The heart of any computer network is the server. If you think personal computers look far from sexy …meet their really ugly cousins, the servers. Servers are designed and built with one single goal in mind – reliability. Everything from the components, operating system, software, functionalities, power supply, etc is chosen and put together so that a server will run problem-free and continue running 24/7/365. Imagine a passenger car that is all tricked out with spoilers, skirting, effect lighting and an entertainment system that can blow out eardrums. That’s your personal computer. Think a rally car where everything not essential is stripped out until it is almost bare metal inside the cabin. That’s a server.

Many operating systems can be used for servers: Microsoft Windows 2000, Windows NT, Windows XP (32bit or 64bit), Windows Sever 2003 (32bit or 64bit), all in the BackOffice and Small Business Server suites including SQL Server & Exchange Server; Linux Red Hat Enterprise 3.0 Update4 (32bit or 64bit), SuSE Enterprise Server 9 32bit; OpenBSD; Mac OS X Server; Sun Solaris; and other flavours of UNIX from AT&T, Sun, HP/Apollo, OSF, X/Open, Novell, SCO, IBM, Siemens, SUN, DEC etc.

Server storage are usually RAID systems. Originally an acronym for redundant array of inexpensive disks, the RAIDs for servers would better be described as redundant array of independent disks. RAID systems can have the feature of greater speed through striping. That’s where data is split and written to two or more hard disks in a round-robin sequence. For increased reliability, mirroring is used where a set of data is written completely on two or more hard disks so that the data is always replicated.

Standard RAID configurations include RAID 0, RAID 1, RAID 3, RAID 4, RAID 5 and RAID 6. A RAID system could be built with RAID systems instead of single hard disks. For such systems called nested RAIDs, common configurations include RAID 01, RAID 10, RAID 50, RAID 51 or RAID 53 and RAID 100. To meet specific needs there are also non-standard RAIDSs, many of which are proprietary.

So how can such a RAID or nested RAID system possibly fail? Especially considering the fact that hard disks used in servers can be of a higher, more reliable grade compared to consumer hard disks.

The RAID controller, which is the brains of the entire array of hard disks, may fail. A virus attack is no respecter of RAID systems. It can bring down a RAID just as surely as it can a single disk computer. If the server room is destroyed by lightning or fire, all the redundancy will be wiped out too. Think it is bad luck if two or more hard disks fail at the same time? Hey, so it lightning striking a server room! Human sabotage or plain stupid mistakes can bring a server to its knees. Actually, it has been estimated that hardware problems and disasters like fire and lightning only account for one fifth of unplanned downtime. Human factors could make up twice or more!

When a hard disk goes down and is replaced, all the remaining disk drives are subjected to much higher stress levels when the RAID goes through reorganisation. This is because the reorganisation is loaded on top of the usual RAID activities. During the process, it is not uncommon for another disk which was already about to fail to give up resulting in the data being in a worse state.

The slow reaction by support staff to replace one defective hard disk can result in data loss. A sustained power outage may beat the best uninterrupted power supply and compromise the server’s data integrity.