud computing
Internet of Things (IoT). Heiss et al. collect challenges in cyber-
that extends B2B to business networks, and the coexistence of
physical systems such as communication quality, interoperability,
applications on-premise and in several kinds of cloud platforms as
and massive amounts of data [50]. As interesting requirements
hybrid applications.
they state “placement” (of integration scenarios), e. g., cloud or
on-device, the demand for global optimization, more intelligent
Cloud Computing. Asmus et al. focus on the migration of enter- devices, networking and cloud and security including data security
prise applications to the cloud [43]. Integration is considered a key and privacy etc., decoupling of layers vs. direct data access for
factor influencing cloud deployment. Several migration patterns on-top applications. Rather than proposing a solution, the indus-
are described as a basis for enabling enterprise cloud solutions. trial and business perspectives on such envisioned platforms are
The following challenges are named in the paper: data volume, described.
network latency, identity and data security management, interop-
erability (i. e., supporting the trends big data, security, and variety Big data. Ritter and Bross suggest moving-up relational logic
as in multimedia). Asmus et al. state that “integration pattern can programming for implementing the integration semantics within
be a starting point in deciding integration options” [43]. The key a standard integration system [51]. For this EIP semantics is trans-
areas addressed in the approach include on premise, off-premise lated to relational logic. For declarative and more efficient middle-
private cloud, cloud integration, cloud service provider, and ex- ware pipeline processing (e. g., parallel execution, set-operations),
ternal users. The integration patterns refer to process to process the patterns are combined with Datalog. The expressiveness of the
and data integration. Overall, the proposed solutions are “patterns approach is discussed, and a practical realization by example is
and processes-based” methods for an initial evaluation of the risk provided. Although no direct solution to the trends is provided the
and effort required to move new and existing applications to a approach directs to “data-aware” integration patterns.
cloud service. In Ritter and Rinderle-Ma, a collection of integration
patterns derived from requirements of hybrid and cloud appli- 2.2.4. General EAI approaches
cations is presented [44], thus propose a solution for cloud and From practical EIP implementations to ideas for new patterns,
patterns. The main challenge described by Merkel et al. is a secure formalization approaches, enabling techniques and domain-specific
integration [45]. The approach proceeds in a top-down manner by work, this section rounds off the literature analysis with further
deriving integration patterns from scenarios and in a bottom-up EAI challenges.
fashion by deriving patterns from case study requirements. It
identifies the need for security (access control, integrity, confiden- Practical Aspects. Scheibler and Leymann present a framework for
tiality) as well as security constraints (e. g., EU Data Protection configuration capabilities of EIP, specifically for code generation
Directive) and presents an evaluation based on an architecture based on a model-driven architecture [52]. In [53], EIP are im-
with major focus on hybrid and multi-cloud setups. The described plemented in IBM WebSphere. Again no solution for the trends
patterns are cross-cloud ESB, usage of ESBs, as well as security is provided, but a solution to the EIP through implementation.
patterns as architecture components such as LDAP. The approach Thullner et al. analyse EIP coverage in open source tools and
only works in private clouds. Merkel et al. propose future work on implement a sample scenario in Apache Camel and Mule [54]. No
public cloud that involves content encryption, key management, solution is provided.
data splitting, computing with encryption functions, anonymiza-
tion, data masking, and encrypted virtual machines. They mention EAI Patterns. [55] presents a pattern language for conversations
Cross-Cloud Balancer, Cross-Cloud Data Distributor, and replication between loosely coupled services, i. e., patterns are suggested as
patterns as further future work. Other challenges mentioned are solution. Gonzales and Ruggia deal with response time and service
cross-cloud monitoring and cloud management. In summary, the saturation issues (more requests than can be handled) using an
proposed solution are new patterns for SaaS integration and adaptive ESB infrastructure [56]. They propose solutions in the
centralized as well as decentralized multi-cloud integration. form of strategies, i. e., Delayer, Defer Requests, Load Balancing,
and Cache.
Business Network. Ritter provides mappings of EIP integration
semantics and patterns to BPMN-based models as well as an Formalization and Verification. Fahland and Gierds present a con-
implementation of a business network scenario example [46,47]. ceptual translation of EIP into Colored Petri nets, hence providing
Both works do not directly propose a solution to the trends a formal model based on a system specification using EIP [57,58].
depicted in Fig. 4, but introduce modeling as a possible solution in The Petri net based formalizations can be used to simulate and
the context of EIP, thus added as category to Table 1. conduct model checking of pattern compositions. Though the
formalization can be understood as solution, it does not address 2.3. Synthesis and discussion of non-functional aspects
any new trends beyond EIP, thus this approach is not contained in
Table 1. A semantic representation of EIP for automatic manage- The second aspect of our analysis of trends are topics that were
ment of messaging resources (e. g., Channels, Filters, Routers) is named by Gartner [6] and Zimmermann et al. [5] as relevant or
presented by Patri et al. [59]. The application is to connect mobile that were identified during the literature review. However, these
customers to Smart Power Grid companies. Data is accessed in topics have a more cross-cutting quality (i. e., relevant for several
form of alerts from a complex event processing engine using trends). We call them non-functional aspects or requirements
SPARQL queries. The proposed solution is a formalization for re- (NFA), which we appended to Table 1 together with the references
source management of integration patterns. Basu and Bultan focus that supported them as challenges (as evidence). They are set
on the interaction behavior in asynchronously communicating into context to important aspects, when working with integration
systems resulting in decidable verification for a class of these scenarios, namely patterns, formalization and modeling. The focus
systems [60]. As the proposed solution (formalization) is not in on patterns comes from the EIP [3] and supported by many related
the context of the trends, it is not included in Table 1. Mederly domains, that capture knowledge and best practices in form of
et al. generate a sequence of processing steps needed to transform patterns (e. g., SOA, Cloud Computing). Panetto et al. [66] bring up
input message flow(s) to specified output message flow(s) [61,62]. the formalization (supported by [5]) and modeling (supported by
The work takes into account requirements such as throughput, [46,47]) as additional relevant topics. We now set these topics into
availability, service monitoring, message ordering, and message context with the references from the literature analysis in Table 1.
content and format conversions. Additionally, it uses a set of For the EIP, we added asynchronous message processing as
conditions, input and output messages, and a set of configuration Asynch to cover the solutions in this space, e. g., by [3,56]. For
options. Control flow ordering is formalized. The work is excluded the NFA, solutions in the area of formalizations are proposed by
from Table 1 because it provides no solution, but rather creates [57,58] for the validation of pattern composition and business
parts of integration solutions from the description of what has to processes. Another NFA is Security, which was seen as challenge
be achieved, not how it should be done. at least by Gartner [6] and in the literature by [31,50,68] (in gen-
eral), by [69] (performance concerns, real-time integration), and by
EAI enabling techniques. The following approaches address dif- [45] (e. g., safe integration, indications that content encryption, key
ferent enabling technologies. However, neither are the presented management and more is missing). Autili et al. [15] mention the
approaches related to the trends, nor do they propose concrete so- need for security integration patterns. The solutions for patterns
lutions. Hence they are not included in Table 1. Architectural pat- are limited to SOA with patterns like Secure Service Consumption
terns (e. g., Remote Process Invocation, Batch Data Synchronization, or Security Handler Information Exchange [30,31].
SOA, Pub/Sub, P2P, Broker, Pipes and Filters, Canonical Data model, According to Gartner, multimedia format handling and process-
Dynamic Router) are contributed by Kazman et al. [63]. This work ing can be seen as a non-functional requirement [6]. This includes
constitutes a guideline for IT architects that combines existing pat- image, video and text image formats, which are increasingly
terns. Land et al. integrate the existing software after restructuring produced through mobile devices and, e. g., interacted on social
or merger, i. e., address the question of how to carry out the inte- media, becoming of increasing interest for (business) applications.
gration process [64]. Multiple case studies and recurring patterns In the context of the big data challenges of integration systems
for vision process and an integration process are provided as well. (from Gartner; e. g., volume, velocity, stability), (synchronous)
Basic concepts of enterprise architectures including integration streaming protocols are seen as one possible solution. The au-
and interoperability are summarized by Chen et al. [65]. thors of [5] mention that patterns as well protocols are currently
missing in EAI.
With more and more communication partners that result
Domain-specific Approaches. Cranefield and Ranathunga integrate from the trends in Section 1.1, (stateful) conversational protocols
agents with a variety of external resources and services using might be required, according to [5] and also Gartner (e. g., device
Apache Camel and the EIP endpoint concept [71]. e-Learning as meshes). First ideas have been sketched by Hohpe [55] with
a growing and expanding area with huge number of disparate an initial collection of conversation patterns, which should be
applications and services is addressed by Rajam et al. [68]. The extended [5]. For SOA web service conversation policies [19] and
approach redefines the Model-View-Controller pattern. It can interaction patterns [37] solutions were provided. Formalizations
be further enriched to encapsulate certain non-functional and have been proposed in [34] for the SOA domain with focus on the
integration activities such as security, reliability, scalability, and controllability of a process. The proposed solutions for SOA might
routing of request. As all these approaches do not propose a be transfered to integration processes as starting point for more
solution directly connected to EIP and the trends, and hence they general conversation patterns.
are not included in Table 1. To handle erroneous situations during message processing, es-
calate them and make systems more fault-tolerant, error handling
EAI Challenges. A survey to motivate some more challenges in the is seen as a major aspect [5,45]. Hohpe et al. [3,70] do only cover
area of enterprise application integration and to link back to the Dead Letter Channel as solution and sketch some ideas about the
trends is presented by He and Xu [69]. Further this work examines topic. Overall, in the literature, the topic is neither addressed from
the architectures and technologies for integrating distributed a pattern, formalization, nor modeling perspective. While [5] men-
enterprise applications, illustrates their strengths and weaknesses, tions missing patterns and formalization, Merkel et al. [45] lists
and identifies research trends and opportunities for horizontal and Balancing and Distribution, as well as [69] mentions Fault-
vertical integration. Though no solution is proposed, the discov- tolerance and Message Scheduling as missing aspects. Similarly, the
ered trends are strengthened, for example, SOA, personal, mobile, insight into the current state of affairs, called monitoring, for ser-
and IoT. The survey also addresses NFA, e. g., security, which are vices and cross-cloud are seen as important topics in [45,61,62].
collected and serve as input for Table 1. Another survey by Panetto The monitoring of integration processes as well as cross plat-
et al. discusses trends and NFA in enterprise integration [66]. form monitoring were only mentioned, however, no solution was
Moreover, modeling and formalization (formal methods such as provided. The Control Bus, a Wire Tap and the Message History
verification) are proposed as challenges, but no concrete solution patterns in Hohpe et al. [3] denote partial solutions, which can be
provided. used to build a monitoring solution on integration process level.
Table 2
System review - horizontal search.
Category hits selected Selection criteria Selected Systems
Commercial 12 7 Gartner and Forrester IPaaS Quadrants Dell Boomi [72], IBM Cast Iron [73], Informatica [74], Jitterbit [75], MS
BizTalk [76], SAP Cloud Integration [77], Oracle [78]
Startup 20 2 cloud/data integration, B2B, API, #followers Tray.io [79], Zapier [80]
Open Source 13 2 application integration, data ingestion Apache Flume [81], Apache Nifi [82]
Wikipedia 34 1 enterprise application integration; non-duplicates Apache Camel [83]
Added Systems n/a 3 expert knowledge Cloudpipes [84] (startup), Tibco [85], WebMethods [86] (commercial)
Removed Systems – –
Overall 74 15
3. System review munity projects. Since the main focus lies on commercial systems
that are known to be less well accessible for a systematic analysis
This section reports on the results of a system review to of their features, we focus on the publicly available material (i. e.,
answer hypotheses H2 the EIP of the 2004 book are all widely without registration or login) and try to get more information by
used in praxis, and H3 current system implementations do support possibly underlying open source systems, where possible.
more patterns as set out in Fig. 3. The system review is based on
the guidelines described in [9] for a horizontal search including
“well-established” commercial application integration systems, 3.1. Processing of selected systems
more experimental systems from startups, open source systems
and public knowledge in form of a Wikipedia search. The selection 3.1.1. EIP Solutions used in system implementations
of systems was conducted on 2016-10-04, and the results of the We start our system review with an analysis of all selected
horizontal search are summarized in Table 2. The NFA are used to systems with respect to their implementation of EIP solutions. The
focus the search in those systems. EIP describe six pattern categories, namely, Messaging Channels,
First, seven commercial systems were collected by taking the Message Construction, Message Routing, Message Transformation,
systems listed in both, the Gartner (Leaders, Visionaries, Chal- Messaging Endpoints and System Management. We focused the
lengers) [87] and the Forrester (Application Integration) IPaaS analysis on the two pattern categories of message routing and
list [88] – out of 12 systems, leading to the following systems: transformation, since they represent the core aspects of integra-
Dell Boomi [72], IBM Cast Iron [73], Informatica [74], Jitterbit tion systems. Furthermore we left out composed patterns (e. g.,
Harmony Cloud Integration [75], Microsoft BizTalk [76], SAP Cloud composed message processor, scatter-gather), when their single
Integration [77], and Oracle Cloud Integration [78]. We scratched parts were already in the selection. Table 3 (from Boomi to Oracle)
MuleSoft due to its similarity to Apache Camel [83], which we and Table 4 (from Flume to Webmethods) depict the solutions
selected as expert addition from Wikipedia (discussed later). found in the system implementations that could be associated to
In addition, two startup systems from the top 20 overall the routing and transformation patterns.
systems were selected due to their number of followers on The Apache Camel system seems to be specifically designed
angel.co,1 namely Tray.io [79] and Zapier [80]. While the for- around the EIP, thus supports nearly all EIP and sticks to the
mer is striving to build an “Integration Marketplace” for enterprise original EIP naming for the respective solutions, which makes
applications, Zapier is a cloud integration startup. it a benchmark for the others. Most notable deviations are the
Out of 13 open source systems of the Github Hadoop Ecosys- Envelope Wrapper (i. e., wrap application data inside an envelope,
tem,2 we selected Apache Flume [81] and Nifi [82] as data compliant with the messaging infrastructure) and Message Trans-
ingestion systems according to the selection criteria (cf. Table 2). lator patterns (i. e., translate one data format into another one; not
We scratched the application integration systems Talend (also in transformation patterns). None of them is directly represented
listed as commercial system), Spring Integration and MuleESB for in Camel, however, can be implemented using UDFs (i. e., user-
their similarity to Apache Camel as well as Apache Beam, Apache defined functions like Camel Processor) or scripting (e. g., Camel
Sqoop and Spring XD for their similarity to Apache Flume. Script), therefore marked as partially covered by parentheses.
The open source integration system Apache Camel [83] does The most common routing pattern solutions are the Content-
not appear in the open source list, however, it was the only based Router, the Splitter and the Aggregator. Since the Message
non-duplicate from the other lists that has to be selected, since Filter is a special case of the content-based router and filter can
it implements the existing EIP from [70] and is a role model for be used to construct the latter, not all systems provide implemen-
many systems like Spring Integration, or Red Hat’s FuseESB. tations for both of them. The splitter is sometimes implemented
The software systems of Tibco [85] and Software AG [86] are according to the description in the EIP, however, some vendors
wide-spread and influencial integration systems for on-premise decomposed it to its iterative core functionality (e. g., For Each in
with a cloud integration offering and are listed among the top for IBM, Oracle, Cloudpipes). The aggregator shows many partial solu-
wide integration and deep integration for traditional on-premise tions that require user-defined functions (e. g., Informatica, Oracle,
by Forrester.3 Hence we add them as expert selected additions. Tray.io), while only few provide its EIP functionality (e. g., Aggre-
We add Cloudpipes [84] from the startup list as cloud integration gator in BizTalk, SAP Cloud Integration, Tibco or ContentMerge in
system. Apache Nifi).
That leaves us in total with 15 systems with a good mix of The transformation patterns seem to play a major role in the
well-established commercial and startup products, as well as com- analyzed systems, since most of them are broadly supported.
However, there seems to be a tendency to provide UDF capabilities
and leave the burden to the user to deal with the semantics.
1 Finally, the dynamic routing patterns (e. g., Dynamic Router),
Angel.co, visited 02/2017: https://angel.co/data-integration
2
Hadoop Ecosystem on Github, visited 02/2017: https://hadoopecosystemtable. those patterns that contain the recipient in their content (e. g.,
github.io/ Recipient List, Routing Slip), and the Message Resequencer, e. g.,
3
The Forrester Wave: Hybrid2Integration, Q1 2014 used for the exactly-once-in-order service quality [89], were
Table 3
Original EIP used in systems; supported , partial (), unknown/not supported.
Pattern Boomi IBM Informatica Jitterbit BizTalk SAP Oracle
Content-based Router – –
Message Filter – – – – –
Dynamic Router – – – – – –
Recipient List – – – – – – –
Splitter ()
Resequencer – – – – – –
Routing Slip – – – – – – –
Aggregator – – () – ()
Envelope Wrapper () () () () – – –
Content Enricher () – – () () –
Content Filter () – – () () –
Claim Check () – – () – () -
Normalizer () – () () () –
Message Translator – () –
Table 4
Original EIP used in systems; supported , partial (), unknown/not supported.
Pattern Flume Nifi Camel Tray.io Zapier Cloudpipes Tibco Webmethods
Content-based Router () –

