ADOPTING VIRTUALIZATION

VIRTUALIZATION
• Virtualization is a perfect solution for small to medium-scale enterprises.

• It does not suit high performance applications and should be avoided. It may be possible to add
overheads, complexity and reduce the performance.

• In virtualization, the IT industry has a high CPU utilization number indicating optimum usage of
hardware; and based on this number one should not come to the conclusion about the application
usage.

• By using virtualization, CPU utilization during peak hours will shoot up to 50% and it should not
override the SLA (service level agreement) of an organization.

• At present, in virtualization, the CPU and I/O overhead for storage and networking throughput are
minimal, when high storage or more hardware I/O requirements are needed, then avoid using VM.
 VIRTUAL CLUSTERING
• Cluster is defined as ‘a type of parallel or distributed system that consists of a collection of interconnected
computers and is used as a single, unified computing resource’.

• Forming a cluster refers to a collection of computers bounded together to form a common resource pool.

• A task or job can be executed on all computers or a particular computer in the cluster.
 BENEFITS OF CLUSTERING
Scientific applications: Applications running on supercomputers can be migrated to Linux cluster (which is more cost
effective).

Large ISPs and E-commerce enterprise with large database: Internet service providers or e-commerce web site that requires
high availability and load balancing and scalability.

Graphics rendering and animation: Film industry is benefited because of clustering, in particular for rendering quality
graphics and animation scenes. Examples include Titanic, True Lies and Interview with the Vampire.

● Fail-over clusters: Using clusters, network services are increased in terms of availability and serviceability and when a
server fails, its services are migrated to another system. For example, failover servers are database servers, mail servers and
fi le servers.

● High availability load balancing clusters: Application can run on all computers and can host multiple applications.
Individual computers are hidden to the outside world as they interact only with the clusters. This kind of clustering produces
best results with stateless application and when executed concurrently.
 BENEFITS OF CLUSTERING
Scientific applications: Applications running on supercomputers can be migrated to Linux cluster (which is more cost
effective).

Large ISPs and E-commerce enterprise with large database: Internet service providers or e-commerce web site that requires
high availability and load balancing and scalability.

Graphics rendering and animation: Film industry is benefited because of clustering, in particular for rendering quality
graphics and animation scenes. Examples include Titanic, True Lies and Interview with the Vampire.

● Fail-over clusters: Using clusters, network services are increased in terms of availability and serviceability and when a
server fails, its services are migrated to another system. For example, failover servers are database servers, mail servers and
fi le servers.

● High availability load balancing clusters: Application can run on all computers and can host multiple applications.
Individual computers are hidden to the outside world as they interact only with the clusters. This kind of clustering produces
best results with stateless application and when executed concurrently.
 BENEFITS OF CLUSTERING
Virtual Cluster Description:-

A virtual workspace is composed of workspace metadata, which is represented as an XML Schema and implementation-
specific information like pointer to the image in VM which implement a given workspace.

The aim of the metadata is to capture workspace requirements in terms of virtual resource, software configuration and
features.

In this part, we illustrate the addition to the workspace metadata and implementation which is necessary to characterize a new
type of workspace called a virtual cluster.

Head nodes and worker nodes are two kinds of nodes in a virtual cluster.

A head node differs from worker nodes in principle and configuration, particularly in software and operational setup.

Though worker node configurations are similar, they may be given different names or their position may be different.
VIRTUALIZATION ARCHITECTURE AND SOFTWARE
 VIRTUALIZATION ARCHITECTURE AND SOFTWARE

• MultiNet Protocol Driver (MPD) (shown in Figure) is used to implement virtualization of wireless adapter
and placed as an intermediate layer between MAC and IP.

• The figure explains how MPD can be used to share a wireless network adaptor and its MAC. MPD sits in
between IP layer and physical layer (MAC).

• When an application wants to use more than one network than using TCP/IP, the path is sent to MPD,
which is responsible for switching and allocating the network and appropriate MAC address.

• MPD exposes the wireless LAN media adapter as always active. IP stack checks the adapters, whether
they are active always , even though only one is connected at a given time.

• Switching and buffering of packets across networks is done by MPD.

 VIRTUALIZATION ARCHITECTURE AND SOFTWARE
• Virtualization of a WLAN card, where a user needs to connect to three networks (wireless).

• The MPD represents three virtual adapters, and everything appears active to IP even though only
Network 2 is active at the moment as depicted in the figure
VIRTUALIZATION ARCHITECTURE AND SOFTWARE
 Virtualization Benefits

 Improve the server’s utilization: When non-virtualization, servers utilization is about 25%. After deploying
virtualization utilization increases to 70% and goes close to 90%.

 Reduce the number of physical servers: It is typical to see consolidation of servers be in the ratio of 5:1 up to 10:1
range.

 Contribute to corporate ‘green’ initiatives: Number of servers can be reduced through server virtualization. It will not
only require less physical data centre floor space but also use a lesser power and need fewer cooling machines for the
same workloads.

