Data centre testing: 10 essential tips

Ten things you need to know

Q: We are planning to integrate devices into our network based on the performance claims in the vendor literature. Is that a reliable strategy?

A: Network device vendors sometimes test their equipment under ideal circumstances which does not take into consideration the critical customised requirements of your network environment. Also, vendors sometimes test their devices according to incorrect performance benchmarks. You may have no way of knowing what the vendor’s results are based on. But even if the vendor did test in a real-world environment, their literature will probably not reveal how it will perform when it interacts with the devices and applications in your network. Choosing a device and vendor without acceptance testing can be a risky strategy and jeopardise the successful and optimal operation of your network. A device may look good on paper, but you know how good it will perform in your network only after you have tested it in a real-world environment.

Q: My highly qualified team has developed a new network design that will take us into the next generation of capability providing the performance and capacity we need for the next two years. How important is it to validate this network design?

A: Modern networks are extremely sophisticated, with a broad range of protocols and applications, complex interdependencies, and unforgiving expectations from customers. It is extremely challenging to deliver on the performance availability, scalability and security requirements of a network deployment, or integration, without robust design validation. This validation is made possible by expert and knowledgeable testing performed with the right test team and tools before the network goes live. Findings from Nemertes Research state that launching an IT initiative without testing can completely wipe out the anticipated operational and cost benefits.

Q: Can’t my team do the end-to-end network implementation, from design to turn up and maintenance?

A: Yes, they can, but because design and implementation are not their necessarily their core areas of expertise, they’ll possibly miss important issues that will affect your schedule as well the performance and reliability of the network. Testing the network that you designed is like taking a calculus test and then grading it yourself. You’re not going to see your own mistakes. A team that designs a problem into a network will probably not know to test for that problem.

Q: We’re launching a new technology. My team is smart. They can learn as they go with the implementation, right?

A: They can learn as they go through trial and error, which is expensive and time-consuming. And as customers are part of the error cycle of this process, it may be damaging to your credibility and reputation. A better solution is for your team to learn as they partner with a team of seasoned and qualified test professionals who have the tools and the expertise to know what to test, how to interpret the results, and how to troubleshoot the problems and resolve them, before you go live.

Q: Why should I spend the time and budget to test a data centre consolidation? I already know the existing network performs OK and all I’m doing is reconfiguring in the same location.

A: Changing one aspect of a network can have a significant impact on performance. Consolidating data centres changes many aspects of the network. Distances can change, which affect latency, transaction response time, and throughput. Changing the topology of the network and consolidating applications are also contributing variables that need to be accounted for. The best way to validate that the new data centre achieves the target performance, availability, scalability and security is to test it.

Q: Do I really need to do scalability projections? Can’t I just add more capacity when I need it instead of doing all that expensive testing?

A: If you haven’t done the required scalability testing during the design phase, you can’t know the limitations of your network, some of which may be inherent in the design and could have been remedied by refining the design before implementation. If the equipment and labour costs are not in the budget when your network hits the ceiling, you’ll have to live with congestion and poor performance while trying to find ways of paying for the extra capacity at premium and unbudgeted prices. Scalability testing reveals the capacity limitations of your network before you hit them, allowing you to plan for expansion cost effectively without a negative impact on your customers, and consequently, your credibility and reputation.

Q: Are the benefits of green testing that significant? Isn’t it really just kind of a popular fad?

A: With cost of energy continuing to rise, not having a strategy for managing your energy consumption does not scale in the long run. Being able to implement solutions that result in reduced power consumption and lower cooling expenses has resulted in numerous organisations saving tens of thousands of dollars per year and more. Different testing scenarios draw different levels of power consumption. Having a granular insight into this data helps you control your bottom line. Also, testing vendors for performance and power consumption helps assure that their marketing claims are verified and that all your devices work in your lab environment in the most efficient configuration possible. Finally, with initiatives in place to implement energy consumption-compliance standards in many regions, it makes sense to be testing ahead of the curve, and be in sync with the standards, rather than dealing with the costly process of catching up with them.

Q: I’m pretty sure the new network will deliver the performance we planned and which is required in our service level agreement (SLA). And if we do have problems, we can fix it when it happens. So why should I test my network against the SLA?

A: There are some lines that are too costly to cross, and an SLA is a big one. The Gartner Group calculated that the average cost of a single hour of network downtime for financial enterprises averages at about $42,000. The Standish Group found that mission critical application downtime can cost millions of dollars each minute. Timely testing is strategically cost-effective and avoids damage to your reputation as well as consequential loss of customer base and revenue.

Q: We’re leaders in our market, so we don’t really see the need to test. We’ve got a winning combination. If it’s not broken, what do I need to fix?

A: No one can take their dominant position in the market for granted. Leadership in any market is possible by virtue of having the right products or services in the market at the right time. Staying ahead of competition is everyone’s goal, and being able to deliver your products or services consistently to the market in a timely manner is really only possible if you assure that these products and services deliver the value you claim. Reliably assuring the delivery of that value is only possible before launch with a testing strategy that is neutral and reliable. Testing is one of the critical tools to help maintain your competitive edge.

Q: I am installing the best firewall devices, with the most up-to-date applications, into my network. I’m not planning to test, because my engineers assure me it will be good to go. Am I missing anything?

A: Research presented at a recent SANS conference revealed that companies who fail to get their online financial operation back up within 72 hours of a service failure may never recover as a business. Relying on the marketing claims of device vendors, as well as the skills of engineers who might be making claims outside their area of core expertise, are not strategies that assure all vulnerabilities in your security infrastructure have been critically identified and addressed before you launch. Experiencing these vulnerabilities after you launch may be too late for the life of your enterprise.