Message Filter – – – –
Dynamic Router () – – – – – –
Recipient List – – – – – – –
Splitter – –
Resequencer – – – – – – –
Routing Slip – – – – – – –
Aggregator () () – – –
Envelope Wrapper – – () – – – – –
Content Enricher () () ()
Content Filter () () () ()
Claim Check – – – – – () –
Normalizer () – () () ()
Message Translator () () () ()
sparsely implemented. This leaves the question on their relevance handling and monitoring, which allow to control the behaviour of
or other components that take over their function. the integration scenarios.
The classic application integration addresses the variety prob-
Summary. While some of the EIP like Content-based Routing or lem for textual message formats [2]. With the availability of
Message Filter, Splitter and Content Enricher can be found in most integration systems for “everybody” (e. g., in form of a cloud
of the systems, others are rarely implemented (e. g., Resequencer, integration system) non-textual formats gain importance.
Routing Slip). The analysis of these patterns and their relevance The trade-off between stateful and stateless message processing
are left for future work, and thus not analyzed further. is represented by storage capabilities in integration systems, for
which nearly all vendors propose a solution and conversations.
3.1.2. New solutions not covered by system implementations The stateful approaches could be represented by conversational
We now analyzed the collected systems with respect to their protocols, which allow to move the state from the integration
functionalities according to the harvested NFA from Section 2.3), systems to the communication partners (idea sketched in [55]).
namely security, media, streaming or more abstract “processing”, Most of the service qualities (e. g., at-least-once, exactly-once pro-
conversations, error handling, and monitoring. Comparing the cessing) [3,89] require stateful integration processes. Consequently,
NFA with the collected system functionalities, while neglecting this would require changes in the applications. Current systems
functionalities covered by the EIP, gives an answer to the question provide only rudimentary support, if at all.
which topics are required and used in addition. Hereby, the system Finally, a broad variety of miscellaneous topics was collected,
functionalities represent an implemented solution as part of an e. g., sentiment analysis, natural language translators, but also
actual integration system. more general functions like sort, loops, as well as explicit format
Fig. 5 depicts found solutions not covered by the EIP by NFA handling, i. e., marshalling and type conversion.
and system vendor. During the analysis new NFA were identified
– not mentioned by Gartner, the EIP authors, or the literature – Summary. Notably, security and error handling (and monitoring)
that seem to play a role in practical terms: stateful integration capabilities are predominantly found. They address the challenges
processes using storage, (pattern) composition (mentioned in of security and management. Furthermore, solutions for the in-
Zimmermann et al. [5]), and system operations. These three new creasing variety of message formats (cf. media) as well as the
NFA were included into the analysis of the other systems as well. volume and velocity handling can be found in the systems are
All non-related topics are collected as miscellaneous (Misc). part of new processing types. The storage of data and message
Notably, all identified NFA are at least partially covered by semantics like quality of service are relevant for integration sce-
system implementations, indicating that solutions in form of narios. This leads to the trade-off between stateful vs. stateless
conceptual definitions are required (e. g., as patterns). According integration processes, which briefly address in Section 5. The
to Mulesoft [90], the major challenges in cloud integration systems (stateful) conversations, which could be part of a solution, are
are security and management. The management includes error currently under-represented.
Fig. 5. Solutions for NFA not covered by the EIP by system vendor.
3.2. System summaries along NFA store is used. The availability of integration scenarios is not only a
stability, but also a security concern. Therefore some vendors like
3.2.1. Security IBM, SAP Cloud Integration provide implicit countermeasures, e. g.,
The aspect of confidentiality or message privacy is solved redundant message stores with high availability and disaster re-
on transport, message and storage levels. The transport level covery, as well as Apache Flume with explicit MorphlineSolrSinks
channel encryption can mostly be specified in the inbound and and Kafka Channel configurations. Finally, changes to the message
outbound adapters in form of the transport protocol (e. g., HTTPS, are tracked for auditing purposes. This is made explicit as Audit
SFTP) and guarantees that the message cannot be read during Trails in Jitterbit and Oracle or Service Auditing in WebMethods.
transmission (e. g., Jitterbit’s Transmission Protection). Once, the
message is received by the inbound adapter and handed to the 3.2.2. Media
subsequent operation in the integration process, message privacy The literature review in Table 1 shows that there are no so-
can be applied or reversed. Therefore many vendors provide lutions for multimedia processing in application integration or
explicit message encryptors and decryptors (e. g., PGP Encrypt related domains (e. g., SOA, EDA). The system analysis does only
and Decrypt from Dell Boomi, AES_ENCRYPT from Informatica or provide few, superficial solutions. For instance, textual represen-
Encrypt / Decrypt in Apache Nifi), or encrypting adapters (e. g., tation of binary content is explicitly configurable in most of the
FileProcessorConnector in Informatica, FileChannel in Apache systems (e. g., Base64 Encode / Decode in Dell Boomi and IBM,
Flume, WSSProvider in Tibco). The encrypted storage of messages Encoder / Decoder in SAP Cloud Integration). These encodings play
helps to protect the message’s privacy in the store, e. g., can be a major role when communicating with remote applications, but
configured in SAP’s DBStorage and Persist operations. The configu- also when calling services (e. g., user-defined operations) using
ration of the message privacy solutions mostly include encryption textual message protocols. In addition, most of the vendors allow
algorithms, key lengths and certificates. Similarly, the integrity user-defined operations in form of scripting capabilities (e. g.,
and authenticity of a message can be ensured on the different Script, Processor in Apache Camel, Expressions in Informatica).
levels. Most of the vendors provide configurations for safe and With that, more complex operations can be performed like the
authenticated transport (e. g., using user and password, certificate compression of – usually bigger – multimedia messages. Despite
or token-based authentication). The transport is considered safe if that, pre-defined compression operators can be found in, e. g., Dell
changes of the message can be recognized by the receiver and the Boomi, Jitterbit, Apache Nifi, which allow to configure the com-
authenticity guarantees that the sender is the expected one. For pression type (e. g., zip). The explicit support of scripting seems to
instance, most of social media endpoints like Twitter and Facebook be a general trend, when representing transformation patterns (cf.
use token-based OAuth authentication. In addition, many vendors Section 3.1.1). This could either mean that the implementations
provide explicit message signers and signature verifiers (e. g., are too diverse to formulate a general solution or indicate that
Digest/Hash function in IBM, Signer and Verifier in SAP Cloud the topic was not considered yet. The support of explicit image
Integration) as well as safe message storage is provided, e. g., by processing operations seems to be limited to Nifi’s ResizeImage
Jitterbit or SAP Cloud Integration. For the storage, the authenticity and ExtractImageMetadata functions as well as IBM’s Read MIME
seems to be implied, since the cloud platform message or data activity. The only real multimedia operation is the image resizing,
since the metadata simply provide a format specific capture of the for tolerance, and message validation, load balancing (cf. [56]) and
defined image’s meta tags. flow control to prevent errors. More recent work [93,94] – not
found in the literature review – covers all of the found system
3.2.3. Processing solutions as patterns and shows their composition. Furthermore,
While the literature review does not show solutions for mes- it introduces the concept of compensations (e. g., for undo
sage processing, especially not for “data-aware” or Big Data pro- operations), which was not found within the reviewed systems.
cessing, the systems implement solutions. The canonical solution
for processing larger amounts of data is to scale-out to multiple 3.2.6. Monitoring
processing units, constituting parallel subprocesses. The parallel The monitoring of integration scenarios gains importance es-
processing of one message in subprocesses using a broadcast can pecially within integration platforms hosted by a third party and
be done, e. g., in BizTalk with Create concurrent flows, SAP Cloud across those platforms in cloud and mobile computing. The major
Integration Gateways, or Apache Camel Multicast. Furthermore, the monitoring aspects found in the systems can be distinguished
explicit configuration of parallel processing within an integration into UI components that show important aspects of the system,
process (i. e., not process parallelization) is supported by, e. g., called monitors, and a rather event-based registration on instance
Dell Boomi using the Flow Control properties, Jitterbit Parallel Pro- level. For example, Dell Boomi supports message change events
cessing, Tibco Non-inline subprocesses and Critical Section, BizTalk by Find Changes, which can be extended to Field Tracking. Oracle
Scope batch property. Alternatively, a more data-centric approach, supports the latter with Business Identifiers. That means, user-
however, impacting the latency of the process, is micro-batching defined events on technical and business level can be tracked via
[91]. Vendors like Dell Boomi and Jitterbit also support batch pro- conditional events. Examples for monitors can be found in most of
cessing of messages using the Flow Control properties or Chunking the systems across all parts of an integration scenario (i. e., from
configurations. The processing of message streams allows the adapter or channel, over component, down to message monitors).
system to handle larger amounts of data than the integration The monitors can be fed by built-in and user-defined message
system resources would allow. This more connection oriented ap- interceptors (e. g., in Apache Flume and Camel), which allow
proach was identified by Zimmermann et al. [5] as missing pattern scenario specific monitoring. When integrating hybrid applications,
category in the context of synchronous message processing. An most systems provide central, cloud monitors instead of local ones.
explicit streaming support is provided, e. g., by Jitterbit Streaming
Transformation and Apache Camel. However, not all integration 3.2.7. Storage
operations or adapters are (conceptually) capable of streaming. An integration system requires persistent stores and queues
to be operable, e. g., for system management and monitoring. In
3.2.4. Conversations addition, message delivery semantics (e. g., reliable messaging like
Gartner [6] as well as Zimmermann et al. [5] mention the “exactly-once-in-order”) [89], secure messaging, and legal aspects
importance of conversations for messaging. These conversations (e. g., “Which messages were received and processed?”) must be
are similar, however, stand in contrast to the choreography (e. g., ensured. In the literature only simple messaging related storage
[15,92]) and interaction patterns for services [37] in SOA because are mentioned like the Message Store [3], for storing messages
they denote more complex tasks than sending and receiving during processing, and the Claim Check [3] to store (parts of) the
data or messages. They target complex (stateful) conversations message during processing and re-claim them later. Consequently,
as partially covered in [55]. Some of the systems allow a timed several system vendors identified the need for additional storage
redelivery of messages in a non-error case (e. g., SAP Cloud Inte- capabilities, summarized to data stores and their access (e. g., DB
gration, Apache Camel). This feature is similar to the Contingent Update in Jitterbit, DBStorage in SAP Cloud Integration), as well as
Requests pattern in [37]. For a conversation, an acknowledgement memoization and caching during one instance of a scenario or be-
mechanism would be required similar to [55]. One technique tween them (e. g., Add to Cache in Dell Boomi, Shared Variables in
of reducing the number of requests to an endpoint is request Tibco, Global Variables in Jitterbit). For secure messaging, security
caching. In Tibco, request caching can be configured by specifying related storage like the Key Store (e. g., in Apache Flume, Camel
time slices and operations in the Caching Stage. The SAP Cloud and SAP Cloud Integration), for storing certificates and secure key
Integration system allows to map synchronous to asynchronous material, and the Secure Store (e. g., in SAP Cloud Integration), for
communications and vice versa. This becomes necessary when the storing secure tokens, users, and passwords, can be found.
endpoints’ message exchange mechanisms do not fit.
3.2.8. Composition
3.2.5. Error handling In Zimmermann et al. [5] the composition of EIP is mentioned
Error handling is a crucial aspect of integration scenarios [5] for as one of the missing pieces. Many system vendors, e. g., Dell
the control and fault tolerance aspects. In the literature we found Boomi, IBM, BizTalk, SAP Cloud Integration, allow subprocess mod-
solution attempts [3,70] like the Dead Letter Channel pattern for eling as well as delegation from the main integration process. One
the collection of failing messages, while the systems implement important solution are integration process templates, which are
various, more sophisticated solutions. The fundamental topic for configurable re-use processes. Many of the vendors support them,
dealing with errors in integration scenarios is the handling of however, under different names, e. g., Template integration process
exceptions. Therefore, most of the systems provide a “catch-all” in IBM, Snapshot of Jitterpack in Jitterbit, Blueprint in Cloudpipes.
capability (e. g., Catch All in IBM), which sometimes even come
with an exception subprocess for more advanced handling (e. g., 3.2.9. Miscellaneous
Exception Subprocess in SAP Cloud Integration). In addition, ven- The most notable, specific features are explicit or implicit loops,
dors like Dell Boomi, IBM, SAP Cloud Integration and Tibco provide keyword search and replace as well as content sort and format
more fine-granular scoping of exception handlers, e. g., down handling. The implicit loop configurations include While Loop
to the single operation. More advanced topics include escalation, activity, e. g., in IBM, Looping in BizTalk, and the Loop Collection
fault-tolerance and eventually prevention techniques. Most notably, connector in Tray.io. Explicit loops are possible in most of the
the systems support escalation mechanism like (partial) abortion systems by back-references in the process. Dedicated search and
of complex processes (e. g., incl. parallel processing) and raising replace functionality is provided, e. g., in Dell Boomi, Jitterbit,
indicators for alerting, as well as message redelivery on exception and Apache Nifi. While most type converters are implicit in most
Table 5
New integration patterns for NFA in the context of system implementations from security to processing without pattern solutions already covered by literature.
Category Scope Pattern name System examples
Security Confidentiality, Privacy Message Encryptor, Message Decryptor, Encrypted Message PGP Encrypt / Decrypt [72], AES_ENCRYPT [74], Encrypt /
Decrypt [82]
Encrypting Endpoint / Adapter FileProcessorConnector [74], FileChannel [81], WSSProvider
[85]
Authenticity, Identity Message Signer, Signature Verifier, Signed / Verified Digest/Hash [73], Signer, Verifier [77]
Message
Storage Encrypted / Encrypting Store DBStorage, Persist [77]
Safe Store Most of the vendors
Redundant Store MorphlineSolrSinks [81], implicit [73,77]
Transmission Encrypted Channel Transmission Protection [75]
Safe, Authenticated Channel Password, certificate, token-based (Most of the vendors)
Audit Log Audit Trails [75,78], Service Auditing [86]
Media Format Type Converter Type Converter [79,80,83]
Encoder, Decoder Base64 Encode / Decode [72,73], Encoder / Decoder [77]
Marshaller, Unmarshaller “Data Format”[83], “ConvertJSONToSQL” [82],
“JsonXMLConverter” [77]
Compress Content, Decompress Content implicit [72,75], Compress Content [82]
Custom Script Script, Processor [83], Expression [74]
Metadata Extractor Read MIME activity [73], ExtractImageMetadata,
ExtractMediaMetadata [82]
Unstructured Image Resizer Image Resizer [82]
Processing Synch / Streaming – Streaming transformations [75], partially [77], streaming
[83]
Parallel Parallel Multicast, Sequential Multicast [76,77,86]
Join Router implicit [83], join [77]
Other Delegate Process Call [77], Direct-VM [83]
Loop Loop Activity [73], Looping [76]
Find and Replace Search/Replace [72], Control Character Replacer [75], Scan
Content [82]
Content Sort Sort [83]
systems, marshalling support is made excplicit, e. g., in Jitterbit, of columns Category and Scope. The coverage of system implemen-
SAP Cloud Integration, and Apache Nifi. More “exotic” functions tations reflected by column System Examples was chosen in order
are text sentiment analysis in Cloudpipes, an Archiving activity in to provide pattern definitions referring to examples (but not all) of
IBM, and a Yandex language translator in Apache Nifi. the corresponding system implementations (if at least one vendor
supported them). Subsequently, we refer only to the categories that
4. Design of pattern catalog have a special relevance for the comprehensiveness of our analysis.
This section summarizes the findings of the literature and sys- Security. Take, for example, NFA Security in combination with
tem reviews in form of a pattern catalog, capturing and describing scope Confentiality (cf. Table 5), for which no comprehensive
the found ad-hoc solutions and functionalities as new patterns. pattern is provided in the literature on the one side, but system
These patterns can be seen as the starting point of a continuation implementations by, for example, Dell Boomi [72], Informatica
of the EIP, but also recent trends to express domain knowledge [74], or Apache Nifi [82] exist. Addressing design goals 1) and 2)
as patterns [95]. In doing so, hypothesis H3 “Some trends are led to the set of suggested patterns Message Encryptor, Message
handled in an (yet) immature and ad-hoc fashion” is targeted. The Decryptor and Encrypting Endpoint. Message Encryptor, for exam-
design goals for the pattern catalog are: ple, covers the system implementations PGP Encrypt / Decrypt,
AES_ENCRYPT, and Encrypt / Decrypt.
1. Comprehensiveness, i. e., coverage of system implementations
that are not in the literature Media. Besides formatting patterns for structured message con-
2. Novelty, i. e., literature coverage of the missing or only partial tent, the media specific patterns for unstructured content are
pattern definitions for NFA under-represented in current system implementations, since there
is only one pattern with direct relation to multimedia processing
The proposed pattern catalog is summarized in Tables 5 and 6 (e. g., Image Resizer [82]). Although there are functionalities for
categorizing the patterns by NFA as Category. The patterns in col- the work on the structured multimedia metadata (e. g., Metadata
umn Pattern Name are further grouped by sub-categories as Scope. Extractor), further research should target the unstructured multi-
Due to lack of space, the descriptions of all patterns contained in media data and processing (e. g., in the context of synchronous,
the catalog are provided as supplementary material [44] and two streaming protocols).
of them are introduced in detail in Section 5. While in this section
we focus on patterns, the supplementary material illustrates the Summary – Comprehensiveness. With the pattern catalog we ad-
modeling of the new patterns for two integration scenarios from dress 94.74% of the NFA scopes or subcategories (i. e., all but 1 out
the quantitative analysis in Section 6. of 19) derived from system implementations. A synch / streaming
pattern elicitation – as also mentioned by [5] – was not conducted
4.1. Goal 1 (comprehensiveness): system implementation coverage in the context of this work, since the system review did not lead
to pattern changes or new patterns, but only adds an additional
In detail, comprehensiveness is evaluated by comparing the processing style. However, we consider this an interesting topic
coverage of patterns with the NFA that are not or only partly cov- especially in the context of the Media and Big Data trends, and
ered by patterns in the literature, represented by the combination propose a separate study for this current gap.
Table 6
New integration patterns for NFA in the context of system implementations from conversations to composition without pattern solutions already
covered by literature.
Category Scope Pattern Name System Examples
Conversations Endpoint Commutative Receiver –

