Does ̶N̶̶O̶̶T̶ Compute: A simple guide to servers for non-IT professionals
"Everyone knows what a server is, don’t they?" asks a member of Bytestock’s own team in disbelief.
Here at Bytestock HQ, we’ve lived and breathed IT hardware for so long it’s easy to forget that not everyone knows the difference between SATA and PcIE connection. Nor should they!
That being said, servers are such an integral part of modern business functions that having even a basic level of knowledge can help an entire team come to technical and purchasing decisions far more effectively.
What is a server?
Simply put, a server is a powerful networked computer which allows for information to be shared. Technically speaking, any computer can operate as a server, however, the term usually refers to the large, function-specific machines that run continuously in offices and data centres across the world.
There are endless types, from web servers which tell your browser how to load a web page, to high capacity storage servers with hundreds of terabytes of hard drive space. For many companies, servers are the central management point for business-critical functions, such as CRM databases, finance platforms, stock management systems and ecommerce websites.
In general, the more information a network has to store or move and the more users needing to connect to it, the more servers it requires to manage the levels of data transfer. A small company may make do with a single server to cover multiple functions, whereas the finance department of a multi-national corporation may have a whole network of dedicated servers.
What are the benefits of a server?
A server or server network allows information to be shared between locations and colleagues incredibly easily. For teams working across multiple projects, being able to do so can have a huge positive impact on efficiency and productivity as it allows everyone to work from the same up-to-date information.
Central servers are also a far more secure approach to storage. By having company data stored and accessible in one controlled place, the need to duplicate documents, send via email or store on portable drives is reduced. It’s an essential safeguard if you have staff who work remotely or are often travelling. After all, who hasn’t lost a USB flash drive?
If you’re a company of any more than one, a server makes managing your business far easier.
On-premises vs cloud servers
Remote or “cloud” servers have become increasingly popular, providing all of the storage and accessibility advantages of on-site servers without the need to accommodate the actual hardware, nor the concerns of maintenance and running costs.
For smaller business, who perhaps don’t have dedicated IT resource among their employees, cloud servers can be the ideal option. Usually billed as a monthly subscription, setup is incredibly easy, as is upping storage space as requirements change. Up to a certain level of storage, cloud solutions present great value for money.
Growing and larger businesses often prefer the advantages of on-premises servers, nearly always having the dedicated IT manager/department necessary to maintain the company’s overall IT infrastructure.
As long as the space and resource are available, developing an on-premises network can eventually become the more effective solution in the long-term, especially if the storage requirements are substantial.
On-premises servers also offer an additional level of security - all data is retained and managed centrally, without the potential access, control or breach concerns presented by a cloud solution (though, such instances are very rare).
Tower servers are where most smaller businesses should begin. Comparable in size to a normal desktop computer unit, they’re the perfect all-in-one solution for companies who lack the space and budget for larger alternatives.
Being geared towards storage, tower servers often have room for a good number of hard drives but usually lack the increased memory (RAM) and processing specs of rack servers. But that doesn’t mean they’re not powerful! Generally speaking, a mid-range tower server should serve an office of 5-10 people perfectly well.
They’re also pretty robust and don’t require any great deal of maintenance to keep running.
Rack servers, as their name suggests, are held within the racking bays of server cabinets - the tall ‘cages’ you may recognise if you’ve ever glimpsed inside a server room. As well as housing the servers, these bays hold everything else needed to run a larger network, such as additional storage devices, networking and security appliances, backup power, cabling and even fold-out workstations for systems management.
In general, rack servers are far more powerful, expandable and upgradeable than towers.
Rack servers are ultimately geared towards providing as much power and storage as space-efficiently as possible, being specifically intended for vertical scaling. Nearly all major racks, rack servers and additional rack devices are measured in a standardised rack unit, or “Us”. In exact terms, each U is 19” wide and 1.75” tall, so a “2U server”, for example, is 3.5” tall, and so on.
Whereas ten tower servers would occupy a wide floor space or require very large shelving units, a 42U tall rack cabinet takes up little more than 1m2 in floor space.
Blade servers are an even more efficient and somewhat specialist solution, capable of fitting potentially hundreds of servers within a single cabinet.
