Some Thoughts on PC Hardware
Now that we've discussed NICs and some of the connectivity devices you will run into when you network computers, we will finish out the chapter by looking at some other critical pieces of PC hardwarespecifically, motherboards, processors, hard drives, and RAM. We need to discuss these hardware components because they inherently affect a PC's overall performance. Understanding how these hardware components fit into the overall PC hardware puzzle will help you when it is time for you to select hardware for network clients or network servers. Since this book serves as a primer to networking and network hardware, you should consult a heavy-duty reference such as Upgrading and Repairing PCs by Scott Mueller, published by Que.
First, let's take a look at some issues related to motherboards. Then we can take a look at processors, hard drives, and memory.
No matter what operating system you use, computers use pretty much the same strategy to identify themselves on the network. This strategy involves broadcast messages, meaning that a message will go out to every other computer on the network, proclaiming a computer's identity or requesting information related to the identity of another computer on the network. Broadcast messages create what is called broadcast traffic, and broadcast traffic can suck up valuable network bandwidth that otherwise could be used to move LAN data. Routers keep broadcast traffic from spreading beyond a particular segment, thus conserving a lot of potential network bandwidth.
The motherboard is the main system board for a PC, and it provides the data highway (or bus system) that moves data between components on the motherboard and the PC's processor. The speed of data moving along the bus is measured in megahertz (MHz). Although 66MHz was once considered fast for the bus speed of a motherboard, motherboard manufacturers quickly provided motherboards with bus speeds of a 100 to 133MHz. Motherboard bus speed has continued to increase with each new motherboard and chip set. There are motherboards now available that provide bus speeds in excess of 500MHz.
At one time the motherboard would never have been considered the limiting factor in a PC configuration but with advances in PC memory and processors, a computer with an older motherboard may only provide few options as far as upgrading; meaning newer memory modules and processors may not be compatible with a legacy motherboard.
Although how fast a computer can work with data is really ultimately decided by the computer's processor, deciding on a particular motherboard should probably be determined, in part, by the features the motherboard's chipset provides. The chipset determines the ceiling for the bus speed. Chipsets provide DMA support (discussed earlier in this chapter in the section "Selecting a NIC") and other features, such as an Accelerated Graphics Port (AGP provides faster video response). These features will also boost the overall performance of the computer.
The processor is really the brain of the computer. It processes binary information input by the user or received from other devices on the computer, such as the network interface card. Processor speed is measured in megahertz (MHz). The Intel processor used on the original IBM PC ran at 8MHz. Processors are now available that run in excess of 2,000MHz (that is, 2GHz or gigahertz).
Processors are manufactured by a number of different companies, including Intel, Motorola, Cyrix, and AMD. The selection of the processor type and speed, as far as networking is concerned, will depend on whether you are configuring a client machine or a network server. Faster (and even multiple) processors are a must on a server that needs to process a large number of calls for data from users on the network.
Basically, you should choose the processor you want to use for a computer and allow that choice to determine the motherboard you use for the computer. This helps narrow the choice of motherboards down to those that support the processor.
There seems to be as many memory types for personal computers as there are types of apples. Computer memory or RAM (Random Access Memory) is the working storage area. It is used by the processor and other devices to temporarily store information, and it's also accessed by software as we work on our computers.
You can't just install any type of RAM on your motherboard; it is actually dictated by the chipset the manufacturer places on the motherboard. This means you need to know the type of RAM that is compatible with a particular motherboard.
RAM comes in two basic formats: SIMMS and DIMMS. A SIMM (Single Inline Memory Module) is a epoxy-coated silicon wafer that contains a number of memory chips. The SIMM fits into a slot on the motherboard. SIMMs must be placed on the motherboard in identical pairs (there are typically four SIMM slots on a motherboard). This means that to have 64MB of memory on the motherboard, two 32MB SIMMs would be installed. You will only run into SIMMs if you are working with an older computer that has a legacy motherboard. Newer motherboards support the DIMM.
A (DIMM) Dual Inline Memory Module contains memory chips like a SIMM but actually provides a greater density of memory chips and therefore more RAM. DIMMs do not have to be installed in pairs, and there are typically three DIMM slots on a motherboard. This means that a computer with 256MB of memory would only require the installation of one 256MB DIMM.
