Problem solve Get help with specific problems with your technologies, process and projects.

Towers, racks and blades -- Server packaging options

When should you use a tower server, a rack server or a blade server? This excerpt from our e-book, "Windows Servers and Storage," analyzes these three types of server packaging.

The following tip is excerpted from the chapter on choosing your server from our e-book, "Windows Servers and Storage." In this tip, the authors analyze the various physical forms of server packaging such as towers, racks and blades, and explain where to use them.

Server Packaging

Servers come in a number of physical form factors or packaging, including:

  • Tower: The simplest servers, such as high-end PCs, are provided in tower cases and share similar restrictions in their extensibility.
  • Rack-mount: Larger servers often take rack-mount form. This is a standard, 19-inch wide enclosure that is some multiple of 1.75 inches tall and designed to be mounted into a larger enclosure termed a rack, which allows a site to create an appropriate mix of server configurations. A rack generally integrates some number of services such as power, storage and network connections.
  • Blade: As the technologies used to fabricate components in building server computers has evolved, those components have become more integrated and substantially smaller. This allowed a server to be built on a single board so a rack-mount-sized enclosure could hold many such boards, each with its own processing, memory, network and (minimal) storage. This led to a new form factor for servers: blades. The developing trend is that blades will encroach and take over rack-mount form factors.

Here is a brief summary of these physical form factors and where to use them:

  • The tower form factor is used for low-capacity servers where scalability beyond two towers is not needed.
  • The rack-mount form factor is used to provide servers of substantial capability and multiple nodes, with each node itself being highly capable and configurable.
  • The blade form factor is best suited for servers containing large numbers of nodes of limited capability, each of which typically builds on PC technology with its low cost and small size.

The following table further describes and compares these three packaging options.


Table 1-1: Packaging/Form-factor Comparisons

  Tower Rack                                        Blade
What it is A collection of server components (processors, memory, I/O controllers and sometimes some peripherals) integrated within a compact floor-standing unit. A 19"-wide rack into which can be inserted either servers or storage in any mix; the installable units come in heights that are a multiple of 1.75 inches ( "1U"); the maximum height of a rack is 42U. An enclosure equipped to allow a fairly large number of PCI-format cards to be installed within it. Each card is itself a computer system providing processors, memory, network connections and storage connections, a typical configuration being up to four processors, memory, and one or two hard disks for local storage.
What it looks like
Figure 1 - Example of the Tower form factor (Dell PowerEdge 6800)

Figure 2 - Example of the rack mount form factor (HP 9000)

Figure 3 - Example of the blade form factor (IBM BladeCenter)
Where to use it Use a tower system when low cost is a priority, fairly minimal configurations are suitable, scalability is limited and your need for connections to storage or the network is limited.

Physical footprint is small.

Use a rackmount system when each node has to be of fairly large capacity (in terms of number of processors, memory size, storage capacity, etc.), when the overall server configuration should encompass an average to a large number of nodes, and when scalability is important.

Physical footprint is moderate.

Use a blade system when each node has to be of reasonably small capacity (in terms of number of processors, memory size, storage capacity, etc.), when the overall server configuration should encompass an average to a large number of nodes, and when scalability is important.

Physical footprint is fairly small.

About the authors: 
René J Chevance is an independent consultant. He formerly worked as chief Scientist of Bull, a European-based global IT supplier.

Pete Wilson is Chief Scientist of Kiva Design, a small consultancy and research company specializing in issues surrounding the move to multi-core computing platforms, with special emphasis on the embedded space. Prior to that, he spent seven years at Motorola/Freescale. 

Dig Deeper on Server hardware and virtualization