How does a blade server differ from a normal rackmount server?

This may be a question but I thought I would ask even so.

Solution 1:

Blade servers are small, high-density, low form-factor computers, designed for maximum power in a small space.

A blade server is mounted within a chassis, and the chassis typically takes on a lot of functions and parts that were previously done by the individual host. The chassis itself will hold the power supplies (resulting in less wasted power from conversion), the fibre-channel cards, the network controllers, SCSI interfaces, and so on. The servers themselves contain only the non-unique parts – CPU, RAM, and hard drive.

The benefits are that you can pack far more computing power into a rack, you have homogenous hardware, and your management is simpler. Instead of monitoring two power supplies, NICs, etc. per server, you only have to watch one set of hardware. Because your hardware is homogenous, you can keep spare parts around without worrying about which model of power supply you need, and so on.

The downsides are the up-front costs. Because you need to buy the (very expensive) chassis, it's cost-ineffective to start with one blade and work your way up. You would typically buy blade servers if you plan on buying a large number of servers at once. Otherwise, the initial up-front cost is excessive.

Solution 2:

A blade server can be considered a rack-mount server that has a variable number of actual servers within it. When you buy a rack-mount server, you are buying a single server. A blade can be expanded as you get more money, buying additional blades without requiring more rack space. For some uses, this gives you flexibility.

A blade takes several "U" of rack space. One blade server we have can host up to eight blades, and I believe it's a 4-U rack-mount box. Thus, blade servers are often higher density than traditional rack-mount servers.

Before you install a bunch of blade servers, ensure that you have the cooling and electrical ability to handle them.

Solution 3:

Blade servers are high density boxes. In addition to typical rack servers, they may contain additional management tools. Also, blade servers are highly modular allowing for the substitution of individual blades (CPU, memory, controlers, etc). Additionally, some blade servers run with disks on back-end SANS.

Solution 4:

With blade you'll generally find as many management functions as possible are centralised so that multiple blades can be dealt with via a single interface (see HP's Onboard Administrators/iLO) for this.

Also, generally, a blade server will offer better power/heat characteristics than a pizza-box servers.

Also of very great importance is the lack of traditional internal PCI/PCIe adapter although many blades have their own proprietary adapter mechanism to allow for lightly-modified PCIe adapters to be created to fit inside. For instance Emulex and QLogic both make manufacturer-specific FC HBAs that fit inside IBM, HP and Dell's blades.

I'm more than happy to help you with any other questions, just ask.