Optimizing the Data Center
How to make those fancy systems work harder and more effectively for you through software tools and best practices.
By Sanjay Gupta
There was a time when nobody in India talked about data centers. What existed, instead, were EDP or Electronic Data Processing departments, with a few servers and storage boxes lying in cramped – even if ‘air-conditioned’ – spaces. While many companies still continue to present that picture, most progressive and fast-paced organizations now showcase data centers as the heart and soul of their enterprises. Faster pace of business, global competition and changing customer dynamics are ensuring that companies ratchet up their computing infrastructures – essentially the data centers – by an order of magnitude.
Talk to any CIO or vendor today about storage and server, and you’ll most likely come across data center consolidation initiatives. There’s no concrete research data available for India, but Network Computing believes that more than half of the country’s top 1,000 companies have undertaken data center consolidation as a key IT project or are going to do so in the next one year. Needless to say, much of their enterprise budgets and resources will be focused on these efforts.
However, data center consolidation is not just about stuffing more blades into the rack or making an inventory of all the storage silos and replacing them with bigger boxes. As any CIO restoring their SANity or making an iMPLSe buy would tell you, the objective is get the biggest bang for their buck (which no longer stops only at the CFO!). Jokes apart, getting the best performance and results from data centers is serious business that can make the difference between a highly available, modular, reliable and secure infrastructure and one that sputters and screeches despite investments in state-of-the-art equipment.
In order to get more juice out of their data center infrastructures, companies across the board are evaluating and implementing software tools and adopting best practices. Such initiatives range from power management software, database tuning tools and enterprise systems management to IT Service Management, IT Infrastructure Library, ISO and the Control Objectives for Information and related Technology (COBIT).
Going virtual
About a couple of months back, Hero Honda undertook a server consolidation exercise with more focus on virtualization and green or energy-efficient computing, says Vijay Sethi, VP, Information Systems. As a result, the company achieved a physical-to-virtual server ratio of 1:3, which will further go up to 1:4. Through consolidation and virtualization, Hero Honda has improved its server utilization from under 10 percent to between 30 and 40 percent.
Another organization, the Center for Railway Information Systems (CRIS), which is responsible for taking care of the huge computing infrastructure of Indian Railways, is looking at further improvements in the way its five data centers work. Says SS Mathur, General Manager – IT, CRIS, “The biggest challenge of them all is the reliability aspect. Then comes the high uptime requirement. At times, maintaining uptime in a complex environment becomes a big challenge for us.” For CRIS, each data center works as a back up for another. The next step for it will be to set up an exclusive site for disaster recovery. Currently using HP’s OpenView for network management, CRIS is evaluating it for the full enterprise system management capabilities. Mathur has also initiated virtualization of storage and is working on enhancing the security of the entire infrastructure – moves that are aimed at enhancing the data centers’ efficiency.
Besides enterprises, hosting and managed service providers such as NetMagic and Sify, which have data centers spread across thousands of square feet, are working toward better management and optimization of their facilities. These companies are even more driven to enhance the profiles of their data centers, which serve multiple clients in an increasingly complex service delivery environment. And the growth they are witnessing is scorching (more heat for them!). Consider this: While NetMagic took six years to fill up a 10,000 sq. ft. data center in the initial years of its operation, within the next six months it had already pre-sold an equal amount of space.
For NetMagic, it is important to leverage the hardware rack density so that its operations and client contracts remain profitable, because its income is calculated as revenue per rack. The company has used virtualization to its advantage. Furthermore, it has set an example by creating its own data center management application entirely using open source tools.
Vendors are also keen to extend the benefits of virtualization beyond servers and storage. According to Aldrin D'Souza, Country Manager - Tivoli, Software Group, IBM South Asia, “IBM’s Usage and Accounting Manager is a solution that aims to optimize data center infrastructure by providing insights into the relationships between virtualized and physical IT assets, who is utilizing shared resources and what and how much they are using.” Such information is critical in a number of ways, he says. For many companies, it offers a means for quantifying how employee and departmental behavior contributes to overall IT TCO. For others, it provides a reliable mechanism to support company chargeback processes. For organizations such as IT outsourcers, the software can serve as a measurement tool underlying billing processes and service level agreement (SLA) compliance.
Database performance
Given the fact that much of today’s content and applications are dynamically generated using various databases, the way companies configure and use their database systems greatly affects the overall health of their data centers. Practices such as database tuning and clustering can come in handy to drive more performance and value out of the databases.
That, however, is easier said than done. According to Seema Ambastha, Director – Sales Consulting, Database Technologies, Oracle India, “Database clustering is not happening to the extent it should because of the confusion in the way data center managers or CIOs compute costs. But it’s catching on, especially for business-critical operations. Some data center heads have looked at clustering as a way of eliminating single points of failure.” Ambastha is of the view that even if data center heads need to keep and manage a multiplicity of operating systems and hardware platforms, they should at least standardize on a single database platform for increased ease of management and better application integration and performance.