Unlike rack servers, which usually house all the components needed to operate, blades are stripped down to the bare computing essentials - usually little more than CPUs and RAM - and tend to be far smaller. Instead of being mounted directly into cabinet bays, blades are first housed within a blade enclosure, which provides the non-computing necessities to the blades
it holds: power, networking, cooling requirements etc. These enclosures themselves can be held in cabinet racks, however it is not strictly necessary to do so.
As you may expect from such an innovative and space-saving technology, blade servers and the infrastructure to support them remain expensive, most often used by tech-centric companies who handle vast amounts of data.
However, blade servers aren’t necessarily “the best”. They have a high power density compared to rack servers, and can be more difficult to adequately cool due to the nature of having so much computing activity occurring in a confined space.
What are the most important factors when shopping for servers?
Investing in a server isn’t something we’d suggest rushing into and you should take the time to consider the key factors - likely: it needs to perform, to be cost effective and to serve its purpose for as long as it’s required.
Considering the incredible number of different server models and compatible components, specifications can quickly become a headache.
If you’re in need of a new server, the ‘best’ specs will depend entirely on what you need it to do. We’d always recommend discussing your requirements with our team but, in general, there are three main performance areas you’ll want to have an idea on - CPU, RAM and storage.
Price is also difficult to quantify without first knowing the server’s intended function and required specs. The range of investment isn’t too dissimilar
to the automotive world - it’s quite possible to spend anywhere from a few hundred pounds to tens of thousands on a single server. However, as a very rough ballpark, £1,500 - £4,000 will purchase a very capable server to function as the central storage for a small to medium office.
Just as important as the initial specs and price is to look ahead and consider how well a server will function in years to come. Ideally, it’ll be a long-term investment.
Chances are, a company’s server requirements will grow along with its own success, and being able to readily upgrade and replace a server’s components will determine how easily that happens. Investing in a basic tower server that only just services your office of 10 will seem a poor decision when that office has grown to 15 people.
Always factor in the future and consider:
How old the server model is - is it still catered for?
Are parts & components for it widely available?
Does it have room to expand/upgrade?
Is it a widely recognised brand - how easy will it be to maintain?
What software licenses will the system require?
New or refurbished?
Bytestock specialise in refurbished servers and hardware, extending usable equipment lifetimes and delivering maximum value for money for our customers.
Understandably, customers unfamiliar with the realities of refurbished hardware often have a number of concerns: is it reliable? Isn’t it all damaged? Does it come with a warranty? New tech is just better, isn’t it?
Any notion of refurbished servers being akin to your local used car forecourt can be dispelled immediately. Unlike used motors, it’s very rare that we see a server which has been run into the ground.
There are of course the odd few which have seen better days but, for the most part, these servers have spent their time with each owner in housings or cabinets, working hard but protected from harm. Cosmetic damage is normal but serious internal failures or problems that cannot be repaired or replaced are rare.
Bytestock customers trust that each and every component which passes through our warehouse is rigorously tested at every stage, and refurbished in the true sense of the word - being reconditioned for use to the highest possible standards.
As well as refurbished systems, Bytestock are also specialist stockists of cancelled order IT stock. As the term suggests, these are new systems and parts which have been ordered through the manufacturer, and cancelled or returned without being used.
Whilst they cannot again be sold as new, the products are completely unused. So, if one of the latest systems is the one you’re looking for, you could make a significant saving on an as-new item.
And, just like our refurbished stock, boxed items are covered by our unbeatable warranty...
Properly maintained, there’s no reason that refurbished IT hardware shouldn’t run almost indefinitely.
Bytestock is so confident in the reliability of our products and service that we offer a 5 Year Warranty across nearly all of our server, workstation and component range. Should a fault develop with a system purchased from us, we’ll repair or replace like-for-like as quick as we can.
The world of IT parts and components can be a little overwhelming. Endless jargon, possible product specifications and item numbers can leave the uninitiated needing a lie down.
If you’re just beginning your search for a server, or need to make sense of what your IT Manager is requesting, the following server part overviews should help.
Just like a traditional computer, a server is really just a box full of individual parts. That box is called a chassis - the metal casing which houses the functional components of the server.
As discussed earlier, there are various types or form factor of server and their chassis indicate whether Rack or Tower, 1U or 2U etc.
The motherboard is a printed circuit that controls much of what a server does - think of it as the server’s project manager, coordinating the processes, routing power where it’s needed and helping different components talk
to each other. Considering that no other component in a server will work without it, the motherboard can certainly make a case for being the most important part.
