SCSI controllers.

The reason the Small Computer System Interface is used so much on servers has two reasons.
The first is simple to understand.
SCSI can process I/O requests concurrently for each device on the chain.
Example: A read command for Disk 0 comes from the CPU.
The SCSI adapter sends the request to Disk 0.
A read command comes for data from Disk 1.
The SCSI adapter sends the request to Disk 1.
A read command comes for data from disk 2.
The SCSI adapter sends the request to Disk 2.
Disk 0 has by now moved the head and found the right sector, and has the requested data.
The SCSI adapter sends the data to the CPU.
Disk 1 now has its data available.
The SCSI adapter sends the data from disk1.
A request comes for another read from disk 0.The SCSI adapter sends the request to Disk 0.
Disk 2 now has found the requested data.
The SCSI adapter sends the data out from Disk 2.
etc.
IDE can not do this.
After the first command, IDE must wait for the response. (With a single HD in a PCthis is not a problem.)
SCSI can also have more disks attached.
Now the most important reason, RAID 5 .
Under RAID (Redundant Array of Inexpensive Disks) , There is both redundancy and data striping.
Redundancy means that you have more than one copy of the data, so you can replace and rebuild any failed disk without loosing data.
Striping means the data is split up and each sector(s) is written to a different disk. This improves throughput, because disks are so slow.
The SCSI bus has usually also had a data throughput advantage.
There are many good white papers on this subject that go into much more depth than is allowed here. Check adaptec.com , ami.com for starters.
In the Adaptec website find and click on "What is RAID" , and you can see some neat graphs that are not allowed here. (This box is text only.)
Regards, Chris