PERSPECTIVE

EFFECTIVE CAPACITY MANAGEMENT

WITH MODELING AND SIMULATION -
ASSISTED PERFORMANCE TESTING

Abstract
In this competitive marketplace, businesses seeking to maximize profitable outcomes
need to ensure their information technology (IT) systems are robust in order to
deliver uninterrupted application availability. This is significant for enterprises as high
performance of business applications is an underlying driver of business growth.
Accordingly, the top priority for system administrators is to maintain the health
of their servers in order to meet stringent application performance Service Level
Agreements (SLAs). However, several organizations are suffering due to a reactive
approach to monitoring servers and applications only in the production phase. Thus,
performance issues can be addressed only after they have occurred, often impacting
operations and business.
To address this challenge, organizations need performance testing to detect issues
before the production phase in the Systems Development Lifecycle (SDLC). However,
to be effective, testing needs to be performed continuously and should consider
changes in workload and performance owing to business growth and transitions.
Organizations need a proactive approach that can deliver a timely and holistic view
of an application’s performance, thereby mitigating instances of application failure
and overcoming performance issues and server downtimes. This paper discusses
such a proactive approach that enables organizations to benefit from predictive
and continuous performance and capacity management of their applications and
infrastructure.
 Introduction
In today’s connected world, the performance of information technology (IT) systems has a critical role play in shaping the success of
an organization. System administrators are required to execute multitudinous operations to monitor and improve the performance
of all individual components of business applications. Some of these activities include monitoring the system performance/workload,
overcoming business outages or poor performance, and system/infrastructure upgrades with quick turnaround times1.

In such a situation, performance testing is critical to ensure that application end users do not face performance issues in the
production environment. However, typically, performance engineering exercises that identify and resolve performance-related issues
are reactive in nature. Further, the scenarios for performance testing change continuously according to business growth and these
transitions are not considered during performance testing exercises. Today, performance engineering exercises are evolving from
reactive to more holistic and proactive methodologies2. The gradual transition is described in Figure 1 below.

To successfully meet system performance

expectations, businesses need to
consider future workload through Performance Performance
capacity management exercises3. Monitoring Monitoring
Currently, such exercises are unable Reactive
to deliver timely inputs regarding Performance Performance
Monitoring Monitoring
deteriorating system performance and
infrastructure upgrades. Thus, to sustain
optimum growth, businesses need to
adopt mechanisms that can continuously Capacity Forecasting and Performance Prediction
monitor capacity management and
system performance, identify errors and
accordingly take corrective measures. Performance Management and Capacity Planning
This can be achieved by linking proactive
performance testing techniques with
Proactive
continuous capacity management Continuous and Predictive Capacity Management

activities in a holistic manner.

Figure 1: The Reactive to Proactive Transition of Performance Engineering Activities

External Document © 2018 Infosys Limited

What is Capacity Management?
A MET A Group study states that capacity manage IT resources with clear inputs from manner’. Additionally, ITIL finds that a
planning is “the most important issue application and infrastructure planning typical capacity management exercise
for large enterprises with employee groups. consists of application sizing, performance
strength of more than 1000 people”. The According to the Information Technology and workload monitoring, resource and
study further states that the key goal for Infrastructure Library (ITIL), capacity demand forecasting, and modeling5.
Capacity Planning and Management is to management is defined as ‘the discipline Currently, a typical capacity management
provide a holistic and measurable view that ensures IT infrastructure is provided process involves several steps as described
of the IT environment and to describe its at the right time, in the right volume and in Figure 2 below6.
relevance in business terms4. To achieve at the right price, ensuring, at the same
this goal, organizations need to proactively time, that IT is used in the most efficient

Many enterprises believe that capacity

Instrument Monitor Characterize
management processes should be run the systems Usage Workload
only once a year. However, ITIL states
that for capacity management to be
effective and successful it should be
Change System System Forecast
continuous and involve measurement,
Parameters Model Workload
analysis, prediction, and tracking of
performance and capacity5, 7.

NO
Are cost and YES
performance DONE
acceptable?

Figure 2: Current Capacity Planning Process

External Document © 2018 Infosys Limited

 Continuous and Predictive Capacity Management – The Proposed Methodology
Continuous and predictive capacity management enables capacity planners to share inputs for infrastructure upgrades prior to testing. This
methodology provides mechanisms that track application use, monitor changes in usage over time, forecast trends, and proactively take
corrective actions. The key steps involved are as mentioned in figure 3.

Institute Capacity Upgrade

Workload Performance and

Modeling SLA Monitoring

Production Capacity
Environment Upgrade Plan

Workload Simulation
Forecasting

Performance
Modeling

Figure 3: Continuous and Predictive Capacity Management Methodology

The workload modeling tool monitors the usage of applications in the production environment and tracks the variation in usage over
time.The historical usage data collected can be used to predict future changes in the system.

A simulation engine is used to predict system performance in the current production environment. The engine comprises system
performance models and is built using the performance test results and the forecasted workload. In cases where the system
performance does not meet the pre-defined Service Level Agreements (SLAs), proactive corrective measures are taken before an issue
actually surfaces.

