Select server components

Overview

Image: Overview

You should already know about choosing a server application and choosing a network operating system. This resource will help you to select server components within an information technology environment.

In this topic you will learn how to:

  • identify server components with reference to required server application and server features
  • identify product specifications and limitations
  • identify and analyse system interdependencies, including hardware prior to selection
  • provide alternative solutions with reference to required server application and server features.

This topic contains:

  • reading notes
  • activities
  • references
  • topic quiz.

As you work through the readingnotes you will be directed to activities that will help you practise what you are learning. The topic also includes references to aid further learning and a topic quiz to check your understanding.

Download a print version of this whole topic: Select server components (211 KB 2766.doc)

Reading notes

Image: Reading notes

Identify components

Identifying suitable server components can be a complex and daunting task. The problem is there are many different hardware, software and networking products available, that analysing them all would take an eternity and cost a small fortune. The job is made a lot easier if we group the components into three distinct areas - software, hardware and peripherals - then analyse each area.

Software

The major restriction on selecting server components is the selection of the server application and operating system. The server application will have a set of technical requirements that need to be met in order for it to function cohesively with the operating system and supporting software. The application requirements will state the operating system version, including any updates and hardware specifications that are compatible with the chosen application(s). Managing software requirements can start to become complicated once all software requirements that are being installed on the server are compiled.

Through analysing the technical requirements of all the application and operating systems being installed, a list of minimum hardware specifications can be complied. The minimum requirements are the recommended minimum specification for installation. This is usually just enough resources for the software to work and provide basic functionality, but may not provide enough resources for the server to function properly or meet demands.

Review the requirements for IBM’s WebSphere Application server by exploring This application can support many different operating systems and hardware components. It is very complicated and requires many different components that depend on the platform selected.

Supporting software applications

A server consists of the operating system and business application, as well as supporting software for management, protection and continuity. Examples of other applications that need to be installed include

  • anti-virus software
  • firewall
  • server management
  • audit management
  • intrusion detection
  • backup and recovery.

Managing the installation of all the supporting software, the applications and the operating system involves recording the requirements for each application in a single master list for overall minimum requirements for hardware specifications.

Hardware

Most software documentation will list minimum and recommended hardware requirements. The recommended requirements are supplied by the vendor by outlining the minimum requirements for the application to function in a working environment. The hardware requirements will feature internal computer hardware including

  • processors
  • random access memory
  • storage
  • optical drive
  • network connection.

Table 1 below outlines the minimum and recommended hardware requirements for installing the premium edition of Microsoft Windows 2003 Small Business Server. The recommended requirements may not be appropriate for your particular situation. As an example, it is not advisable to have 500 clients connecting simultaneously to use the server resources available on the recommended requirements. The processing power and RAM available would be inadequate to service so many clients. It is advisable to analyse and research the role of the server and the functions it will need to provide.

Table 1: Hardware requirements

Requirement / Minimum / Recommended
CPU speed / 300 MHz / 550 MHz or faster
RAM / 256 MB / 512 MB (to a max of 4 GB)
Hard disk / 5 GB of available space (2 GB if upgrading from Small Business Server 2000) / 5 GB of available space (2 GB if upgrading from Small Business Server 2000)
Drive / CD-ROM / CD-ROM or DVD-ROM
Display / VGA (Video Graphics Adapter) / Super VGA (800 x 600) or higher
Other devices / Hardware that supports console redirectionEthernet NIC / Keyboard and Microsoft mouse or compatible pointing deviceTwo Ethernet NICs
Client workstations / No minimum listed / Microsoft Windows XP Professional or Windows 2000 Professional for client operating systems

Processors

A computer processor manipulates data and instructions from a piece of software. There can be multiple processors in a server, with each processor having a particular role. The most important processor in a server is the central processing unit, or CPU. The central processing unit can be thought of as the brains of the server. Some common CPU architectures can be found at When evaluating, remember that processor factors such as address bus, internal registers and clock speed will affect server performance.

Other processors in the server are referred to as the server chipset. The chipset connects the remaining server processors to the CPU and typically contains the front side bus (FSB), memory controllers, bus controllers, I/O controllers and power management etc.

