Computer Technicians 301: Planning computer configurations – the right tool for the right task

Servers and workstations

The ideal server is the most expensive piece of machinery in the whole office network – these beasts will require a lot of money to plan out, set up and keep running. Unlike a SoHo machine, a server box has a much higher requirement list, but it will vary according to its role. However, some things apply for all of them – a very powerful PSU (or even multiple), high amounts of memory, high CPUs, dedicated 1Gbps PCI NICs and redundancy arrays.

A datacenter server is home to all the backups and shared data of a network, in certain cases working as a ghosting disk for all the drives on the network. To that end, it will require a big tower chassis with a high-power PSU and an attached UPS, a large amount of high-capacity drives, either Serial ATA or SCSI (often up to 8 ), any number of DVD+RW drives for rapid backups (might also be SCSI) decent cooling (without the emphasis on silence), at least 1 GB of memory, a purebred CPU, and finally, a SATA or SCSI RAID controller. Obviously, a graphics card isn’t needed, and neither is sound, or even the keyboard and mouse, once the server has been set up. The reason for all of those is as follows: a datacenter is supposed to churn out its shared data at an enormous rate, explaining the purebred CPU and large amount of memory, and the whole thing needs to run stably and without interruption, hence the PSU and UPS. However, the real kicker is the RAID controller. By connecting the hard-drives into either a RAID 10 or 0+1 hybrid field, you’re ensuring your data’s survival to a certain level, in case of bad sectors. If the RAID controller is expensive enough, RAID 5 will be an option, which is currently the ultimate in data security because it creates checksums for all data on all the adjoined hard-drives, therefore effectively negating data errors. All the drives will pull out as much as 150W out, which is why a decent PSU is neccessary.

Standalone and application servers are both created to run high amounts of applications and services, with the difference being that standalones are usually used for testing applications for external (i.e. Internet) usage. Depending on which role it’s used for, an application server will need a pure-bred single-core or dual-core CPU, or if the applications they run are very intensive, even specialized motherboards and CPUs such as Xeons and Opterons, usually 2 GB of memory or more, along with at least two disks joined into a RAID 1 array for security. Of course, a UPS is neccessary to keep it operational at all costs.

Finally, a workstation can commonly be considered the younger cousin of a server. Whire there’s rarely a need for a RAID array, what those workstations need is raw mathematical power, lots of memory for the stored data and specialized equpiment. As a very common example, a rendering station will need a dual-core CPU, 2 GB of dual-channel memory, a 1 Gbps network interface in case the used application can share rendering across the network, and a dedicated graphics card optimized for OpenGL rendering, such as an nVidia Quadro or ATi FireGL. A workstation of this sort can be used in various forms, from CAD/CAM/CAE to GIS processing and 3D design, modelling and motion rendering.

Gaming

Gaming machines are arguably the most difficult to plan out. SoHo computers go for maximum cost-reduction while retaining basic office or end-user usability, and servers are built for high data capacity and transfer rate. At the same time, gaming machines push for maximum CPU, RAM, graphical and audio performance, as well as capacity, data rate and plug-and-play ability for various 5.1 audio systems, gamepads, joysticks, HOTAS systems, steering wheels, pedals and so forth. In short, gaming machines should not be created with exclusively one thing in mind – before everything, a gaming computer needs to be scalable and upgradeable. Since most gamers know how to hand-pick components and assemble the computer, should you still be employed to construct one, you’ll have a lot more management on your hands than with either of the other categories because gamers’ funds vary wildly – some will be able to shell out enough for a CrossFire or SLI-powered machine, pulling out all stops, while others will be willing to do with less-than-ideal components. In either case, you’ll have to have done a great deal of research to know which components you’ll be able to cut back on. Depending on the gamer’s preferences, 1 GB of RAM might not be enough, but they might be able to squeeze past with just a Radeon X800, or they might require an nVidia 7800GTX, but would be willing to cut on RAM speeds. Or, alternately, might only need a small hard-drive because they stick to specific games or do regular backups. Or they play games that need fast reactions (like FPSes) and want a USB laser mouse and USB keyboard. The possibilities are literally endless, and need a lot of tailoring in order to get the best for the price tag. As per usual, a high price tag doesn’t mean quality, and neither does a specific brand. Investigate the market almost too well so you can avoid pitfalls like problematic graphic cards, and investigate your gamer client’s wishes and needs – then you’ll be able to construct a perfect gaming machine for them.

 
Written by Boris M