ANALYSIS: The Never-Ending Quest

By Chris Gloede

Does it ever seem as though you are buying new network servers several times per year? Sometimes, an upgrade will do, but you find that the machine you have is simply maxed out. You are not alone, the problem is everywhere.

Each day, someone in every IT department or business unit comes up with a new use for the network. This is great, as it makes use of the technologies at our fingertips. The down side is it's a rare occasion when the network resources exist to perform the requested function. What often results is a network with a bunch of servers, some with redundant uses, some under-utilized and some over-utilized, and little or no logical order to the deployment. What is worse is that the management of all of these servers and the costs of licensing new servers rather than upgrades can put a tremendous burden on any department's budget.

There are several ways to evaluate the value of a server, and price is NOT the most important one. Remember, ultimately, you get what you pay for. An excellent example is my recent debacle with Intel. If you were unaware of this, Intel sells servers too. As a matter of fact, they are quite inexpensive as compared to the majors, but, as usual, there is a reason for that. On receipt of my new Intel server I had some problems with the RAID 5 controller causing the server to not recognize any of the disk drives. When I called Intel for support, they were unsympathetic to my plight and ultimately suggested that I call the manufacturer of the RAID Controller that they supplied! They did not even offer a phone number! If this is support, I am the Dali Lama. On the other hand, I have experience with Digital (Pre-Compaq) calling about an Alpha. Now, they know support. Professionalism, courtesy and knowledge--they had it all. Obviously, the price difference was more than made up for in the time that I lost playing around with this thing.

I am often asked to configure servers for my clients to support the applications we are installing. Invariably, I configure systems that will last through at least one major software upgrade, yet these are rarely the configurations that are ordered. If I ask for quad processors, I get dual processors. If I ask for 1 GB of memory, I get 512MB and so on. Moreover, when I recommend a model with some breathing room, upgradeability without replacement, I rarely see that installed. Since I am rarely the supplier of the hardware, I often wonder why my clients think I am "oversizing" the equipment. While slashing my configurations may save my clients some cash in the short haul, I can all but guarantee the long-term costs will be higher, and potentially significantly higher. These costs can be measured in terms of performance (slower response time), new servers, upgrade costs, licensing costs, down time, personnel time, contractor time and so on.

Instead of running around, continuously adding undersized, overtaxed servers each time someone wants to add a directory or a service, any time taken to evaluate the current network and how this fits in is time well spent. Buy the right sized servers and buy the right name in servers. It is money well spent and usually pays unexpected dividends. Servers are not the place to skimp. Think about the costs when you are down, or unable to provide the services required to run your business.

If it seems as though you are talking to your hardware supplier more often than your own family, perhaps it is time to revisit your network architecture and your hardware sizing methodologies. If you are guilty of undersizing, underbuying or poor network utilization, take a step back. Examine the real consequences of your decisions and re-evaluate your tactics.

A veteran of the IBM midrange arena since 1983, Chris Gloede is executive VP for Business Solutions Group in Wayne, Pa. cgloede@thebsg.com.