Timed Redelivery until Acknowledge –
Fault tolerant Timeout synchronous request –
Failover Request Handler Failover Client [81]
Resources Request Collapsing –
Request Partitioning –
Monitoring Processing Message Cancellation [76,82]
Usage Statistics [77,78]
Immediate Insights Raise Indicator [72,75–77]
Detect [76,82]
Message Interceptor [81,83], implicit [77]
Monitors Component Monitor [77,84]
Channel Monitor [77,78,80,84,85]
Message Monitor [77–79]
Resource Monitor [77,85]
Circuit Breaker [83]
Hybrid Monitor [77]
Storage Data, Variable Data Store [73,75,77]
Store Accessor DB Update [75], DBStorage [77]
Transient Store Add to Cache [72], Shared Variables [85], Global Variables [75]
Security Key Store, Trust Store, Secure Store [77,81,83]
Composition Integration Subprocess [72,73,76,77]
Integration Process Template Template Integration Process [73], Snapshot [75], Blueprint [84]
4.2. Goal 2 (novelty): literature coverage multimedia [6], synch / streaming [5], conversations [5,6], mon-
itoring [45,61,62], and pattern composition (from system review
Now, we set the pattern findings from the literature review for Section 3, [5]). In addition, the system review raises a demand for
the NFA – summarized in Table 1 – into context with the new pat- storage patterns that was not mentioned in the literature.
tern proposals derived from the system analysis. We exclude the
solutions from the original EIP [3], and Media, Synch / Streaming
and Composition (not in NFA, however, came up during system
analysis and mentioned in [5]), for which no solutions were found 4.3. Solutions for future challenges
in the literature. In addition, we excluded Error Handling, since
it is comprehensively covered from a pattern perspective and We propose several new conversation patterns, of which none
compared to system implementations in prior work [93,94] (not was found in the system implementations. The proposed endpoint-
found in the literature review). specific patterns Commutative Receiver and Timed Redelivery Until
Acknowledgement (similar to the Contingent requests pattern in
[37,55]) – that together denote a solution for a critical trade-off for
Security. Although some security patterns were proposed in the
scalability inspired by [95] – are discussed in detail in Section 5.
SOA domain [30,31], they only provide partial solutions with
The other patterns are further discussed in the supplementary
respect to the NFA and no solution in the context of the system
material [44], and target the additional conversation scopes: Fault
review.
tolerance and Resources. The multi-tenant processing, conversation
patterns (e. g., Cross Scenario and Cross Tenant) patterns that are
Conversations, Processing. In terms of conversation patterns, the mostly required in hybrid and cloud computing setups, are already
system implementations only showed basic support (cf. Table 6), covered by prior work [89], thus not shown.
however, some more can be found in the literature, showing Toward a more stable system, the Timeout Synchronous Re-
that this is an area for integration systems to add more features. quest and Failover Request Handler patterns are improving the
Although, Barros et al. [37] mostly reiterate the original EIP, fault-tolerance of the messaging. Especially in the Big Data con-
there are few patterns that are new in the context of the system text, the resources of an integration scenario or platform become
implementations. In the category of multi-lateral communication, crucial for their stability. Therefore, the Request Collapsing pattern
the One-from-many pattern [37] is a special case of our more reduces the number of requests within a conversation. In addition,
general Join Router that we found in the system implementations the Request Caching reduces the amount of duplicate requests to
(e. g., Apache Camel, SAP Cloud Integration). The One-to-many the same endpoint, while Request Partitioning optimizes requests
send pattern [37] is similar to the (parallel) Multicast – found in to endpoints and confines errors to one request aspect. Together
the systems (e. g., Apache Camel, SAP Cloud Integration), however, with other patterns from the literature review and the proposed
some systems have variants that we captured as Sequential Multi- patterns in this work, we see a clear evidence for further research
cast, which routes messages of the same type to multiple receivers required. Since none of the patterns was found during the system
in sequence to guarantee the successful delivery to all recipients. review this indicates potential for current integration system and
application endpoint implementations.
Summary – Novelty. From the functionality required by system The monitoring of integration scenarios reaches from real-time,
implementations, 59 distinct, new patterns are derived that were scenario-specific processing to near-real time monitors. One fur-
not found in the analyzed literature. However, for 5 out of 7 NFA ther challenge – also identified by [45] and partially covered by
(compare to Table 1), the literature indicates missing patterns as the systems with mixed on-premise and cloud integration – is
research challenge (cf. Section 2.3), thus supporting the extension the monitoring across different platforms (e. g., cross-cloud, across
of the integration pattern catalog for security [15,31,45,50,68,69], on-premise and cloud).
5. Example integration pattern realization gained through better scalability and lower latency of the con-
versational approach – by not waiting for the failure of a sent
As example from the catalog, we selected patterns related to message – is contrasted by the more resources overall required in
the non-trivial trade-off between stateful and stateless integration case of many failures.
processes (inspired by cloud computing challenges [95]). Espe-
cially the system review shows that all vendors provide extensive 5.2. Patterns and pattern formats
storage capabilities – beyond the EIP, leading to stateful processes.
However, the under-represented conversation patterns could of- To formalize the new challenges and the resulting, required
fer an alternative, thus allowing for stateless processes. While capabilities within an integration system, thus coming to less
stateless processes have scalability benefits, they come with some immature and ad-hoc solutions (cf. H3), we propose to express
drawbacks that have to be considered. We selected this trade-off them as patterns. Similarly, expert knowledge and best practices
due to its relevance in the context of Big Data, its relevance for were already collected for software design by Gamma et al. [96],
Cloud Computing, and because it addresses one well-represented EIP by Hohpe et al. [3], and recently for cloud computing patterns
(i. e., storage) and the currently under-represented, but important by Fehling et al. [95]. For a suitable pattern representation, we
area of (stateful) conversations. Subsequently, we describe the compare their pattern formats in Table 7, and select common
trade-off as problem description, discuss a suitable pattern format categories for our proposal.
and conclude with two pattern descriptions and their realization. From the analysis of several known pattern formats in Table 7,
we selected: name, icon, driving question, context, solution, re-
5.1. Problem description: stateful vs. stateless integration processes sults, and known uses to round-off the description. We leave out
the separate categories of forces (i. e., problem constraints) and
Operating an integration system requires persistent stores and implementation (i. e., pattern variants), which we include into
queues, e. g., monitoring, key or secure store to achive security, the selected context and known uses categories, respectively. The
or auditing for legal reasons. In addition, transactional message pattern descriptions in the supplementary material [44], add a
processing (e. g., aggregator pattern) as well as message delivery Data Aspects category (not further discussed here), which gives
semantics (e. g., reliable messaging like “exactly-once-in-order”) even more insight into the configuration of the pattern solutions.
[89] require some persistent state. While the system operability
avoids influencing the message processing by not using shared 5.3. Pattern examples and realization
states between integration scenario instances, the transactional
processing and message delivery semantics of the stateful message From the problem description we take three capabilities that
processors (i. e., patterns) usually require shared states. For ex- are required to represent an EOIO, while keeping the integration
ample, when a stateful aggregator – as part of a scenario instance processes stateless. We summarize the capabilities to the following
– processes a sequence of messages, a second scenario instance two patterns, for which we explain the realization: Commutative
could be used to distribute the load. However, in the absence of Receiver and Timed Redelivery until Acknowledge. In addition, we
“process stickiness” (i. e., messages of one sequence are only sent require the Quick Acknowledgement pattern from [55].
to one instance), the stateful aggregator in the second instance
has to be able to complete a sequence the other instance started, Commutative Receiver. The commutative receiver accounts for two
thus shared state. Hence, the shared states imply complex state tasks: message deduplication and out-of-order handling. Therefore,
handling across integration processes in compute clusters or cloud the application’s state is re-used, hence no additional state in the
environments, and this may have a negative impact on their scala- integration process is required.
bility. Alternatively – following the ideas on (stateful) conversation
patterns from Hohpe [55] – some of the discussed messaging
related storage and message delivery semantics could be moved to
“smart” message endpoints (i. e., applications), which already have
a persistent state, thus making the integration processes stateless.
For example, Fig. 6 illustrates the trade-off between Exactly- How to ensure idempotent, in-order message
once In Order (EOIO) delivery semantics within the integration processing without intermediate state in form of persistent integration
scenario (i. e., requires a stateful Message Store, Resequencer and scenarios?
Idempotent Receiver [3], and transactional Message Redelivery (Icon: the icon uses the icon notation from [3], combining the
on Exception [94]) in Fig. 6(a) and as a (stateful) conversational in-order sequencing as well as the idempotent storage.)
approach in Fig. 6(b). The integration processes are represented in Context: Out-of-order communication with end-
BPMN 2.0 similar to [46]. An EOIO delivery requires a transactional points/applications.
redelivery in case of an exception, a message ordering step accord- Solution: Guarantee that endpoint/application handle arriving
ing to a sequence of messages in form of a Resequencer and an out-of-order messages will be stored within their sequence and
Idempotent Receiver, which is able to deduplicate the messages. only then processed, if the sequence is (partially) complete and in
Fig. 6(a) depicts the instance spanning state for the retry and the the correct order.
resequencing. To avoid stateful integration process, both capa- Result: This solution handles out of order messages and applies
bilities can be moved to the endpoints (cf. Fig. 6(b)). While this them in-order within the application endpoint.
will not work for legacy, packaged applications, it results into an Relations to other patterns: This pattern is an extension of the
improved scalability within the integration process and moves the Idempotent Receiver from [3] with additional Message Sequence
resequencing decision to the receiver. To eventually stop sending, handling.
the sender – redelivering the message periodically – requires a Known uses: not found in literature or system review, how-
stop event (i. e., an acknowledgement) from the receiver. ever, Microsoft advices developers to implement commutative
The solution’s trade-off are the several messages that are sent endpoints in the context of micro services4 .
by the receiver until an acknowledgement is received, while being
able to process all messages in parallel using stateless integration 4
Designing Services, visited 02/2017: https://msdn.microsoft.com/en-us/library/
process instances. In other words, the performance improvements ee658114.aspx.
(a) Stateful EOIO (b) Stateless EOIO

