The reliability of a server and the services it provides is a key factor for the success of your business. In the world of hosting, service up-time is extremely critical, as SLA’s are mostly based on service up-times.

How can you ensure that your server will perform and maintain its functions under normal circumstances, as well as in hostile or unexpected circumstances? The solution is to have your server’s components exposed to extreme stress and ensure that each component is good to use.

Why stress?

Let us see how this works in an analogous case: An athlete is given training and is subject to rigorous exercise schedules, as preparation for a sporting event. All this pressure is never seen as something that degrades the performance or affects the players sporting life-span.

When it comes to servers, the cost of failure is enormous. Adding to it, components used in assembling a server usually either fail early, or very late in their life cycle. And it is this case of early failures that is easily avoided by initially performing a “Burn-In” test. This type of test will bring to surface any intermittent or dormant problems, allowing us to use only those components that passed the test.

Burn-in is the process by which components of a system are stressed prior to being put in service. In the case of newly provisioned servers, much of it depends on what procedures are followed by your data-center. Many data-centers deploy only fresh hardware components, and others re-use old hardware. So it is imperative that you take measures to check the components before you convert your server to a production server.

How do these work:

Basic components such as CPU, RAM, hard-disks and Network Interface cards are usually tested the most.

• 1) CPU testing :CPU stability is a vital part for your server performance. The tests usually require the assistance of sensors that report over-heating; unless you have physical access to the server.

  The test uses the available core(s) in full throttle. If the hardware has an issue or if you happen to have a cooling issue, you would face a crash or instability as the outcome . These tests usually comprise of mathematic algorithms that would require very high processing power.

• 2) RAM testing :
  RAM testing algorithms usually fill the memory cells with patterns, and later checks them to see if the pattern has changed. It repeats similar operations and performing repeated checks to ensure that your RAM is indeed in good health. Time duration of test is usually prolonged, as the performance of the RAM modules is expected to vary with increase in temperature.

• 3) Hard-Drive testing :Almost all hard-drive vendor define parameters and attributes which are monitored by drive testing tools to predict a possible failure of the hard-disk. A few alarming situations would include increased heat, problems with reading and writing of data, an increase in the number of damaged disk sectors, and even at times excessive noise.

  The tools provided by the hard-drive vendor or third party tool relies on the raw values collected for these parameters while it scans the hard-disk. These values can be compared to desirable results to ascertain the overall health of the drive.

• 4) Network Interface Card testing :Here again, tools simulate maximum traffic load for the NIC and the network traffic speed and performance is reported. These tools are highly effective in the diagnosis of a network performance problem.

  Under normal circumstances, their effectiveness is questionable. However, you may rely on them for benchmarking the cards, which would help you locate issues at a later stage.

The right tools :

There are hardware testing suites available in the market. A few of them are free-ware(s), others are licensed.
However, the broad range of tools that are part of these suites are not always fully used. They are mostly heavy on your wallet and a better option would be to rely on different tools for different purposes.

A simpler method is to ask your data-center for a ‘hardware test result’ for each server that is provisioned. You could very well consider that as a point when selecting your data-center.

About the Author :

Sankar works as a Senior Software Engineer in Bobcares. He joined Bobcares back in April 2006. He loves grooming/mentoring people. During his free time, he listens to music, and enjoys singing..