The performance model of a system is always built in line with the actual system in the production environment. It is designed by
continuously monitoring the use of infrastructure resources and the performance of the system. If deviations are found between the
performance model predictions and actual system performance, the performance model is modified to mirror the changes in the
actual system.

External Document © 2018 Infosys Limited

Information Workflow in Proactive Performance Testing
Monitoring the production environment Performance testing is conducted before a further used to strengthen the test results
provides data on system workload and system moves into the production phase and provide informed recommendations
performance. The current workload of a of the Systems Development Lifecycle to augment the performance of an
system can be obtained by analyzing log (SDLC). The results from the performance application. If required, these models can
files on a web server. A large set of this testing as well as system architecture and be revised to overcome the differences
workload data can then be used to predict infrastructure details are used to build between the test and production
workload. Such predictive forecasting also the initial performance models for the environments using industry benchmarks.
considers possible changes in workload test environment. This modeling and
owing to business growth. simulation-enabled performance testing is

Server access Workload Requirement

Modeling Document

Workload Business
Intelligence
Workload Drivers &
Forecasting Requirements
Forecasted Workload
Concurrent Hits
System Performance
feed from Performance Predictive Model
Monitors Performance Modeling Creation

Capacity
Legends Projection

Manual Work Inputs

Production
Environment Servers

Figure 4: The Information Workflow in a Continuous and Predictive Capacity Management Methodology

The projected workload is an input for The proposed solution methodology and performance modeling process.
the performance model and is used to drives significant benefits and helps It enables system administrators to
predict performance under the modified organizations to: address infrastructure requirements for
workload. If the hardware and software Analyze the effect of external consistent system performance over
performance shows deviations from the changes such as workload growth a period of time and deliver upgrades
pre-defined SLAs, the system capacity is on the performance of a business when required.
upgraded accordingly using data from application. This helps minimize risk, Maximize their return on investment
multiple rounds of performance modeling improve quality and enhance end user (RoI) for infrastructure and ensure
and several configurations. The system experience. uninterrupted application availability,
environment can then be continuously thereby helping them achieve their
Upgrade system infrastructure
monitored to ensure SLA adherence and to business objectives.
through a monitoring, forecasting
correct the performance models.

External Document © 2018 Infosys Limited

 Conclusion
Organizations today need uninterrupted methodology can be leveraged to deliver end user. Further, system administrators
application availability in order to ensure advanced capabilities such as timely benefit from instant visibility into system
sustained profitability. They require tracking, measurement and analysis of issues and ongoing requirements for
solutions that can improve their existing performance models before migrating infrastructure upgrades. The solution
performance testing capabilities to to the production environment. Such a methodology ensures consistent
ensure superior system and application proactive methodology enables businesses business application availability, thereby
performance. With the aim of improving to accurately predict and adjust system maximizing RoI, enhancing the end-user
testing by ensuring robust predictive and performance in the actual environment experience, and ensuring superior system
management mechanisms, the continuous with minimum risk of errors, thereby performance.
and predictive capacity management ensuring a better experience for the

External Document © 2018 Infosys Limited

References
1. “Analyzing Computer System Performance with Perl::PDQ” by N. J. Gunther, Published at Springer-Verlag GmbH, 2005.
2. “Performance Engineering of software systems” by C.U. Smith, Published at Addison Wesley, 1990.
3. “Model Based Performance Prediction in Software Development: A Survey” by Balsamo, S., Di Marco, A., et. al. and published at IEEE Trans.
Software Engg., pp. 295-310, Vol 30., No. 5, May 2004.
4. “How Does Your IT Organization Measure Up to Current Industry Wide Spending and Performance Metrics”, Meta Study, November 2003.
5. “ITIL Capacity Management”, Web Source: http://www.itlibrary.org/index.php?page=Capacity_Management
6. “The Art of Computer System Performance Analysis: Techniques for Experimental Design, Measurement, Simulation, and Modeling” by Raj
Jain, Wiley, April 1991
7. “Information Technology Infrastructure Library” Web Source: http://en.wikipedia.org/wiki/ITIL#Capacity_Management

About the Author

Nidhi Tiwari
is a Senior Technical Architect with Infosys Labs, the research wing of Infosys. She has over ten years of experience in varied software
technologies. She has been working in the performance engineering and cloud computing for six years. Her research interests include
adoption of cloud computing, cloud databases and performance modeling. She has authored papers for many international conferences,
journals and has a granted patent as well.
She can be reached at [email protected]

Amit Gawande
works as a Technology Lead at Infosys Labs. He has over five years of experience with a focus in the areas of performance engineering
methodologies and cloud computing. His research interests include performance modeling and simulation techniques.
He can be reached at [email protected]

External Document © 2018 Infosys Limited

For more information, contact [email protected]

© 2018 Infosys Limited, Bengaluru, India. All Rights Reserved. Infosys believes the information in this document is accurate as of its publication date; such information is subject to change without notice. Infosys
acknowledges the proprietary rights of other companies to the trademarks, product names and such other intellectual property rights mentioned in this document. Except as expressly permitted, neither this
documentation nor any part of it may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, printing, photocopying, recording or otherwise, without the
prior permission of Infosys Limited and/ or any named intellectual property rights holders under this document.

Infosys.com | NYSE: INFY Stay Connected