Motherboards

Just as important as selecting the CPU or chipset is the selection of the motherboard, sometimes called the systemsboard, mainboard or planner. Motherboards are selected by form factor, referring to the size and shape of the motherboard. Common classes of form factors are mini, tower, pedestal, and rack mounted. Universally acknowledged form factors include

  • ATX
  • MicroATX
  • BTX
  • MicroBTX
  • Server Systems Infrastructure Thin ElectronicsBay (SSI TEB)
  • ServerSystemsInfrastructureCompactElectronicsBay (SSI CEB)
  • ServerSystemsInfrastructureMidrangeElectronicsBay (SSI MEB)
  • Blade Servers.

Each form factor is designed to perform specific tasks and work at different loads. Some form factor designs cater for small loads and are single socket CPUs, while others are designed for large loads and use up to 8 CPU sockets. Most server motherboards have plenty of expansion slots and room for scaling up. When selecting the motherboard, make sure you are familiar with all the specifications and limitations.

Storage

Primary Storage

RAM is the primary storage of a computer and is used to hold program code and data during computation. RAM and the CPU have a major influence on server performance. The two main types of RAM are static RAM and dynamic RAM. Static RAM (SRAM) contains it contents as long as there is electrical power. Alternatively, dynamicRAM (DRAM) needs to be constantly refreshed to retain its data. SRAM is mainly used in CMOS, while DRAM is used as main memory.

Other important features of DRAM to consider are listed below:

  • Registered memory – increases the accuracy of data access
  • Error correcting code – type of memory that includes special circuitry for testing the accuracy of data as it passes in and out of memory.
  • Form factors – describe the pin configuration and size of a memory module. Common form factors include
  • SIMMs – Single Inline Memory Module
  • DIMMs – Dual Inline Memory Modules
  • RIMMs –the trademark name for Direct Rambus memory module
  • DDR DRAM – Double Data Rate synchronous DRAM allows the doubling of the bus clock rate by performing two transactions on one clock cycle.
  • EDO – Extended Data Out DRAM speeds access to memory locations by as much as 10%.
  • ECC – Error Correcting Code memory tests for and corrects memory errors automatically.
  • Registered/Buffered –has a temporary holding area or buffer for data travelling to and from the CPU to RAM, used to increase systems reliability.

For more information on server storage, go to

For more information about Intel Dual-Channel DDR Memory

Architecture, read the White Paper at

For more information about RAM, go to

Secondary Storage or direct attached storage

Secondary storage is used to hold long term data and requires the use of I/O channels to store and retrieve information. A common form of secondary storage is the hard disk. There are many technologies you must be familiar with before selecting secondary storage for a server. Whichever technology is chosen, it must meet the performance, technological and business requirements. When selecting disk based storage options, consider the following:

  • Spindle speed – the higher the spindle speed, the better the disk performance. Current spindle speeds are 5400, 7400, 10000 and 15000.
  • Cache buffer size – as the disk cache buffer stores accessed data, it improves performance. Accessing cached data is much faster than accessing data from the disk platter.
  • Internal media transfer rate – is the rate the hard disk can read/write bits from the disk platters and is expressed in Megabits per second (Mb/s).
  • Interface type – the main interface types for servers are
  • EIDE
  • SCSI
  • SAS
  • fibre channel.

For more information on server storage, go to and review the references at the end of this topic.

Network Storage

Network storage is storage that is accessed via a network connection. Storage Attached Network (SAN) and Network Attached Storage (NAS) are two common forms of network storage. The purpose of network storage is to centralise data storage into one or more manageable areas. Having a central area for storage allows better utilisation of disk performance. Disk space can be rearranged, reallocated and easily increased without the need to reconfigure or upgrade the server. Centralised storage simplifies data protection and security, and reduces costs and failure rate through consolidation. Network storage is a solution usually associated with large enterprises that consist of many servers.

To learn more about SANs, go to and complete the tutorial on SANs.

For more information about NAS, go to and read the article ‘Introduction to NAS—Network Attached Storage’.

Chassis

You may be limited by your choice because of the motherboard form factor and mounting points. Before selecting the chassis, determine the number of internal drives and expansion slots required and if the server should be rack mounted, wall mounted or a desktop.

The different types of chassis are

  • Pedestal - looks like an ATX chassis but is designed to fit the larger form factors of SSi-EEB.
  • Rackmounted
  • Wall mounted rack
  • ATX
  • BTX.

Thermals

Thermals play an important part in keeping the server running. A build-up of heat can adversely affect server reliability, causing component failure or in extreme cases combustion and fire. Major sources of heat include

  • the processor
  • memory modules
  • hard disks
  • chipsets.

Heat sources of more concern include power supplies, high speed disks and processors. To reduce the amount of heat, cool air must circulate correctly through the chassis. By using exhaust fans working in tandem, air should travel in one direction.

