Server Strategies

Making the right server management moves

In an evolving business environment, CIOs need to optimise server performance by getting server sizing, management, and upgrades right. by Soutiman Das Gupta

The enterprise data centre has evolved along with the increasing complexity of the overall business environment.

Enterprises have seen changes in their business environments with the addition of new products and services, geographical expansion, new personnel, targeted sales and marketing campaigns, and need for compliance. As a result, the enterprise data centre has grown in size and has assumed a position of greater importance among corporates.

To keep pace with the growth CIOs now have to balance a complex mix of the right software, configurations, patches, and vendor relationships. There is pressure to align with fast-paced business growth, and at the same time, minimise system downtime and maximise productivity.

In such a scenario it’s important to keep a close watch on server architecture by nurturing, growing, managing, and updating it with the help of the right policies and technology.

Server management challenges

Keeping a close watch on the server infrastructure is a great way for the CIO to ensure that the business is in safe hands. However, things are not easy on this front since server management involves more challenges than meet the eye.

These include getting the correct server sizing for the various enterprise applications, creating suitable strategies for jobs such as centralisation and planned upgrades, managing a high rate of server growth which requires a lot of manpower for deployment, configuring, administering and keeping track of configuration changes, and frequent updates to ensure that critical applications are available.

These challenges have made it extremely important for CIOs to take server management and architecture seriously.

The right size

Any enterprise today uses a wide range of enterprise applications. There may be ERP, databases, messaging platforms, Web platforms, proxies, CRM, and SCM applications.

Before deployment, as a part of the IT strategy, it is necessary to size server architecture to match the performance levels that the application will have to deliver. Sizing servers is an art that requires a combination of logic, careful study, educated guesswork and calculation.

“From the perspective of an organisation, procuring a server and getting server sizing right is important because one needs to be able to tune the applications deployed on a server so that the predicted number of users obtain a consistent response time,” says Sagar Sule, President, Cyquator Technologies. “At the same time, correct server sizing will enable the available server resources to be utilised optimally.”

Sanjit Sinha, Senior Manager, Hardware Research at IDC India adds, “The right sizing of servers gives CIOs the confidence that they’ll never get complaints about speed and underutilisation. The key is to strike the optimum balance between possible peak application load, and normal load.”

Sizing strategies

	There is a trend among medium and large enterprises in India and the PAC to consolidate their server infrastructure. The aim is to get better performance and availability from the infrastructure
Sanjit Sinha Senior Manager Hardware Research IDC India

Here are a few strategies for server sizing. First, it is necessary to gather as much information as possible about the intended use of applications. You need to know the number of potential users and the number of concurrent users as well.

Will the use be light, moderate, or heavy? For instance, in a bank there will be frequent access to the customer database and comparatively less access to the Intranet portal. In a manufacturing set-up, the production department will need to access the ERP servers more often during the day than the HR department.

The next step will be to identify a benchmark that is the closest to intended usage levels. Most application software vendors have freely available benchmarks that define standards of the systems you plan to use. For example, the MAPI Messaging Benchmark is the benchmarking standard for MS Exchange.

The benchmark will usually have information on the recommended hardware such as the server model, number and speed of CPUs, amount of cache and memory, disk subsystem, and network adapter.

A point to note is that although published benchmarking standards are useful, they assume that the server will be exclusively used for a single application (100 percent CPU utilisation).

	It's always a good idea to use a scale-out strategy rather than a scale-up strategy. You can deploy four 2-way servers instead of a single 8-way server
Anil Jain Vice-president Wipro Personal Computing Business

Prototypes and guesses

You can also build and test a prototype, if you have the luxury of time and money. Another method is to make an educated guess based on your knowledge and experience. This is surprisingly useful especially for very experienced CIOs, who can combine the published recommended server configurations with personal calculations.

It’s always a good idea to adopt hardware that can be readily upgraded by adding processors, memory, hard drives and even network interfaces if required.

Here are a few thumb rules to follow while planning:

• For proxy servers: The RAM should be high (1 GB and more).
• For VPNs: The CPU power should be high.
• For Web servers: The most important factor for sizing is the number of visitors, which is usually underestimated.
• For ERP servers: There should be emphasis on failover and load balancing.

Server consolidation

Enterprises perform server consolidation periodically to gain a number of business benefits. The idea is to do more with less by consolidating server resources and tools, and simplifying training. This helps minimise business disruptions and improve customer satisfaction.

“There is a trend among medium and large enterprises in India and the APAC to consolidate their server infrastructure. The aim is to get better performance and availability from the infrastructure,” explains Sanjit Sinha.

“It helps an organisation respond quickly to changes in the business needs and makes for better manageability,” concurs Anil Jain, Vice President, Wipro Personal Computing Business.

Time to upgrade or replace

Whatever the reason, it is always necessary to look at long-term instead of near-term benefits when upgrading or purchasing new hardware. One shouldn’t wait for a catastrophic failure before upgrading or replacing equipment.

A number of server performance and network monitoring tools in the IT infrastructure give out various signs that indicate whether server hardware needs to be replaced or simply upgraded.

If the performance requirement of the particular application exceeds the current capabilities of the server then it will become apparent if it is time to upgrade. There may be frequent instances of downtime or unexplained application ‘seizures’. Sometimes the specifications of the deployed hardware do not match recommendations. In such cases, an upgrade is desirable.

Sometimes, the speed is sharply reduced and you find yourself spending too much time troubleshooting and fixing server issues. Depending on the nature and extent of these issues, it may be a good idea to replace the boxes.

Upgrade strategies

	From the perspective of an organisation, procuring a server and getting server sizing right is important because one needs to be able to tune the applications deployed on a server so that the predicted number of users obtain a consistent response time
Sagar Sule President Cyquator Technologies

It’s useful to create an upgrade strategy so that it’s performed in a planned and phased manner that minimises wastage.

Sinha says, “One has to first understand and prioritise applications running on servers. Server platform choices are driven by the type and availability of solutions, and the long-term benefits associated with a platform.”

While undertaking a consolidation exercise, CIOs need to consider how they can get better utilisation, availability, performance, and provisioning advantages.

CIOs should look at aspects such as platform monitoring, management and control, and infrastructure and application provisioning. In the next level, enterprises should look at infrastructure virtualisation and service-level automation.

The main factors that Cyquator Technologies considers during server upgrade processes are: the product lifecycle especially from the predicted End of Life (EOL) perspective, the ease of migration of existing applications without downtime, the scalability of CPU, RAM, network interfaces, and disk space, and the amenability of the new hardware for augmentation to the load-balanced layer of the current servers.

On another note, Jain says, “It’s always a good idea to use a scale-out strategy rather than a scale-up strategy. You can deploy four 2-way servers instead of a single 8-way server.” Obviously fans of the SMP approach who run mission-critical applications on 64-way or bigger boxes will beg to differ but for smaller deployments, a scale-out approach can work.

Soutiman Das Gupta can be reached at

soutimand@networkmagazineindia.com