Someone once said you can never be too rich or too thin. Too rich is not in my future, so I thought I would talk about too thin this month thin servers, that is. Blade servers, which are essentially servers on a card, are starting to garner some interest and sales. I saw a research report (from Gartner) that predicts blade-server sales would jump from 84,000 units in 2002 to over one million in 2006. OK, I know I gave them a double yawn last month, but I suspect that thin is here to stay. I mean most servers are built on a desktop hardware model. That PC paradigm has been around for over 20 years. It is based on an open extensible architecture that allows a system to be extensively modified to suit any particular user or purpose. Isnt it time we started looking at building servers that are optimized to do only what they are required to do?

What Are They?

Blade servers really are servers on a card. A blade server consists of a chassis that is capable of hosting multiple blades. Each blade is a fully functional computer system on a mother board with a processor, memory, network connections, and usually some sort of local disk storage. The blades plug into a chassis that provides the necessary infrastructure, such as power supplies, cooling fans, cabling, networking, video, and keyboard interfaces. The chassis typically is designed for standard mounting racks with a form factor of 3U (5.25) or more. (The term form factor is a generic computer term used to define the physical shape and size of a device.) The first obvious advantage to these things is their small footprint. It is possible to mount a half-dozen servers (six blades and a chassis) in the space often used for a single standalone rack server. There is a wealth of benefit to be derived from this model. Right now I use a minimum of seven cables for each working server (two power, video, keyboard, mouse, two Ethernet, etc.). Imagine running six servers with the same number of cables! Still, there is a lot more to this picture than cables. Lets take a look.