Most chassis manufacturers have incorporated multiple mounting points for fans. Fans come in a range of sizes and ratings. Common sized fans are 40, 80, 92, or 120 millimetres and ratings of 12-400 CFM (cubic feet per minute).

Processor heat sinks

There are two types of processor heat sinks: those with a fan and those that are fanless. Fanless heat sinks rely on chassis cooling fans to extract heat, so special consideration must be taken when designing air throughput. Heatsinkswithfans have the fan mounted on top or on the side.

Electromagnetic interference and electrostatic discharge

All electronic devices give off electromagnetic interference (EMI), which is radiation caused by electrical or magnetic activity. EMI is also known as electrical noise. The problem with EMI is that emissions from one device can interfere with other devices and potentially cause problems. EMI can lead to such things as data loss in memory, distortion of the display on a monitor, as well as other problems on a server or for network clients.

A server may be affected by EMI emissions generated by itself, as well as EMI emissions generated from other devices. If a server is located close to a noisy device, this can directly affect the server.

Servers and PCs do not typically cause much interference to other devices because the Federal Communications Commission (FCC) requires them to be certified as Class B devices. This FCC certification is used to verify that a system conforms to standards that limit the amount of EMI that a computer can produce. EMI is contained by the metal lining of the case and cover of a server or PC. This limits the amount of EMI that escapes from the chassis.

More common, though, is electrical interference that iss transmitted over electrical power lines. EMI problems are not all that common, but a server affected by EMI can be difficult and frustrating to troubleshoot. EMI problems can be avoided by spacing the devices in a rack about 5cm apart (see the manufacturer’s recommendations for rack spacing). Another method for reducing EMI is to avoid placing a server next to a pop machine or a power transformer.

There are many ways you can protect your server from electrical interference. Three of the most important methods are listed below:

  • Physical isolation: Systems should be kept at a reasonable distance from one another. Server rack mount cabinets usually limit your ability to spread devices apart too far. Typically, a rack bay allows around 48 rack units (approx 5cm to a rack unit). Depending on the size of the server(s), additional components such as UPS, patch panels, switches and routerscan be installed. These components together with the spacing between devices and cables can often be a tight fit. Fortunately, most rackmount servers are designed to withstand EMI from the devices above and below it. Very thick metal chassis - especially on the top and bottom - help to prevent EMI radiation from escaping or entering the server chassis.
  • Use dedicated circuits: These employ separately grounded and isolated power circuits which are dedicated for use in the server room. Separating the server room circuits from others in the building removes the threat of interference being passed to the server room equipment from other electrical devices. For example, avoid having servers share the same circuit as the tea room vending machines or air conditioning units. Isolating the server room electrical supply has the added benefit of improving the quality of the power used by the equipment in the server room.
  • Install a UPS: The use of an uninterruptiblepowersupply (UPS) is essential in a server room. A UPS filters out interference caused by other devices that share a line with your servers and also provides back-up power in case of an unexpected power interruption.

Security

Physical security is important when considering chassis specifications. If the server is located in an area that is easily accessible, you’ll need to protect the server and its components against theft. Depending on the manufacturer, there are a few ways this can be done. These include

  • cablelocks which enable a computer with a security lock loop to be locked in place
  • solenoidlocks which attach to your PC and act as a hood sensor and cover lock (allows chassis to be locked and unlocked remotely over the network)
  • keylocks which are located on the front bezel, enabling the components to be secured.

Network interface cards

Network interface cards can make or break your server performance. The last thing you want is a multiprocessor server that breaks all benchmarks using a 10MB five dollar NIC. The most common type of NIC protocol is Ethernet. Ethernet comes in a variety of types that require different interfaces. Common Ethernet server standards are

  • 100BaseTX
  • 1000BaseT
  • 1000BaseF
  • 10GbaseT

Data bus

A data bus is a type of communication path which exists between computer components and is usually controlled by the system chipset. There are many different data busses within a computer. Common busses are the systems bus and I/O bus.

Systems bus

The systems bus, sometimes referred to as the frontsidebus (FSB), is data path between the CPU and the systems board chipset. It connects the CPU with main memory and any other system such as hard disk, AGP, PCI and USB. The process operates at a multiple of the systems bus. For example, if running a processor at 3.2Ghz and 800 FSB, the clock multiplier will be set at 4 (3.2 / 0.8 = 4). By changing the FSB clock multiplier it is possible to over-clock or under-clock the processor.