Fig. 6. Conversational approach for Exactly Once in Order (EOIO).
Table 7
Common pattern formats: Enterprise Integration Patterns (EIP) [3], Cloud Computing Patterns (CCP)
[95], Design Patterns (DP) [96].
Categories Description EIP CCP DP
Name pattern identifier

Icon visual representation –
Problem / Driving Question / Motivation difficulty as question
Intent statement about design issue – –
Also known as other pattern names – –
Context / Motivation introduces problem domain
Forces, Appilcability problem constraints –
Solution how to solve the problem –
Sketch, Structure illustrate solution –
Participants, Collaborations participants, responsibilities – –
Results / Consequences how to apply the solution
Next / Related Patterns related patterns, differences –
Sidebars / Implementation / Code pattern variations –
Examples / Known Uses real system examples
Timed Redelivery until Acknowledge. The commutative receiver instances of the same message with configurable timings until the
moves the message redelivery on exception from the integration actual receiver endpoint acknowledgements (e. g., Quick Acknowl-
process to the sender application, while conducting an asyn- edgement [55]) reach the sender. Requires an Idempotent [3] or
chronous communication. Hence, no exceptions are propagated Commutative Receiver for certain message delivery semantics [89]
back to the sender, however, the redeliveries are stopped by Result: Send copies of the same message asynchronously until
asynchronously received Acknowledgements from the receiver (via the receiver’s acknowledgement reaches the sender.
the integration system). Until then, the messages are resent with Relations to other patterns: This pattern is an extension of the
increasing delay to reduce the load of duplicate messages. Retry pattern in [55], and related to the Redelivery on Exception
pattern in [94].
Known uses: - (not found in literature or system review).
Solution Summary. As an extension a Timed Redelivery until Ac-

knowledge pattern would be required that makes multiple attempts
How to ensure that a message will be received to deliver a message (potentially with exponential back-off delay).
without intermediate storage, e. g., in form of Redelivery on Exception That might result to duplicate message instances, sent to the
[94]? receiver. Assuming a stateless integration process, an idempotent
(Icon: the icon uses the icon notation from [3], combining receiver [3] is required to detect and handle the duplicates. The
delayed message send with asynchronous reception of acknowl- sketched conversation works fine for exactly-once processing
edgments using a transactional store.) semantics [89]. However, for ensuring in-order message processing
Context: This pattern is used for asynchronous communication (e. g., create sales order, before update), it would not be sufficient.
with message delivery guarantees A stateless integration process cannot reliably re-order the in-
Solution: Instead of relying on intermediate storage and retry coming messages, delegating this task to the receiver application.
within the integration system, the application sends multiple The receiving application has to handle incoming messages in an
associative and commutative way (e. g., handle update, before cre-
ate). An implementation of this Commutative Receiver pattern can SAP Financial Services Network (FSN). In contrast to C4C, FSN [98],
be found in the microservice context (e. g., service applications). abbreviated fsn, is a cloud-based, business network that connects
When the receiver got all messages belonging to one message banks and other financial institutes with their corporate customers
sequence (i. e., without duplicates), it sends an Acknowledgement (e. g., for payments, bank account management). The integration
message that is asynchronously processed by the sender, which style is mostly process invocation [2]. Besides reliability, the major
stops redelivering corresponding messages immediately. focus lies on secure message exchange.
6. Quantitative analysis SAP Cloud Integration eDocument/Electronic Invoicing (eDocument).

