In a network, there are lots of data and multiple people sharing it. The file server at the centre of it is under constant use, with a heavy workload. The discs are being driven hard, and the consequences of a disc failure are high. To help mitigate against this, it is normal to have more than one disc drive, configured in what is called RAID. RAID is short for Redundant Array of Inexpensive Drives. There are several variants of this, but the most common are RAID 1 and RAID 5. In RAID 1, there are two disc drives and one is an exact mirror of the other; if one fails, the other carries on regardless until it can be replaced. RAID 5 is more complicated; there are at least three but preferably more more drives. Data is written across all the drives according to complex mathematical algorithms, such that if one drive fails then the system can work out what is "missing" and recreate it from parity information on the other drives (this is a simplified explanation - a lot more information can be found on Wikipedia, for example). Speaking very generally, a file server in a small organisation will commonly have RAID 1, whilst that in a larger organisation will use RAID 5 or something even more sophisticated.
It will be appreciated that a normal desktop or laptop computer does not have RAID. On a server, it is done using an additional component called a RAID controller. A RAID controller is a relatively complicated piece of electronics and therein lies the rub: this makes it correspondingly expensive. This may not be an issue in a large organisation but for a small business or school it is a consideration. For instance, a HP Proliant Microserver costs around £250. A RAID controller for it costs the same amount again, thereby doubling the overall price.
To get around this issue, chipset manufacturers have developed a low-cost alternative. This may be referred to as 'SATA RAID' or 'Host-based RAID' or 'Software RAID'. Rather than a dedicated hardware controller, the standard electronics in the computer are used (possibly with an additional small chip to handle the setup) and all of the work is done by the computer, controlled by a software driver. There is a slight overhead on the computer, although not too much for a modern machine. Most variants can handle RAID 1, although don't do RAID 5. As it adds little or no cost to the server, it seems an attractive option.
The problem is when things go wrong. Because it depends upon a software driver for Windows and the processing takes place in Windows, when the operating system has a hiccup then it can be very bad news indeed. For instance, if Windows becomes corrupted (power problem, bad update, flakey hard disc etc) and cannot start, then the discs may not be visible i.e. the data has gone. Because Windows won't start, you can't access the limited tools that may be available to try and manage the discs. A true chicken-and-egg situation. The problem doesn't happen with a proper hardware RAID controller because it is "intelligent" in its own right and operates quite independently of Windows. It also has more tools and options to fix problems.
Because of this problem, low cost RAID systems that do not use dedicated hardware controllers are known in the trade as Fake RAID. A Fake RAID system may work perfectly well for the lifetime of the system, but there again it may not and when it does fall then it is likely to be catastrophic. In a recent example, a customer had a Dell T110 server that would no longer start. Investigation revealed that the RAID had failed, and moreover it was the built-in PERC S100, which is a Fake RAID. Long conversations with Dell tech resulted in the conclusion that nothing could be done, and the comment that "it's a cheap server... what do you expect?" (actually the T110 is a fine little server, provided it is used with a proper hardware RAID controller).
Bottom line... this is not an area for compromise or false savings. When buying a file server - even an economy one - it is important to get a decent, hardware RAID controller.