When fertilizer major IFFCO consolidated its multiple data centers into a centralized location in Delhi, it deployed several performance management tools. Among them is Toad from Quest Software. “We are using Toad to optimize the day-to-day operations of all Oracle applications,” says Executive Director SC Mittal. The company’s objective is to achieve maximum efficiency from its data center consolidation initiative.
The cooling factor
Power consumption in data centers continues to be a big worry for all companies. Says Deepinder Singh Bedi, Director, International Business & Marketing, Tulip Telecom, “A big concern is the constraint to operations and growth represented by the power consumption of IT equipment. Energy costs already are the second-largest line item associated with data center operations and are predicted to rise to 50% of the overall IT budget within the next few years.”
Cooling in the data center is the biggest component of this power cost. According to HP, 60 percent of the power consumed in the data center is used for cooling the heat dissipated by the IT server infrastructure. Durgadutt Nedungadi, Director, Marketing & Alliance, Technology Solutions Group, HP India says, “HP has developed a software that sits on top of virtualization software that can shift functions virtually from a hot server to a cooler server without affecting the ongoing operation.” The company’s Dynamic Smart Cooling technology consists of advanced software residing in an intelligent control node to continuously adjust air conditioning settings in a data center, based on real-time air-temperature measurements from a network of sensors deployed on IT racks.
Software-based energy-monitoring solutions are catching on, with several vendors jumping into the fray. IBM’s System Director Active Energy Manager helps companies monitor energy consumption to allow better utilization of available energy resources. Says IBM’s D'Souza, “The software enables customers to trend actual energy consumption and corresponding thermal loading of systems.”
Such advanced software solutions help data center managers and CIOs measure and monitor cooling or energy consumption costs across various departments, users or infrastructure components so that they can get a better handle on their current energy needs and plan requirements for the future.
Embracing best practices
CIOs recognize that having the latest hardware and software at their disposal is one thing and using them in the best manner possible is quite another. That’s why they are looking to adopt the best practices that go with implementing a technology. Hero Honda, for one, is refining its data center processes in accordance with the best practices ingrained in ITIL and COBIT. The objective, says Sethi, is to increase efficiency, improve reporting and ensure smooth operations through well-documented processes and shared information. The company has also adopted internal policies that aim to optimize the life of servers used in the organization. Says Sethi, “We have a policy to optimize the use of our servers by moving them from production environment to test environment after certain number of years of usage.” Another policy measure the company has undertaken is to negotiate its AMC and support contracts with the vendors so that it gets optimum returns from the equipment it buys.
There are some who are establishing their own standards for their particular needs. Hosting provider Sify, for instance, has created SIMS or Sify Infrastructure Maintenance Standard, which helps the company manage its data center locations and solve any incidents arising, says Saji PK, VP - Technology, Sify Technologies.
Another best practice among companies, according to Anand Naik, Director of Systems Engineering, Symantec, is the standardization of security infrastructure. The need for such standardization arises because information at a data center needs to be “secured dynamically,” says Naik. No wonder the company now offers advisory and professional services that include a comprehensive status report on the customer’s data center. The idea is not only to make the customer aware of its existing set up but also to highlight some pain points and newer business problems it might encounter when it needs to scale up.
Companies are also adopting best practices associated with reducing their energy consumption in the data center. For instance, NetMagic has incorporated energy efficient practices to ensure that Hot Aisle / Cold Aisle anomalies do not exist. The company is taking this further in ensuring that now the cooling is restricted to only those racks in which the equipment is placed rather than waste energy in cooling each and every nook and cranny.
Then there are other important areas like management of service level agreements (SLAs) to ensure higher uptimes. Says Ambastha of Oracle, “All performance in the data center has to be SLA-driven, but it can happen only if one is able to measure it accurately.” For one, SLAs need to be more granular and specific, she says. “Instead of specifying, say, 4 millisecond as the general response time for the intranet, IT managers should further break it down to which application and for which users.” On the choice of tools available for measuring SLA performance, she says that the right tools should be able to measure all “the components that exist from the time the user hits the Enter button to the time they get back the response from the back end.”
The way forward
For many data center heads, breaking up all the components and defining various business needs and then analyzing them is a big challenge. However, with the right tools and best practices, they can overcome it.
Another challenge is how to take the computing resources utilization level from the current 25 percent to the globally recognized standard of between 60 and 70 percent, according to Ambastha. Besides, how efficiently CIOs and data center mangers are able to virtualize server and storage in their facilities is another metric they must look at.
As data centers become more powerful and expansive in their scope and business impact, their efficiency of energy consumption and operation will become a competitive differentiator for companies. Those who manage and run their own data centers will need to fill the skills and technology gap in their IT organizations, while those who outsource them to hosting providers must become better negotiators, monitors and enforcers of contracts.
(with inputs from Jamsheed Gandhi and Faiz Askari)
|