The SAP Cloud Integration eDocument/Electronic Invoicing is a
In this section we conduct a quantitative analysis of integration solution for country-specific electronic document management
scenarios to study the usage of original EIP and the new patterns [97]. The edocuments solution helps cooperations to interact with
from the pattern catalog in Section 4. We selected the SAP Cloud legal authorities (e. g., implement the new EU Data Protection
Integration system from the review (cf. Section 3), for which we Regulation5 ).
found “real-world” examples of all scenarios from Fig. 2. With
over 1,0 0 0 customers and several hundred integration scenarios SAP Predictive Maintenance and Service (PdMS). In PdMS
delivered as standard content SAP Cloud Integration represents [97], machine data is collected using an Internet of Things (IoT)
a cloud integration system for application and data integration. platform and enriched with business information coming from,
The analysis targets the hypotheses “The original EIP of the 2004 e. g., SAP Business Suite. This allows real time monitoring of the
book are all widely used in praxis” and “H4: Solutions not in EIP machine that triggers alerts resulting in service tickets in SAP CRM
can be found in real-world integration scenarios for the trends”. or ERP. Based on that any unusual trends or behavior becomes
Therefore, we firstly select several scenarios along the identified visible and appropriate action, potentially avoiding unnecessary
trends and briefly describe them. Secondly, we evaluate found service costs, can be taken.
original EIP and new solutions.
6.2. Scenario analysis
6.1. Integration scenarios
For the analysis of the SAP delivered standard content in this
The new trends – set into context denoted by edges via paper, we prototypically implemented data discovery and mining
the integration system node in Fig. 2 – can be summarized to capabilities into the SAP Cloud Integration system, which identified
integration the scenario domains: a total of 154 distinct integration scenarios (c4c-erp: 42, c4c-crm:
37, fsn: 56, edocument: 13, pdms: 6).
• On-Premise to Cloud: Most organizations productively run on The total number of patterns for all scenarios is 1501 (w/o
packaged, on-premise applications. They need to connect these adapters, endpoints). For the more “classical” integration scenarios
applications with cloud applications for various reasons, e. g., in c4c-erp and c4c-crm nearly all integration patterns could be
extensions for legal reasons or new functionality, to connect covered by original EIP from [3] (apart from second level con-
with business partners. This integration domain is called hybrid figuration on monitoring and exception handling). For the cloud
integration [4]. integration scenarios fsn and eDocument as well as for the pdms
• Cloud to Cloud or Business Network (including social): Native IoT scenario, 466 new requirements (and 597 complementing,
cloud applications or cloud integrated on-premise applications second level configurations) were needed in total, out of which
interact with business partners in business networks (e. g., 66% are covered by existing EIP (i. e., 1025 capabilities in total).
payment, financial, supplier relationships), connect to social This means that for these integration scenarios approximately 13
networks (e. g., social marketing) or consume cloud services of the required patterns are not covered by the original EIP.
(e. g., machine learning).
• Device to Cloud (including mobile and personal computing): Pattern Solutions covered by the EIP. The distribution of patterns
What starts with business applications on “bring your own de- covered by original EIP is depicted in Fig. 7. Not all EIP were
vice” for mobility, extends to intelligence brought to machines required in the scenarios of the integration solutions, however,
(e. g., sensors and actuators in smart logistics or production) nearly all of them facilitate the tasks of (i) Message Construction:
and eventually comes down to sensors and devices for personal solutions Document Message and Request-Reply; (ii) Messaging
computing. Channels: solution P2P Channel; (iii) Message Routing: solutions
Content-based Router, Splitter, and Aggregator; (iv) Message Trans-
We left out the conventional On-Premise to On-Premise appli-
formation: solutions Content Enricher, Content Filter, and Message
cation integration and other variations due to our focus on new
Translator.
trends. For the quantitative analysis, we selected one application
integration solution for each of the new scenario domains, and we
New Pattern Solutions. Additional pattern solutions are covered by
added one for cloud to cloud and business networks due to the
integration capabilities, depicted in Fig. 8. We grouped the new so-
slight focus on these domains. The solutions can be visited as SAP
lutions according to the pattern catalog from Section 4 and added
Cloud Integration standard content [97].
the corresponding pattern proposals for each of them. While
approximately half of the new patterns from the catalog are used
SAP Cloud for Customer (C4C). The C4C solutions for the com-
in the real-world scenarios, the new conversation patterns, are not
munication with on-premise Enterprise Resource Planning (ERP)
yet used in any of the scenarios. For example, the confidentiality
and Customer Relationship Management (CRM) applications, ab-
requirements covered message confidentiality or privacy patterns
breviated c4c-erp and c4c-crm, can be considered typical hybrid
(Message Encryptor, Message Decryptor, Encrypted Message, En-
corporate to cloud application integration [97]. The dominant
crypting Endpoint, Encrypting Adapter) are called Msg. Privacy, and
integration styles – according to the classification in [2] – are
the authenticity and integrity requirements (Message Signer, Sig-
process invocation and data movement. The state changes (e. g.,
nature Verifier, Signed / Verified Message) are summarized to Msg.
create, update) of business objects (e. g., business partner, oppor-
tunity, activity) as well as master data in the cloud or corporate
applications (e. g., ERP, CRM) are exchanged with each other. 5
EU – General Data Protection Regulation: http://goo.gl/Ru0slz.
Fig. 7. Scenarios using original EIP.
Auth.. Thereby the message confidentiality is exclusively required guaranteed rollback for all branches in case of an error and as
for the communication within the FSN business network, while Parallel Multicast in PdMS for parallel processing purpose.
message authenticity is also used for exchanging eDocuments with This behaviour is complemented by a Stop All setting in the
the legal authorities and for PdMS. FSN, PdMS and partially CRM scenarios, consequently stopping
In the category multimedia, currently no real media format the message processing in all parallel instances of the integration
handlers (Type Converter, Encoder, Decoder) are used (e. g., im- scenario. Another error handling functionality is the Rethrow,
age, video). However, we grouped the used functionality into which allows to (re-) throw exceptions. The Rethrow is mainly
three patterns. The Encoder and Decoder patterns represent the used in eDocument, FSN, PdMS and CRM scenarios to indicate
handling of binary message content, exclusively used in FSN and that a situation is still unresolved. Especially in FSN, PdMS and
eDocument scenarios. This is due to the various communication eDocument scenarios, it is important to inform a business expert
partners, using different encodings, as well as third party services or administrator about erroneous situations. The Raise Indicator is
(e. g., financial document mapping engines), which requires special used for this purpose. To prevent from uncontrollable behaviour
encodings. The Custom Script allows to add versatile User-defined and to customize the information returned in case of an error in
Functions, which are mostly used as auxiliary in FSN scenarios. The synchronous scenarios, a Catch-all exception process (with several
compression algorithms (Compress Content, Decompress Content), steps) is used in FSN and eDocument scenarios.
which would be immensely relevant in real multimedia scenarios, To remember parts of a message or additional information
are used in FSN scenarios due to larger messages sizes (e. g., generated by adapters or message processors, a Transient Store
aggregated FSN payment details). Finally, marshalling (Marshaller, (cf. [56]) is used in FSN. The Store Accessor, used by FSN and
Unmarshaller) support is required in FSN scenarios, since some eDocument, is mostly used for stateful pattern compositions and
communication partners require JSON to XML conversion and vice for legal reasons (Data Store, Audit Log). Especially in FSN, most
versa. of the message stores are encrypting (Encrypting Store), which
The processing of messages is mostly done by moving mes- means that the messages are stored confidentially.
sages through the pipeline. However, especially the FSN and CRM The composition in terms of the Integration Subprocess pat-
integration require streaming and parallel processing. This is again tern (excluding the exception sub-processes) indicates complex
due to scenarios with larger messages to be processed (e. g., processing logic, which can mostly be found in FSN, PdMS and
IDoc segments in CRM) and stream-based splitting in PdMS. The eDocument scenarios. Sometimes composition is used for re-usable
Multicast pattern is used as Sequential Multicast in FSN to allow process parts.
Fig. 8. New capability categorization.
In summary, the analysis shows the need for new patterns and to these new trends and NFA where the systems provide a more
solutions as seen in the system review. While the hybrid integra- comprehensive support. Solutions mentioned in the literature
tion scenarios simply extend the on-premise integration into the comprise patterns, formalization, and modeling, covering the spec-
cloud relying on transport level security, all other new integration trum from a more abstract description as patterns (cf. Section 5)
scenario domains require more on security and control over the to the implementation and execution as well as towards the user’s
message processing in form of error handling. This becomes more point of view. For this reason, patterns are regarded as the “glue”
obvious, the more exclusively the integration scenarios are running for which a comprehensive description is required first. Hence,
in the cloud. For example, the small amount of device integration this work (together with the supplementary material [44]) aimed
scenarios still relies on integration logic on the devices, thus show at filling the gaps in existing patterns for new integration trends
only few security, error handling and processing capabilities. The and NFA (cf. Section 5): security, (ideas on) conversation, mon-
scenarios are less complex – compared to the cloud to cloud cases, itoring, storage. Nonetheless, the different reviews and analyses
hence require limited composition options. Furthermore, with conducted in this work indicate open research challenges. These
increasing cloud focus, the trade-off between more complex, but are subsequently summarized and discussed.
expressive, stateful and simpler, better scalable, stateless message
processing seems to lean towards the interaction with storage cur- 7.1. Research challenges
rently. The conversation patterns – including stateful conversations
– are still mostly unexplored. The same is true for the increasingly The literature review shows that for the trends and NFA differ-
important topic of multimedia processing, which will give a new ent solutions have been proposed, mainly patterns, formalization,
edge to the variety and interoperability problems in EAI [2]. and modeling.
7.1.1. Patterns
7. Challenges, limitations, impact Patterns are the predominant solution proposed in literature
(cf. Table 1). Moreover, this work has closed gaps by providing pat-
The literature (cf. [5]) and the quantitative analysis of real- terns for NFA not present so far. Still pattern descriptions would be