RAM speed has been historically measured in nanoseconds (ns). The lower the nanosecond rating for the RAM, the faster the RAM. For example, 10ns RAM would be faster than 30ns RAM. The speed of newer RAM memory is now measured in MHz (as is the motherboard bus speed and processor speed).
As mentioned earlier, a number of different RAM types are available, and the type used is dictated by the motherboard used on the computer. Although memory has changed dramatically since the advent of the PC, the RAM types listed here are all types you might find on a motherboard designed for a Pentium processor (clone processor motherboard):
Fast Page Mode (FPM) memory. This type of RAM is mounted in SIMM modules of 2, 4, 8, 16, or 32MB and is the traditional RAM type. FPM RAM is typically found in 60ns and 70ns versions. You cannot mix different speeds on the same motherboard.
Extended Data Output (EDO) RAM. This RAM type is an improvement of FPM RAM that provides for the faster reading of data. EDO RAM is usually sold in 60 and 50ns versions. The 50ns version is available at a higher cost. EDO RAM is mounted in SIMM modules.
Synchronous Dynamic RAM (SDRAM). This RAM type is still considered the standard for desktop computers but is rapidly being replaced in new computer systems with DDR-SDRAM (discussed in a moment). SDRAM is typically referred to as PC100 or PC133 RAM, depending on its speed. SDRAM comes in DIMMs and has an access time of only 12 to 8ns (PC100 and PC133 RAM, respectively). This type of RAM is called synchronous because it's actually able to synchronize itself with the speed of the motherboard.
Double Data Rate-Synchronous DRAM (DDR-SDRAM). This type of RAM is actually able to transfer data twice during the motherboard's clock cycle (which is measured in MHz). This means that this type of RAM can be twice as fast as other RAM types. DDR-SDRAM comes in DIMMs and can provide speeds of 400MHz. This is the most commonly used RAM type on new computer systems.
Rambus DRAM (RDRAM). RDRAM was developed by Rambus, Inc. RDRAM is a very fast type of DRAM that uses a two-byte wide data channel to move data at high speeds. Rambus memory sticks are referred to as RIMMs because they look similar to SIMMS. Rambus memory is becoming the standard for Pentium 4 motherboards and is available on both desktop and server computers.
While the type of memory you use can enhance PC performance, the true bottom line is that the more RAM you have on a system the better the system's performance. Adding RAM to any computer will increase system throughput. The addition of RAM, in many cases, will even provide more of a performance jump than switching to a faster processor.
The first hard drive available for the original IBM PC boasted a capacity of 10MB (10 million bytes). Now it is common for hard drives or fixed disks to have capacities in excess of 40GB (that's 40 billion bytes). In fact, a 40GB hard drive is now considered quite small in comparison to the large capacity drives that are available.
Hard drives come in two flavors: IDE drives and SCSI drives. An Integrated Drive Electronics (IDE) drive is a hard drive (or other device such as a CD-ROM) where the controller for the drive is built in to the drive itself. An IDE drive is connected to the motherboard using a ribbon cable. Each motherboard IDE connection supports up to two IDE drives. Motherboards typically have two IDE connections, meaning a maximum of four IDE drives (this includes hard drives and CD-ROM drives) could be installed on the computer.
The newest drive standard available is Enhanced IDE (EIDE). It provides faster data rates and can support larger storage devices than the IDE standard.
Small Computer System Interface (SCSI) hard drives are attached to a SCSI controller card (placed in one of the motherboard's expansion slots or built right onto the motherboard, in the case of servers). SCSI controllers allow for the attachment of more drives (up to seven); therefore, SCSI drives are pretty much the standard for server computers.
In terms of server hardware, SCSI hard drives are preferred over IDE drives because SCSI drives boast a 12% performance boost over their IDE counterparts. Also, only two IDE drives can be connected to the motherboard via the same ribbon cable (limiting a computer to a maximum of four IDE drives connected to the motherboard). On the other hand, one SCSI controller will support up to seven drives, which makes it easier to deploy multiple-drive configurations, such as a RAID array (RAID is discussed in the section titled "Working with RAID," in Chapter 18, "Protecting Network Data").
Network Server Considerations
Network servers require large storage capacity, fast processors, and ample memory. Server tower cases also are larger than client computers and provide much more space for additional hard drives and other peripherals.