As with every piece of IT hardware, motherboards come in various shapes, sizes and specs, and your server’s chassis and requirements will determine which should be used. The vast majority of servers chassis will be supplied with a motherboard and you will likely only need to consider buying one separately if upgrading or replacing a damaged/faulty original.
At Bytestock, every server is supplied with a dedicated motherboard as standard and we stock a range of replacement and upgrade options.
Daughter Card / Daughterboard
Daughter cards - a.k.a daughterboards, piggyback boards, mezzanine boards etc. - are supplementary circuit boards that plug-in or otherwise attach to a motherboard to extend its capabilities.
Network Interface Controller
Sometimes referred to as a Network Card, a Network Interface Controller (NIC) is a circuit board or chip that facilitates a server’s connection to a network.
There are myriad types of connections - whether wired, wireless,
via Ethernet, USB or Bluetooth etc. - and an equally large number of different functions they facilitate, such as data transfer, direct memory access or partitioning. Each connection and function requires a different protocol that is managed by a NIC.
The NIC is an essential go-between, managing communication between the server and any networks it’s connected to.
CPU or Processor
Another contender for the “Most Important Part” title, the Central Processing Unit (CPU) is the brain of a server. A relatively small part, considering its importance, a CPUs primary function is performing the almost unfathomable number of calculations a system’s applications require to function.
A CPUs performance level and price are determined by its clock speed and the number of processing cores it has.
When considering CPUs, you’ll often see something like “2.2GHz” quoted; this gigahertz measure is the processor’s clock speed, with higher numbers indicating a faster operating rate.
So, the higher the number the better, right? Well, yes and no. Comparing
CPU clock speeds is only relevant when doing so with CPUs from the same product line or generation. Processor and silicon chip technology advances at such a rapid rate that a 2GHz CPU from 2019 may well be faster than a 3GHz CPU from 2016 due to the performance improvements in areas other than clock speed.
One of those such areas is processing cores. Processors used to have only a single processing core but modern CPUs feature multiple cores - dual-core, quad-core, six, eight or twelve-core etc. - meaning they can perform multiple instructions simultaneously.
CPUs can be a tricky area to navigate, especially when factoring in the specific hardware and software capabilities they determine your system can use.
If you’re not up to date with the latest technologies and it’s all too easy to get lost down a rabbit hole of performance figures. Investing many thousands on the very latest generation 28-core Intel Xeon Platinum would be overkill for the average 10-person office, whereas only spending £100 may see that office’s server struggle to perform as required.
To avoid making a potentially costly mistake, speak to the Bytestock team.
Memory / RAM
Often confused with storage, Random Access Memory (RAM) doesn’t actually ‘save’ or store files permanently. A better way to think of it would be as ‘short- term memory’, designed to very quickly take stock of what the server needs to do, temporarily remember that information and only deliver it to the CPU and other components when they’re ready to deal with it.
It’s important to note that RAM does not retain information without power, meaning it ‘forgets’ everything once its system is powered down.
RAM has three primary aspects which define its performance specifications: generation, capacity and speed.
The current generation of RAM is DDR4, however many servers still in use require the previous generation, DDR3. As with any computing component, it’s important to check minimum and compatible specifications when replacing memory, and your server model will determine which generation of memory you should buy.
The memory capacity in your system is a general guide to how many applications it will be able to run smoothly at once and, in general, the more the better. Some servers may run with 8GB or 12GB of RAM, whereas far higher-end systems may feature 512GB or more - it all depends on how much intensive work that system has to handle. However, having excess capacity past a certain point may not necessarily help your system run any quicker.
It’s also important to bear in mind the speed of RAM. Measured in Megahertz, most DDR4 RAM runs between 2,133MHz and 3,000MHz. Some high-end memory can run upwards of 4,000MHz, however it comes at a
Just like CPUs, RAM can be a complicated area to research. For complete confidence, speak to one of the Bytestock team who’ll be happy to advise exactly what you should be looking for.
Storage / Hard drive
If RAM is a server’s short-term memory, hard drives are its long-term memory. Commonly referred to as ‘storage’, hard drives are the units on which all saved data is permanently stored.
Server hard drives are specifically designed for fast access times, as well as the ability to use multiple hard drives internally.
Hard Disk Drive (HDD)
Hard Disk Drives or HDDs are the most popular form of hard drive storage. They use a spindle and magnetic write-head to physically record data onto a number of magnetic platters, which rotate on a spindle. The speed at which the platters spin determines the ‘write speed’ of the HDD and is recorded in Revolutions Per Minute (RPM) - common speeds for server HDDs are 7.2K, 10K and 15K.
