Telescope 2007

Data centres: Application acceleration

With an increase in the number of users, more data being stored at the back-end, and a consequent rise in the number of requests that a database server has to address, it is harder than ever to figure out where a bottleneck lies in this tangled mess. But solutions are available in the market that help CIOs optimise application performance. By Aishwarya Ramani

An ERP system, a BI suite perhaps, a messaging server and an Intranet are typical applications that run in a corporate data centre today. The aim here is to examine the bottlenecks within a data centre and explore the various technologies available to optimise application performance.

CIOs have to ensure that a data centre is equipped to deal with bottlenecks as and when they arise due to the large volumes of data which are being processed. Vishal Dhupar, Managing Director, Symantec India feels that it is not uncommon for organisations to focus their IT operations on service availability, while performance issues are reacted to—rather than anticipated—and treated in a proactive manner.

Applications aren’t getting any smaller, data mountains are turning into Everest, all of which contribute to make fine-tuning application performance that much more complex a task.

Today’s data centres are static and hard-wired, making them inflexible and hard to reconfigure. Any change within a data centre requires too many people and involves a number of steps. Vendors such as HP opine that apart from this, since today’s data centres are built at one-go, there is a lot of wastage in terms of power, cooling, space, people and money.

“While factors like data protection, compliance, availability and management are driving the consolidation of IT resources, many applications must be deployed centrally. With these come challenges related to application performance,” says Sumit Mukhija, Business Development Manager, Cisco Systems India & Saarc.

Applications are becoming more complex and inter-dependent. In addition to this, multi-tier architectures are the norm. Owing to these factors, performance management is harder than ever. Organisations are struggling to find a repeatable and understandable way to measure the service levels at each tier within an application infrastructure, and to correlate transactions across application tiers. Adding to this problem are the frequent changes made to applications with changing business environments with no or minimal testing performed before deploying an application in a production environment. Another challenge that organisations need to overcome is that of taking corrective action once a problem is identified and building intelligence around the same to avoid recurrence.

With business environments changing, frequent changes are made to applications. Deploying applications without conducting proper tests leads to confusion and chaos. The traditional method of tuning applications in a multi-tier environment does not address performance issues. The traditional approach involves the use of data that has been captured by various components and correlated in terms of time slices. This means that the diagnostician or performance analyst studies the data points for the various components along the path for a slice of time and then uses his best judgement to connect the dots. Not only is this extremely difficult, it also involves trial and error.

A data centre has several components—application servers, database servers, storage, back-up servers, networking equipment, power back-up systems, etc. Typically, a CIO allocates the management of each of these entities to different third-parties. Agrees Hilal Issar Khan, Head, IT, Honda Siel Cars India, “We have multiple SLAs with multiple partners.” He says that the problem is aggravated when multiple vendors are roped in, as each of them refrains from taking responsibility for bottlenecks that arise. T G Dhandapani, Corporate CIO, Sundaram Clayton, also agrees that this disparity contributes to exacerbate the problem.

Inadequate storage and under-utilised resources contribute to these bottlenecks as well. Dhandapani explains that one of the reasons for the underutilisation of resources is simply lack of knowledge vis-à-vis the proper utilisation of available resources. Khan seconds this view by saying that the knowledge base of manpower is wide but not deep.

Although vendors cannot do much about inadequately-skilled manpower, they do suggest ways of handling a resource crunch. Additional equipment—such as servers that help handle problems faced due to inadequate resources—do not come cheap, and it is necessary to examine the different ways in which a resource crunch can be dealt with.

According to Rajendra Dhavale, Consulting Director, CA, one needs to go in for a solution that enables a new level of management control across the enterprise by integrating and automating the management of IT applications, databases, networks, security, storage and systems across departments and disciplines to realise the full potential of each of these elements.

And Dhupar says that taking a proactive approach towards monitoring / analysing infrastructure and application behaviour allows an organisation to allocate the right resources to each requirement. This optimum and correct allocation of IT resources helps an organisation anticipate bottlenecks in advance and gives sufficient time for risk mitigation. The approach not only helps allocate right but it also helps ensure optimum utilisation and increased efficiency—vital elements in any capacity-on-demand strategy.

Kallol Hazra

Informs Kallol Hazra, Practice Director, HP Services, Technology Solutions Group, HP India, “When there is a sudden increase in the number of users for a particular set of applications, we will need additional servers in the application layer to maintain the same performance levels. If the servers are already wired up they have to be added to the network in a distributed architecture. If, on the other hand, we have a data centre with server blades, we have to simply reconfigure a few blades through software and add the resources to the task. The network capacity has also to be augmented to cater to the additional load.”

Considering all these factors, there are several solutions available today that help CIOs identify bottlenecks and help optimise application performance in a data centre.

The root-cause

What companies really need to look for are enterprise-wide event management solutions with appropriate root-cause analysis supplemented by business services that help zero in on a problem. Several solutions are now available to help CIOs find bottlenecks and make it easier to resolve them.

Dhavale is of the opinion that identifying bottlenecks is more time consuming than repairing a problem, while Dhupar points out that the combination of multiple tiers and disparate technologies makes the diagnosis of problems a complex task as the traditional approach of inward focus within a specific tier of applications does not solve performance problems. “What is required is an outward approach of performance management which correlates performance metrics across multiple tiers to understand and improve the performance of transactions from an end-user perspective.”

Busting bottlenecks

Proactive application performance requires a combination of people, processes and technology. In identifying and implementing performance management technologies to support a proactive strategy, it is important to select a tool that offers both broad and deep performance analysis. It must be able to monitor and manage performance across the entire transaction path, and offer deep analytics within each tier. Without visibility across all tiers of the architecture, the IT head may think that a problem is being solved, but in reality it is a symptom of the underlying problem—or he may be addressing the wrong thing.

An effective performance management method should have elements to identify symptoms that indicate a performance problem, identify the tier where the problem lies, discover the root-cause of the problem within that tier, take the steps required to improve performance, and make sure that the steps taken have achieved the desired goal. These stages combine to form a process that provides a systematic approach to finding and resolving all kinds of performance issues, both predictable and unforeseen.

Hazra stresses on the need to create and transform all resources from being physical and static to being virtual and adaptive. This, he says, will definitely help in reconfiguring resources on the fly. The cost of operations can be lowered by automating most of the processes and using them consistently.