When someone mentions servers, the mind conjures up images of a stuffy room with huge machines installed wall-to-wall, humming methodically. Big or small, stored in dingy or airy sweet-smelling rooms where journalists come to inspect the IT implementation in your organization, they are the ones that make life possible at the front-end touch points. Irony of fate, the reader might say, but while working on this story, the servers (rather the lifeline of our organization) crashed for a brief period and from a calm serene working atmosphere, we had people tugging their hair out and cursing the fast-looming deadlines. Chaos reigned as people were unable to access the network and the data stored on their drives. Floppies were in great demand and the only question being flung back and forth across the DQ floor, in the canteen and at the dhaba outside, was—"When will the server be up and running? When will our comps function normally?" Though it was a disguised respite from the routine stuff, people seemed more than anxious to get back to work.

Imagine the same scenario re-enacted in a bank or financial institution. The financial loss is not entirely fathomable, but the lack of face that ensues is irreparable. And the wizards who wield the wand that can make or break life in a computer-dependent atmosphere are the back-room men—those who play around and fix servers... the ones at whose mercy our sanity at times lies.

But no more drab thoughts on the importance of servers in this tech-savvy age. It is a foregone conclusion that an organization should make a prudent choice while shopping for servers to take care of their information and workload needs—both present and future. Given the multitude of configurations available in the market, there are some simple rules to decide on the type of server that will fit or cater to your organization’s needs.

Entry-level servers: If your requirement is to take care of the e-sharing, messaging (e- mail), and print-sharing tasks, then look for servers with the lowest price points. However in this case it should be kept in mind that scalability could be a problem in the long run. The typical minimum specifications (specs) should include an Intel Pentium 4 processor/up to 533MHz front-side bus speed, up to 2GB maximum memory, at least one HDDs, up to three SCSI or IDE. And of course some manageability tools.

Mid-range sfdervers: A mid-range server would fit your bill if you are looking to perform multiple functions like departmental applications, files, and print and can exist in a cluster configuration. The usual specs for these range of servers include up to four Intel Xeon processors, memory ranging from 512MB-12GB, Hot-pluggable: 6 x 64-bit/100MHz PCI-X (supports 3V or Universal PCI Adapters), up to 10 high-speed SCSI drives in a RAID configuration, Hot-pluggable hard drive bays and multiple external components, including Fibre Channel SANs (storage area networks), or other attached storage devices. Also, these would have high reliability features.

High-end servers: If you are looking at mission critical applications, and a 24/7 uptime, then these fit the bill. They are expensive but are useful for organizations depending on a large database, data warehousing, and e-commerce applications.

These highly scalable workhorse servers are usually found in airlines, banks, government agencies, large retail distribution companies, and universities. This category offers quite a few choices. If you are looking at the Intel based servers, then they are quite a few to choose from.

Apart from IBM, Hewlett-Packard and Dell, you will find a host of Indian companies like HCL Insys and Wipro ready with their offerings. However, if you are looking at the RISC-based servers, then the choice is limited to a handful—IBM, Hewlett-Packard and Sun Microsystems, and Indian players merely distribute and integrate the same.

Quick thoughts
Once you have decided on the apps and the type of servers that best fit your requirement, it’s time to go shopping. Obviously, the first consideration would be the existing relationship with the vendor. For example HDFC Bank has systems from Sun as it started out with and developed the expertise on the Sun platform. So buying a new system—the choice obviously being Sun given past experience—was an easy job for the IT head at HDFC Bank. However, even when you share a very close relationship with a single vendor, it’s still not a bad idea to explore other options. You still have a budget to think about.

According to studies by the Meta Group, "The greatest factor contributing to aggressive server vendor discounting is the extent to which vendors perceive a threat from competitors. In the current buyer’s market, we have seen discounts of up to 50% for single high-end Unix systems." The studies add—"It is not a bad idea to introduce generic third-party systems management tools (e.g. CA), which would be common across hardware platforms."

Another aspect to conveniently ignore are benchmarks. Since each vendor will produce specific benchmarks where his product tops the charts and the results do not portray real-life situations, it should hardly have an impact on the buying decision. Talking of real-life situations, a better test is to run the server continuously for 72 hours with specific apps. If it works fine for the given time-frame, it will work for the next couple of years without glitches.

Trends
One thought that reigns in the minds of CIOs across the globe for server purchases is whether the system should be based on UNIX or Windows. Till recently, Unix-based ones were the mid-range to high-end systems—the de facto servers for mission-critical applications. Windows and Intel were relegated to the low-end server category. However, the Wintel combination has grown much beyond expectations. Of worldwide server purchases in 2001, 88% was made up of what IDC calls ‘Standard Intel Architecture Servers’. This was a market dominated only a few years ago by proprietary systems like the IBM 3090, Digital Vax, and the Hewlett-Packard 3000.

Call it the effect of standardization or what you may. But the question that arises is why has this trend not moved to the high-end lot also? According to IDC, it is because of the perception that not all software programs, databases and functions are available or currently suitable for these devices. Will the same continue or are there more options available to the CIO? Linux could prove to be the answer given the fact that most of the Unix vendors are also supplying the Linux version.

Then again, the price of these systems could take a beating, or Intel-based systems could move up the value chain. This will lead toward greater standardization, as research reveals that the significant benefit of using standardized products is the lower cost of the product itself, lower software costs as well as the lower cost to deploy and operate. It boils down to easier deployment and lower cost of services.

Yograj Varma