world integration scenarios (cf. Fig. 7) show that some of the necessary in the context of the following trends and NFA. At first,
enterprise integration patterns (EIP) described by Hohpe and multimedia functions are under-represented. Due to the access to
Woolf [3] in 2004 are still widely used, thus denote best-practices the end user, multimedia becomes more and more interesting for
in the area of application integration. This supports the assumption all kinds of applications (e. g., sentiment analysis, monitoring in
of the EIP authors that the patterns are still practically relevant different domains like medicine or agriculture). For application
[5]. Literature and system reviews also reveal that since 2005 integration, this targets the volume, variety and interoperabil-
several new trends and non-functional aspects (NFA) for enterprise ity problems. The resulting increase of heterogeneity of media
application integration have appeared that are not covered by the formats and communication partners (e. g., cloud applications,
EIP from 2004. Literature as well as systems partly offer solutions mobile devices, camera phones) demands for revisiting the EIP in
the context of multimedia operations and their semantic aspects. e. g., allowing to “query by sketch” conditions, for integration sce-
Consequently, the increasing message sizes require the evaluation narios would provide a more adequate, non-textual configuration.
of optimization techniques (e. g., message indexing), and more ef- In addition, editors and visual data science (incl. machine learning)
ficient processing styles like streaming, which the EIP authors also tools for scenario-based, runtime monitoring, which are capable of
acknowledge [5], or data-aware integration pattern solutions (e. g., dealing with large amounts of data, could lead to smarter (cross-)
[91]). In general, to address the Big Data challenges of volume integration platform administration of integration scenarios. In
and velocity requires corresponding benchmarks for pattern (e. g., this context, the development of visual modeling notations, new
EIPBench [99]) as well as for end-to-end system implementations, editors together with extensive user studies become necessary.
which are currently missing. As additional NFA, only few of the
conversation patterns are supported. For instance, Section 5 showed 7.2. Limitations
that conversation patterns can provide alternatives to improve the
current processing and might become useful in more complex Limitations of the work concern the literature and the system
application or device interactions (e. g., device mesh [6]). The review. For both the searches were led by the selection of key-
monitoring of integration scenarios across multiple platforms (e. g., words and criteria due to the vast amount of existing work and
mobile, on-premise, cloud) – including aspects like raising indica- in order to not loose focus of this study. Nonetheless, conducting
tors in case of an event – remains a challenge. This also hints on further vertical searches and expert additions that were not found
further work required for Mobile Computing and Internet of Things, based on the keywords could be included in the analysis. The
e. g., standardized protocols, conversation or interaction patterns selection of representative systems is supposed to reflect the
(incl. data collection, device reconfiguration), energy efficiency. Fi- current system offering. Different types of systems (open source,
nally, as new trends and NFA might constantly arise, their analysis commercial, and startup) were considered. In summary, both re-
with respect to pattern support becomes a continuous task. views were envisioned to be comprehensive, but not complete. The
quantitative analysis aims at covering a broad range of applications
7.1.2. Formalization based on five use cases. One might argue that the use cases are
Starting from the pattern view, formalization is an important all provided by the same organization. However, this provides the
step to precisely specify the semantics of the pattern realizations, chance to analyze real-world scenarios instead of toy examples.
i. e., formalization constitutes an important step towards the
implementation and execution of the patterns in integration sce- 7.3. Impact
narios. As shown in Table 1 formalization approaches have been
predominantly proposed in the context of service oriented archi- The impact of a continuous analysis of integration trends
tectures (SOA) for validating and optimizing compositions by, for and NFA on research and practice is enormous. The impact on
example, mapping them to Petri nets. Notably, a more formal def- research is reflected in the open research challenges stated in
inition of integration pattern composition (also suggested by the Section 7.1. In order to address these challenges, a plethora of
EIP authors [5]) is required not only for structural validations using new approaches is necessary. The importance of the topic from
Petri nets – as in the literature review, but also semantic, runtime a practical perspective is already paramount as the system and
validations and optimizations on static scenario as well as dy- scenario analyses of this paper show. Facing new trends that often
namic, workload data is missing. First work on the latter was con- stem from practice will perpetuate the importance of this work
ducted by [100], however, has to be revisited in the context of the in the future. Putting the focus on the human aspect in addition
trends and NFA as well as new technical capabilities (e. g., machine to a more technical treatment of the topic will also lead to a
learning of / for workload patterns, routing conditions, condition multitude of new research questions and practical implications.
orderings). Furthermore, cloud, mobile and device computing raise While the original EIP from 2004 are still relevant for many of the
new questions about optimal runtime placements of integration new trends in 2016 and beyond, new capabilities are required to
processes. In general, there is still an enormous potential for elab- address requirements (e. g., non-functional aspects) resulting from
orating formalizations for both, trends and NFA, specifically, as a these trends (cf. hypothesis H1).
more or less comprehensive set of patterns has been proposed by
now. A follow-up research question is how to implement patterns References
and pattern compositions in solutions using formal models.
[1] S. Conrad, W. Hasselbring, A. Koschel, R. Tritsch, Enterprise application inte-
gration, Spektrum Akademischer Verlag, 2005.
[2] D.S. Linthicum, Enterprise application integration, Addison-Wesley Longman
7.1.3. Modeling Ltd., 20 0 0.
Except few works in the SOA domain providing modeling [3] G. Hohpe, B. Woolf, Enterprise integration patterns: designing, building, and
support for compositions, no attention has been paid to model deploying messaging solutions, Addison-Wesley Professional, 2004.
[4] Forrester ResearchInc., The Forrester Wave: Hybrid Integration For En-
integration-specific aspects so far. For compositions, business pro- terprises, Q4 2016, 2016. (https://www.forrester.com/report/The+Forrester+
cess modeling notations such as business process model and nota- Wave+Hybrid+Integration+For+Enterprises+Q4+2016/- /E- RES131101).
tion (BPMN) can be used, however, the integration-specific aspects [5] O. Zimmermann, C. Pautasso, G. Hohpe, B. Woolf, A decade of enterprise in-
tegration patterns: a conversation with the authors, IEEE Softw. 33 (1) (2016)
exceed the modeling capabilities. However, conveying information 13–19.
on the integration scenarios to users is of utmost importance for, [6] GartnerInc., Gartner Newsroom Emerging Technologies from 2005 to 2017,
e. g., maintenance and adaptations of these scenarios. Also here 2017. (http://www.gartner.com/newsroom/id/492152, http://www.gartner.
com/newsroom/id/495475, http://www.slideshare.net/dinhledat/dinh-ledat-
patterns might help to form the basis for different modeling and
top- 10- technology- trends- 20072014- gartner, http://www.gartner.com/
visualization proposals. It could be envisioned to base integration newsroom/id/530109, http://www.gartner.com/newsroom/id/777212,
flows on existing business process modeling languages (e. g., http://www.gartner.com/newsroom/id/1210613, http://www.gartner.com/
newsroom/id/1454221, http://www.gartner.com/newsroom/id/1826214,
BPMN) in order to keep the mental map of users, but to enhance
http://www.gartner.com/newsroom/id/2209615, http://www.gartner.com/
them with integration-specific icons. In [44], a first idea of equip- newsroom/id/2603623, http://www.gartner.com/newsroom/id/2867917,
ping BPMN with integration icons is depicted for the SAP Cloud In- http://www.gartner.com/newsroom/id/3143521, http://www.gartner.com/
tegration eDocument use case (due to lack of space we refer to the newsroom/id/3482617).
[7] Forrester ResearchInc., The Top 10 Technology Trends To Watch: 2016 To 2018,
technical report). In general, NFA like security and multimedia have 2016. (https://www.forrester.com/report/The+Top+10+Technology+Trends+To+
to be further analyzed. Therefore, new visual configuration editors, Watch+2016+To+2018/- /E- RES120075).
[8] R. Wieringa, Design science methodology for information systems and soft- [39] F.B. Vernadat, Interoperable enterprise systems: principles, concepts, and
ware engineering, Springer, 2014. methods, Annu. Rev. Control 31 (1) (2007) 137–145.
[9] B. Kitchenham, Procedures for performing systematic reviews, Technical Re- [40] G. Grossmann, M. Schrefl, M. Stumptner, Exploiting semantics of inter-process
port TR/SE-0401, Keele University, Keele, UK, 2004. dependencies to instantiate predefined integration patterns, in: Proc. of the
[10] C. Hentrich, U. Zdun, Patterns for business object model integration in pro- 26th international conference on Conceptual modeling, 2007, pp. 155–160.
cess-driven and service-oriented architectures, in: Proceedings of the 2006 [41] H. Taylor, A. Yochem, L. Phillips, F. Martinez, Event-driven architecture: how
conference on Pattern languages of programs, 2006, p. 23. SOA enables the real-time enterprise, Pearson Education, 2009.
[11] C. Hentrich, U. Zdun, Service integration patterns for invoking services from [42] O.P. Patri, V.S. Sorathia, A.V. Panangadan, V.K. Prasanna, The process-oriented
business processes., in: EuroPLoP, 2007, pp. 235–278. event model (PoEM): a conceptual model for industrial events, in: Interna-
[12] C. Hentrich, U. Zdun, A pattern language for process execution and integra- tional Conference on Distributed Event-Based Systems, 2014, pp. 154–165.
tion design in service-oriented architectures, in: Transactions on Pattern Lan- [43] S. Asmus, A. Fattah, C. Pavlovski, Enterprise cloud deployment: integration
guages of Programming I, Springer, 2009, pp. 136–191. patterns and assessment model, IEEE Cloud Comput. 3 (1) (2016) 32–41.
[13] U. Zdun, C. Hentrich, W.M. Van Der Aalst, A survey of patterns for service-ori- [44] D. Ritter, S. Rinderle-Ma, Toward a collection of cloud integration patterns,
ented architectures, Int. J. Internet Protoc. Technol. 1 (3) (2006) 132–143. arXiv preprint arXiv:1511.09250 (2015).
[14] U. Zdun, Pattern-based design of a service-oriented middleware for remote [45] D. Merkel, F. Santas, A. Heberle, T. Ploom, Cloud integration patterns,
object federations, ACM Trans. Internet Technol. (TOIT) 8 (3) (2008) 15. in: European Conference on Service-Oriented and Cloud Computing, 2015,
[15] M. Autili, A. Di Salle, A. Perucci, M. Tivoli, On the automated synthesis of pp. 199–213.
enterprise integration patterns to adapt choreography-based distributed sys- [46] D. Ritter, Experiences with business process model and notation for modeling
tems, ArXiv e-prints (2015). integration patterns, in: European Conference on Modelling Foundations and
[16] V. Gacitua-Decar, C. Pahl, Ontology-based patterns for the integration of busi- Applications, 2014, pp. 254–266.
ness processes and enterprise application architectures, Semantic Enterprise [47] D. Ritter, Towards more data-aware application integration (extended ver-