Network storage capacity is no longer limited to the hard drives you have on your network servers. The recent introduction of network attached storage (NAS) provides a way to add storage capacity to a network without adding an actual server computer. An NAS device is really just a box of hard drives that is directly connected to the network. NAS devices, such as the Quantum SnapServer, run their own proprietary OS and can be configured for file sharing using remote management software that can be run from a network client or server. NAS devices don't need a monitor or keyboard because you remotely configure them over the network. NAS devices provide an easy way to add hard drive storage capacity to a large network or even a peer-to-peer network without the expense of a network operating system and dedicated server hardware.
Server computers often have motherboards that provide space for more than one processor. Most network operating systems are built to take advantage of multiple processors. In fact, some network operating systems support up to 32 processors.
Servers also typically contain many more hard drives than you would find on a client machine and usually contain SCSI hard drives rather than IDE drives (which you would typically find on a client computer). This is not only to provide adequate storage space but also to build in some redundancy for storing files. We will talk about how drive redundancy can be used to help keep important data safe in Chapter 18. The SCSI controllers are also typically part of the server's motherboard rather than an add-on card.
The hardware configuration for a server will, to a certain degree, be dictated by the network operating system running on that server. Each network operating system provides a baseline configuration for a server that will run that NOS. We look at network operating systems in Chapters 8, 9, and 10; where we discuss Novell NetWare, Microsoft Windows Server 2003, and Red Hat Linux, respectively. You will see that each of these network operating systems have recommended baseline hardware server configurations (detailed in each of the chapters).
Network operating systems are also generally less forgiving when you attempt to run them on computers with odd configurations or atypical hardware. NOS software companies often provide a hardware compatibility list that allows you to view what types of server hardware have been tested with the particular operating system.
Before you actually finalize your server configurations, you need to do some research on the NOS you will be running. Novell, Microsoft, Red Hat, Sun Microsystems, and other NOS providers typically provide white papers and other material that allow you to look at case studies and performance data related to a particular NOS and various hardware configurations.
Servers often will contain drive bays that contain "hot-swappable" drives. This means a drive can be added or swapped while the server is running. These drive bays allow you to access the drive by sliding the drive out of the case; it's not unlike opening a drawer. This means that you can deal with drive issues without powering down the network server or opening the server case.
Network Client Considerations
Network clients don't typically need the processing power and storage capacity required by a server computer. Network clients do need, however, to be able to properly run the client operating system they have been configured with. Be sure that a client machine has at least the minimum hardware configuration to run the client operating system.
As with network operating systems, client operating systems will also often have hardware compatibility lists that provide you with a listing of the type of computer hardware that has been tested with the OS. As with servers, the more robust the hardware configuration of the computer, the faster (and better) the OS will run on the system.
Microsoft, Red Hat, Novell, and Sun all provide the hardware requirements to run their network operating systems. Check out these sites:
A Final Word on Computer Hardware
To buy or to build? That is the question. Although the very low price of computer components may seem like a fairly seductive reason to build computers for a network, prebuilt PCs are configured with components that the computer manufacturer knows (or should know) work well together. Computer manufacturers also provide warranties and service plans. And in most cases they also provide excellent support. So, why build PCs? Adding RAM or hard drives to a PC is one thing, but building a bunch of computers from components may be nothing more than a major headache.
With computer hardware prices continuing to drop, you can purchase PCs from any number of computer manufacturers that will provide you with excellent desktop and server systems. Do your research before buying and make sure that you configure your servers and desktop computers to take care of the job for which they were intended.
The Absolute Minimum
In this chapter we took a look at networking hardware devices such as the network interface card. We also took a look at networking connectivity devices such as hubs, switches, and routers. Computer hardware considerations were also discussed.
Network interface cards (NICs) supply the connection between the computer and the network's physical medium. The NIC supplies the MAC or hardware address that is used to identify a computer on the network.
Internetworking devices are used to connect network computers and other devices. Hubs, bridges, switches, and routers are devices that provide different strategies for connecting PCs and managing (or in the case of a hub, not managing) the network bandwidth.
A motherboard is the main circuit board for a PC, and it provides the connection point for the processor and peripheral cards. The motherboard also supplies the data bus used to move data from various devices to the computer's memory and processor, and vice versa.
Network servers require more processing power, RAM, and storage space than network client computers.