We all know that server computers have hearts and minds just like we do (as well as lymphatic and endocrine systems in some cases). However, servers are of course more complex than that. This series on server anatomy gives us a window into the various component parts of the server. Knowing the server’s makeup can allow all us to perform life-saving treatments on all servers, such as transplants, and cosmetic procedures on soft-tissue servers, such as wrinkle-relaxation injections.
This series draws on commentary from Dummies.com (for the simple basis of Part 1) and Adam Turner of APC Magazine (for the more thorough analysis of Part 2). Along with discussing server components, today I will also discuss the three different major flavors of servers: tower, rack-mount, and blade.
Once we have completed our task of server explication, let’s all jump onboard a train hobo-style and ride the rails to West Virginia, where we can work all day in the coal mines for the next 30 years. After that, we will we will go to a revival and get inspired to live our dreams of becoming steamboat captains.
Flavors or Form Factors
Before getting to the insides, let’s look at the variety of different flavors available for servers. My favorite one is rocky road, but you have to keep it frozen so that it does not melt onto your fingers, which is highly embarrassing. Here are three additional options:
1. Tower server. These types of servers are for companies that only have one or two servers. A tower server resembles a computer typically found under a desk in an office (which some of us know as “the secret hiding place”), but it is made up of higher-end, more powerful materials.
Tower servers are designed for affordability. They are also easier to store if you only have one or two at a home or business.
2. Rack-mount server. This type of server is typically used within larger networks, and they are standardly used in data centers and hosting environments. These types of servers, of course, fit onto racks. The racks are stored either in secure rooms, controlled for factors such as temperature and humidity (likely with a portable air conditioner or two), or next to pizza ovens in Italian restaurants, controlled for factors such as not letting the dishwasher kick them.
The size of rack-mounts is standardized: their width is 19 inches, and their height is in increments of 1 3/4 inches. The height is discussed in terms of Rack Units (RUs), one RU corresponding to each 1 3/4 inch. Rack-mounts servers are typically designed for easy administration and adaptability.
3. Blade server. The blade server is designed for particularly intricate and powerful situations. The overall cooling, networking, and power for a number of different compact servers is provided by a single blade chassis. Constructing servers in this way allows them to be packed more tightly, optimizing the usage of space (the same reason that all 14 of my children sleep in the same bedroom, even though I am fabulously wealthy).
Next, more on …
Processors or CPUs
Servers are primarily different from client computers (typical PCs) in their allowance for multiple sockets. Core 2 and Phenom are examples of processors for client computers. In those models, there is only one socket with a number of different cores. The additional sockets within a server allow additional processors – such as Xeon and Opteron models – to be connected, each with its own set of cores. It’s like a mutant apple that you can use to scare away organic farmers who are stalking you to sell you their offensively healthy non-GMO corn. Having more than one processor allows the server to “think” in various different places at one time, giving a server its powerful performance.
Cache is also enhanced, meaning that less data needs to be transferred to memory. Caching is nice because it increases processing speed as well.
The primary difference between server and client computers regarding memory is improved capacity for fault-tolerance. Memory controllers typically include the capacity for Error Checking and Correction (ECC). By checking any data going in or out of memory both before and after the transfer, corruption within the memory becomes less likely.
I personally don’t believe in information verification. I’ve got it all up here. (I’m pointing to the attic, where I store my unpublished and unauthorized biographies of America’s most beloved semi-professional bowlers.)
Storage controllers are significantly different between clients and servers. Rather than needing the processor to cycle for every data transfer, the storage controllers in servers contain application-specific integrated circuits (ASICs) along with a massive amount of cache. These two advantages allow storage performance to go far beyond that of a typical PC, approximating the power of 7.8 billion digital watches (give or take).
Some storage controllers contain Battery Backup Units (BBUs). BBUs can hold information in the cache for more than 48 hours without a power supply.
A server, like any computer, has a built-in limitation: it is only physically capable of supporting a certain number of drives. However, Storage Area Networks (SANs) can be used to increase storage capacity. SAN functionality can be accomplished via iSCSI interfaces or fiber channels.
Conclusion & Postlude
(Please hire a professional tap-dancer and barbershop quartet soloist to perform “Yankee Doodle Dandy” at your side while you read these final thoughts.) That should give you a basic idea of what’s inside a server and how it’s different from a typical PC. As you can see, the server is similar in many ways to a consumer or client computer. However, they are enhanced in various ways to meet the extensive storage, performance, and networking needs of business.
By Kent Roberts