Application Integration for Business Processes: Service-Oriented Frameworks, sion), CoRR abs/1504.05707 (2015).
IGI Publishers, Hershey, PA (2009) 36–60. [48] D. Mansor, Moving to the cloud: patterns, integration challenges and oppor-
[17] M. Heller, M. Allgaier, Model-based service integration for extensible enter- tunities, in: Proceedings of the 7th International Conference on Advances in
prise systems with adaptation patterns, in: e-Business (ICE-B), Proceedings of Mobile Computing and Multimedia, 2009. 9–9.
the 2010 International Conference on, 2010, pp. 1–6. [49] H. Buckow, H.-J. Groß, G. Piller, K. Prott, J. Willkomm, A. Zimmermann, In-
[18] E. Kaneshima, R.T.V. Braga, Patterns for enterprise application integration, in: tegration strategies and patterns for SOA and standard platforms, in: GI
Proceedings of the 9th Latin-American Conference on Pattern Languages of Jahrestagung (1), 2010, pp. 398–403.
Programming, 2012, p. 2. [50] M. Heiss, A. Oertl, M. Sturm, P. Palensky, S. Vielguth, F. Nadler, Platforms for
[19] K. Umapathy, S. Purao, Designing enterprise solutions with web services and industrial cyber-physical systems integration: contradicting requirements as
integration patterns, in: IEEE International Conference on Services Computing drivers for innovation, in: Modeling and Simulation of Cyber-Physical Energy
(SCC’06), 2006, pp. 111–118. Systems, 2015, pp. 1–8.
[20] Y. Zheng, H. Cai, L. Jiang, Application integration patterns based on open re- [51] D. Ritter, J. Bross, Datalogblocks: relational logic integration patterns, in: In-
source-based integrated process platform, in: International Conference on In- ternational Conference on Database and Expert Systems Applications, 2014,
formation Computing and Applications, 2011, pp. 577–584. pp. 318–325.
[21] C. Gierds, A.J. Mooij, K. Wolf, Reducing adapter synthesis to controller syn- [52] T. Scheibler, F. Leymann, A framework for executable enterprise applica-
thesis, IEEE Trans. Serv. Comput. 5 (1) (2012) 72–85. tion integration patterns, in: Enterprise Interoperability III, Springer, 2008,
[22] R. Seguel, R. Eshuis, P. Grefen, An Overview on Protocol Adaptors for Service pp. 485–497.
Component Integration, Technical Report, Technische Universiteit Eindhoven, [53] T. Scheibler, F. Leymann, Realizing enterprise integration patterns in Web-
2008. Sphere, Technical Report, University of Stuttgart, 2005.
[23] V.N. Gudivada, J. Nandigam, Enterprise application integration using ex- [54] R. Thullner, A. Schatten, J. Schiefer, Implementing enterprise integration pat-
tensible web services, in: IEEE International Conference on Web Services terns using open source frameworks, Softw. Eng. Tech. Prog. (2008) 111–124.
(ICWS’05), 2005, pp. 41–48. [55] G. Hohpe, Conversation patterns, in: The Role of Business Processes in Service
[24] W. Deng, X. Yang, H. Zhao, D. Lei, H. Li, Study on EAI based on web services Oriented Architectures, in: number 06291 in Dagstuhl Seminar Proceedings,
and SOA, in: International Symposium on Electronic Commerce and Security, 2006, p. 7.
2008, pp. 95–98. [56] L. González, R. Ruggia, Addressing QoS issues in service based systems
[25] C. Mauro, J.M. Leimeister, H. Krcmar, Service oriented device integration-an through an adaptive ESB infrastructure, in: Proceedings of the 6th Workshop
analysis of SOA design patterns, in: 43rd Hawaii International Conference on on Middleware for Service Oriented Computing, 2011, p. 4.
System Sciences (HICSS), 2010, pp. 1–10. [57] D. Fahland, C. Gierds, Using Petri nets for modeling enterprise integration pat-
[26] Y. Liu, X. Liang, L. Xu, M. Staples, L. Zhu, Using architecture integration patterns, Technical Report, bpmcenter.org, 2012.
terns to compose enterprise mashups, in: Software Architecture & European [58] D. Fahland, C. Gierds, Analyzing and completing middleware designs for en-
Conference on Software Architecture, 2009, pp. 111–120. terprise integration using coloured petri nets, in: International Conference on
[27] Y. Liu, X. Liang, L. Xu, M. Staples, L. Zhu, Composing enterprise mashup com- Advanced Information Systems Engineering, 2013, pp. 400–416.
ponents and services using architecture integration patterns, J. Syst. Softw. 84 [59] O.P. Patri, A.V. Panangadan, V.S. Sorathia, V.K. Prasanna, Semantic manage-
(9) (2011) 1436–1446. ment of enterprise integration patterns: a use case in smart grids, in: Data
[28] D. Braga, S. Ceri, F. Daniel, D. Martinenghi, Mashing up search services, IEEE Engineering Workshops (ICDEW), 2014, pp. 50–55.
Internet Comput. 12 (5) (2008) 16–23. [60] S. Basu, T. Bultan, Automatic verification of interactions in asynchronous sys-
[29] S. Cetin, N.I. Altintas, H. Oguztüzün, A.H. Dogru, O. Tufekci, S. Suloglu, A tems with unbounded buffers, in: International conference on Automated
mashup-based strategy for migration to service-oriented computing., in: In- software engineering, 2014, pp. 743–754.
ternational Conference on Pervasive Service, 2007, pp. 169–172. [61] P. Mederly, M. Lekavỳ, M. Závodskỳ, P. Návrat, Construction of messag-
[30] X. Qu, X. Yang, J. Zhong, X. Lv, Integration patterns of grid security service, ing-based enterprise integration solutions using AI planning, in: IFIP Cen-
in: Sixth International Conference on Parallel and Distributed Computing Ap- tral and East European Conference on Software Engineering Techniques, 2009,
plications and Technologies (PDCAT’05), 2005, pp. 524–528. pp. 16–29.
[31] D. Shah, D. Patel, Dynamic and ubiquitous security architecture for global [62] P. Mederly, P. Návrat, Construction of messaging-based integration solutions
SOA, in: The Second International Conference on Mobile Ubiquitous Comput- using constraint programming, in: East European Conference on Advances in
ing, Systems, Services and Technologies, UBICOMM’08., 2008, pp. 482–487. Databases and Information Systems, 2010, pp. 579–582.
[32] M. Fisher, S. Sharma, R. Lai, L. Moroney, Java EE and. net interoperability: [63] R. Kazman, K. Schmid, C.B. Nielsen, J. Klein, Understanding patterns for
integration strategies, patterns, and best practices, Prentice Hall Professional, system of systems integration, in: System of Systems Engineering, 2013,
2006. pp. 141–146.
[33] C. Ouyang, E. Verbeek, W.M. Van Der Aalst, S. Breutel, M. Dumas, A.H. Ter [64] R. Land, I. Crnkovic, S. Larsson, Process patterns for software systems in-house
Hofstede, Formal semantics and analysis of control flow in WS-BPEL, Sci. integration and merge-experiences from industry, in: Conference on Software
Comput. Program 67 (2) (2007) 162–198. Engineering and Advanced Applications, 2005, pp. 180–187.
[34] N. Lohmann, P. Massuthe, C. Stahl, D. Weinberg, Analyzing interacting [65] D. Chen, G. Doumeingts, F. Vernadat, Architectures for enterprise integration
WS-BPEL processes using flexible model generation, Data Knowl. Eng. 64 (1) and interoperability: past, present and future, Comput. Ind. 59 (7) (2008)
(2008) 38–54. 647–659.
[35] A. Kumar, Z. Shan, Algorithms based on pattern analysis for verification and [66] H. Panetto, R. Jardim-Goncalves, A. Molina, Enterprise integration and net-
adapter creation for business process composition, in: OTM Confederated In- working: theory and practice, Annu. Rev., Control 36 (2) (2012) 284–290.
ternational Conferences, 2008, pp. 120–138. [67] K. Wang, M. Dumas, C. Ouyang, J. Vayssière, The service adaptation machine,
[36] J.-m. Jiang, S. Zhang, P. Gong, Z. Hong, Message dependency-based adapta- in: European Conference on Web Services, 2008, pp. 145–154.
tion of services, in: IEEE Asia-Pacific Services Computing Conference (APSCC), [68] S. Rajam, R. Cortez, A. Vazhenin, S. Bhalla, Design patterns in enterprise ap-
2011, pp. 442–449. plication integration for e-learning arena, in: International Conference on Hu-
[37] A. Barros, M. Dumas, A.H. Ter Hofstede, Service interaction patterns, in: Inter- mans and Computers, 2010, pp. 81–88.
national Conference on Business Process Management, 2005, pp. 302–318. [69] W. He, L. Da Xu, Integration of distributed enterprise applications: a survey,
[38] T. Köllmann, C. Hentrich, Synchronization patterns for process-driven and ser- IEEE Trans. Ind. Inf. 10 (1) (2014) 35–42.
vice-oriented architectures., in: EuroPLoP, 2006, pp. 199–228. [70] G. Hohpe, Your coffee shop doesn’t use two-phase commit, IEEE Softw. 22 (2)
(2005) 64–66.
[71] S. Cranefield, S. Ranathunga, Embedding agents in business processes using [88] Forrester ResearchInc., The Forrester Wave: iPaaS For Dynamic Integra-
enterprise integration patterns, in: International Workshop on Engineering tion, Q3 2016, 2016. ( https://www.forrester.com/report/The+Forrester+Wave+
Multi-Agent Systems, 2013, pp. 97–116. iPaaS+For+Dynamic+Integration+Q3+2016/- /E- RES115619).
[72] DELL Boomi, AtomSphere User Guide, 2017. (http://help.boomi.com/ [89] D. Ritter, M. Holzleitner, Integration adapter modeling, in: Conference on Ad-
atomsphere/GUID- A98714FA- 9EAB- 4B69- BCC8- 7D8984F0B0EC.html). vanced Information Systems Engineering, 2015, pp. 468–482.
[73] IBM, WebSphere Cast Iron Cloud integration, 2017. (https://www.ibm.com/ [90] MuleSoft, Integration: The Cloud’s Big Challenge, 2017. Accessed: 01/2017.
support/knowledgecenter/SSGR73). [91] D. Ritter, Towards more data-aware application integration, in: British Inter-
[74] Informatica, Cloud-Integration, 2017. ( https://www.informatica.com/de/ national Conference on Databases, 2015, pp. 16–28.
products/cloud-integration.html). [92] C. Peltz, Web services orchestration and choreography, IEEE Comput. 36 (10)
[75] Jitterbit, Harmony Cloud Integration, 2017. (https://www.jitterbit.com/ (2003) 46–52.
harmony/). [93] D. Ritter, J. Sosulski, Modeling exception flows in integration systems, in: En-
[76] Microsoft, BizTalk Server, 2017. (https://msdn.microsoft.com/en-us/library/ terprise Distributed Object Computing Conference, 2014, pp. 12–21.
dd547397(v=bts.10).aspx). [94] D. Ritter, J. Sosulski, Exception handling in message-based integration sys-
[77] SAP SE, SAP HANA Cloud Integration, 2017. (http://www.sap.com/product/ tems and modeling using BPMN, Int. J. Cooperative Inf. Syst. 25 (2) (2016)
technology- platform/hana- cloud- integration.html). 1–38.
[78] Oracle, BEA WebLogic Integration, 2017. ( https://docs.oracle.com/cd/E13214_ [95] C. Fehling, F. Leymann, R. Retter, W. Schupeck, P. Arbitter, Cloud comput-
01/wli/docs81/index.html). ing patterns - Fundamentals to design, build, and manage cloud applications,
[79] Tray.io, Tray.io Docs, 2017. (http://docs.tray.io/). Springer, 2014.
[80] Zapier, Zapier v2 Documentation, 2017. (https://zapier.com/developer/ [96] E. Gamma, R. Helm, R. Johnson, J. Vlissides, Design patterns: elements of
documentation/v2/reference/). reusable object-oriented software, Addison-Wesley Longman Publishing Co.,
[81] Apache Foundation, Apache Flume, 2017. (https://flume.apache.org/). Inc., 1995.
[82] Apache Foundation, Apache Nifi, 2017. (https://nifi.apache.org/). [97] SAP SE, SAP HANA Cloud Integration - Prepackaged Content, 2017. (https://
[83] C. Ibsen, J. Anstey, Camel in action, 1st, Manning Publications Co., 2010. cloudintegration.hana.ondemand.com/)
[84] Cloudpipes, Cloudpipes Documentation, 2017. ( https://docs.cloudpipes.com/). [98] SAP SE, SAP Financial Services Network, 2017. ( http://www.sap.com/product/
[85] Tibco, Tibco Cloud Integration Documentation, 2017. ( https://integration. financial- mgmt/financial- services- network.html)
cloud.tibco.com/docs/index.html). [99] D. Ritter, N. May, K. Sachs, S. Rinderle-Ma, Benchmarking integration pat-
[86] Software AG, Webmethods Integration Cloud, 2017. ( http://www.softwareag. tern implementations, in: International Conference on Distributed and Even-
com/corporate/products/cloud/integration/default.asp). t-Based Systems, 2016, pp. 125–136.
[87] GartnerInc., Magic Quadrant for Enterprise Integration Platform as a Service, [100] M. Böhm, D. Habich, S. Preissler, W. Lehner, U. Wloka, Cost-based vectoriza-
Worldwide, 2016. ( https://www.gartner.com/doc/3263719/magic-quadrant- tion of instance-based integration processes, Inf. Syst. 36 (1) (2011) 3–29.
enterprise-integration-platform).