Solid State Drive (SSD)
Solid-State Drives or SSDs are rapidly replacing HDDs as the default form of non-volatile memory.
Unlike HDDs, SSDs have no moving parts, which provides huge advantages: quicker access times, lower power consumption, far higher reliability and quieter operation. And, whilst still more expensive in terms of £s per GB of storage, SSD technology is improving swiftly and becoming far more affordable.
The majority of SSDs utilise a SATA connection, which has a maximum transfer rate of 750MB per second. However, recent advancements have seen a new type of connection become more popular - PCIe - which offers rates of up to 1.5GB per second.
Did you know?
The first hard drive was introduced by IBM in 1956 at a cost of $50,000 for a HUGE 5 MB of storage. Hard drive technology has progressed significantly since that time and is now one of the most affordable components in a server.
Graphics card / Graphics Processing Unit (GPU)
A miniature computing system in its own right, a Graphics Processing Unit (GPU) houses its own processor and RAM, and is dedicated to performing the incredible number of calculations necessary for rendering graphics.
The most powerful GPUs have more transistors and many more cores than the average CPU - the difference being GPUs are designed to run thousands of similar mathematical equations over and over, simultaneously, whilst CPUs are capable of a wider range of more complex tasks which they tackle one-by-one.
Think of a CPU as a Swiss Army knife and a GPU as a bread slicing machine (yes, it’s a strange analogy but we think it works!). Efficient servers require both GPUs and CPUs.
A server’s CPU works hard, a hard working CPU generates heat and too much heat is bad for performance.
To maintain optimum performance levels, all computer systems require some type of cooling. From small fans to vast water cooling setups, ensuring computer systems don’t overheat is usually the noisiest and most energy- hungry factor, but it’s a vital one.
Did you know?
In 2015, the world’s data centres used roughly 1.5x the electricity of the entire UK, and that the ICT sector could account for 20% of the globe’s power consumption by 2025.
Significant effort is now being dedicated towards developing far more efficient cooling technologies.
A heatsink is usually installed directly above the CPU and is dedicated to cooling the processor and server in general. Most servers hold up to two CPUs, with a few able to hold four or more, and each of these must be cooled by its own heatsink.
There are two types of heatsinks: passive and active.
Passive heatsinks have no moving parts, dissipating heat purely through convection. As long as there is a steady stream of cooler air flowing over the heatsink, it will be doing it’s job.
Active heatsinks are enhanced, most often by a dedicated fan or blower attached directly to it. Whilst active heatsinks are generally very reliable, they are more prone to malfunction that passive heatsinks as they feature moving parts..
Fans work in tandem with the heatsink, feeding it a constant supply of air with which to cool things. Each server model will hold a different number of fans depending on how many CPUs and heat-generating components it houses.
Liquid cooling, also known as water cooling, is the latest and most efficient way of keeping a server/computer cool, and uses a system of pipes and pumps to move cool water around the server. It relies on thermodynamics - the process of heat moving from warmer objects to cooler objects, so as the cooler object (the liquid) gets warmer, the warmer object
(the server) gets cooler.
As well as the increased efficiency offered by liquid cooling, it’s also a lot quieter than traditional fan cooling.
Did you know?
Liquid Submersion Cooling
Submersion cooling is emerging (does that count as a pun?) as the leading approach for server and data centre cooling in the future.
As the name suggests, liquid submersion cooling involves servers being permanently submerged in liquid dielectric, a specialised liquid material which poses no risk to the electronics it surrounds. The liquid has a heat capacity hundreds of times that of air, meaning
it can cool server systems far more efficiently than traditional air- conditioning approaches.
Despite being an old technology, submersion has only recently begun gaining popularity due to the huge energy and setup savings to be made at large scales.
Power Supply Unit (PSU)
No matter how impressive the spec of your server, without a power source it’s useless!
The primary specification of a PSU is its wattage, which tend to range from 350W (Watts) up to around 2700W. As with many of the other components discussed, the specific PSU/s required are determined by the server.
It’s very important to note that not any PSU can be used in any server. If you’re unsure of exactly which PSU your server needs, one of the Bytestock team will be happy to help.
For more information or advice on server requirements, get in touch with the Bytestock team!