 Give a fast return on investment (ROI): Business leaders demand one year of less time for ROI.
 BEST PRACTICES IN SETTING VIRTUALIZATION
Avoid Monopoly: The first important step in virtualization standards and best practices is to employ an
experienced partner to work for the organization

Proactively Engage Stakeholders:- Involve business stakeholder and get help to manage the transition to
virtualization and ensure that it supports the overall business goals. Be proactive about instructing all affected
stakeholders and users about how resources will be assigned, shared and also clarify about the benefits of
moving the business to virtualization.

Balance Resource Needs:- Many applications requesting for the same resources may leave those
applications contending for inadequate RAM, processor capacity, disk I/O or network bandwidth. Take stock
of your applications and their computing requirements before moving everything to a virtual server.

Do not Overload Physical Servers: So that while taking a single host server down for maintenance, multiple
mission critical applications are not down at the same time. Also keep in mind that virtual servers like
physical servers have cyclical resource needs that can spike dramatically when business processes
are most demanding, whether it be weekly or once in four months.
 BEST PRACTICES IN SETTING VIRTUALIZATION
Prevent Virtual Server Sprawl:- New virtual machines can be created so simply and rapidly, that it can feel like a free
and limitless resource.
 BEST PRACTICES IN SETTING VIRTUALIZATION
Prevent Virtual Server Sprawl:- New virtual machines can be created so simply and rapidly, that it can feel like a free
and limitless resource.
 PITFALLS OF VIRTUALIZATION
Poor preparation: Adding virtual OS to boost computing power may create problems

Insufficient server capacity: Virtualization does not increase the computing resources, but
only the usage.

Mismatched servers: When multiple servers are virtualized which uses different chip technologies
(Intel and AMD), there will be various problems while migrating VMs between them.

Slow network communications:- Upgradation of network’s communications capabilities are required

Slow mechanical disks: Present disk technology cannot retain with the read/write demands of multiple servers in high-
demand peak hours, you will face some latency

Uneven workload distribution: Finetuning the distribution of processing requirements across all physical servers is needed
to maximize usage of data centre computing power.

Security risks: Virtualization does not improve network security by itself. Firewalls, antivirus needs to be installed and
keep them repaired and updated.
 VIRTUALIZATION IN GRID
• The main focal point in grid computing lies in secure resource sharing in accessing computers, software and data in a
dynamic atmosphere.

• Sharing of those resources has to be fi ne-tuned and handled in a highly controlled manner.

1.Nontrivial qualities of service delivery

2. Habit of using standards in all places
3. Resources coordination
 Grid Computing and Virtualization

• Virtualization is not a solution for enterprises to manage their resources although it provides richer capabilities in
managing and moving the OS in different hardwares.

• It helps to run multiple workloads in a single machine with clear distinction between them.

• Virtualization can do suspending, resuming and migrating images in run-time.

• Like grid, virtualization is also a trend. Instead of having more hardwares to meet peak demands, organizations can use
virtualization approaches to get better utilization out of the existing hardwares which are not fully utilized.

• Also virtualization is possibly becoming a mainstream approach in managing network resources.

 Using Virtualization in Grid Computing

Grid Computing Level

Grid architecture represents the internal structure of grid computers.

The following layers are broadly present in a grid node:

1.The top layer consists of high-level applications, such as an application to perform predictive modeling.

2.The second layer, also known as middleware, manages and allocates resources requested by applications.

3.The third layer consists of available computer resources such as CPU, memory, and storage.

4.The bottom layer allows the computer to connect to a grid computing network.

gLite Middleware has strong dependencies to Scientific Linux as operating system which is used by the LHC (Large Hadron
Collider) Computing Grid High Energy Physics Virtual Organizations (VO)
 Using Virtualization in Grid Computing

• LRMS( Local Resource Management System):-The key research work in grid computing is to use virtualization in
resource management system locally.

• When virtualization is adopted locally, it supports job managements and migration for virtual machines instead of
jobs.

• In fact, virtualization exhibits with live and delayed migration of virtual machines even on added feature.

• Merging the above mentioned LRMS features (suspension, check pointing) with migration, helps the LRMS in
varying the current resource allocation dynamically.

• In addition, alteration of resources (e.g. number of CPUs, RAM) allocated to a virtual machine is possible and also
it allows dynamic updating in service quality that can be done on-the-fly or by suspending and resuming depending
on the virtualization platform
 Using Virtualization in Grid Computing
Virtual Machines

Virtual machines can check the execution of applications and also they are useful tool grid system administrators.
VMs can easily manage the systems in terms of isolating and partitioning the amount of resources needed.

It also explains, how computational services can be deployed virtually in grid.

The architecture consists of physical machines, physical worker nodes, infrastructure manager,GridWay, virtual
workspace service (VWS) and grid middleware components such as monitoring and discovery of services (MDS),
grid resource allocation manager (GRAM) and GridFTP.

 User’s requests are submitted to GridWay (meta-scheduler). Depending on the policies, SLAs and the
grid load, an infrastructure manager deploys it in worker nodes (VO-specific).

 The deployment of the VMs supporting the worker node is done through the virtual workspace service (VWS).

 When worker node is up, it registers the information stored in information service (MDS). Now GridWay can
detect the slot and jobs are submitted through GRAM.
 Using Virtualization in Grid Computing
Virtual Machines

Architecture developed for dynamic provisioning of computational services in grid.