Ritter 17

Uploaded by

Copyright:

Available Formats

Ritter 17

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ritter 17

Uploaded by

Copyright:

Available Formats

Information Systems 67 (2017) 36–57

Contents lists available at ScienceDirect

Patterns for emerging application integration scenarios: A survey

Fig. 1. IT trends since 2005 in the context of application integration.

Fig. 2. IT trends since 2005 an their relationship to application integration.

Fig. 3. Design science methodology used in this paper.

Fig. 4. Distribution of topics mentioned in literature over time.

Trends Patterns [3] Formalization [5,66] Modeling [46,47,66]

Category hits selected Selection criteria Selected Systems

Pattern Boomi IBM Informatica Jitterbit BizTalk SAP Oracle

Pattern Flume Niﬁ Camel Tray.io Zapier Cloudpipes Tibco Webmethods

Content-based Router () –

Category Scope Pattern name System examples

Category Scope Pattern Name System Examples

Conversations Endpoint Commutative Receiver –

(a) Stateful EOIO (b) Stateless EOIO

Categories Description EIP CCP DP

Name pattern identiﬁer

Solution Summary. As an extension a Timed Redelivery until Ac-

6. Quantitative analysis SAP Cloud Integration eDocument/Electronic Invoicing (eDocument).

Fig. 7. Scenarios using original EIP.

Fig. 8. New capability categorization.

You might also like