Wednesday, March 25, 2009

Scsi Card

Scsi card
A SCSI card is a card that will control the interface between SCSI versions of hard drives, CD-ROM drives, CD-ROM burners, removable drives, external devices such as scanners, and any other SCSI components. Most fit in a PCI slot and there is a wide range of types. The three main types of connectors on these cards are 25-pin for SCSI-1, 50-pin for Narrow SCSI, and 68-pin for Wide SCSI (and Ultra-Wide SCSI, Ultra2-SCSI, Ultra160 SCSI, and Ultra 320 SCSI - all of which use a 68 pin connector).


SCSI controllers provide fast access to very fast SCSI hard drives. They can be much faster than the IDE controllers that are already integrated your computer's motherboard. SCSI controllers have their own advanced processing chips, which allows them to rely less on the CPU for handling instructions than IDE controllers do.
For the common user, SCSI controllers are overkill, but for high end servers and/or the performance freaks of the world, SCSI is the way to go. SCSI controllers are also much more expensive than the free IDE controller already included on your motherboard. There is also a large premium in price for the SCSI hard drives themselves. Unless you have deep pockets, there isn't much of a point in going with a SCSI controller.
Many people buy SCSI controllers just for use with their CD-ROM burners and CD-ROM drives (these drives must be SCSI drives of course).
SCSI cards also have the ability to have up 15 devices or more per card, while a single IDE controller is limited to only 4 devices (some motherboards now come with more than one IDE controller though). SCSI cards allow these drives to be in a chain along the cable. Each drive on the cable has to have a separate SCSI ID (this can be set by jumpers on the drive). The last drive on the end of the cable (or the cable itself) has to "terminate" the chain (you turn termination on by setting a termination jumper on the drive - or use a cable that has a terminator at the end of it).
Notes on SCSI and IDE RAID:
RAID stands for Redundant Array of Independent Disks. RAID arrays combine multiple hard drives to act as one. You have to have at least two drives for a RAID array, and in some cases you need more. RAID is generally considered overkill for home users, although it is very important in the server market. For those really interested in high peformance, a RAID array may be worth setting up. The primary benefits of RAID include Redundancy and Performance, although not always both together. There are several different levels, or types, of RAID arrays. I'll discuss the four primary forms, their benefits, and their disadvantages:
RAID 0 is first, and it is also known as a Striped Array. In this level with 2 drives, part of the data is stored on one drive, and part of the data is stored on the other. This way, the RAID card can write a small amount of data to both drives at the same time about twice as fast as a single drive could write the same amount of data. The RAID card can also read data from the two drives at once, making read speeds twice as fast. This RAID setup is all about performance and does not provide any redundancy. If one drive dies, all data is lost. RAID 0 is good for desktop situations where data is not mission critical, but it is a good idea to make backups often. Since data is not reproduced, two 36 GB drives would be seen as one 72 GB drive.
RAID 1, which is also called Mirroring, uses 2 drives and has the exact same data on each drive. The major advantage of RAID 1 is redundancy. If one drive dies, the computer can keep going normally. Read speeds are also sped up because the drive can read different parts of the data from each of the two drives at the same time. Write speeds are slower than a single drive though because the RAID card has to write the same data to both drives instead of just writing once to one drive (or across multiple drives like in RAID 0). RAID 1 can be good for keeping backups of your important data, while also increasing read speeds. It can also be helpful in web servers, since reading is what is done most often. Since data is reproduced, two 9 GB drives would be seen as one 9 GB drive, so you don't get the full benefit of all drives.
RAID 5 is Striping with Distributed Parity. This configuration requires at least 3 drives. It provides the same performance as RAID 0, and also adds the redundancy of parity. A "parity bit" of the data is distributed across all the drives so that if one drives dies, the array can continue working okay, but you would need to replace the third drive. The calculations can slow write speeds though. This also gets expensive since 3 drives are required. RAID 5 is most commonly used in big enterprise servers. Most IDE RAID controllers do not support RAID 5.
RAID 0,1 is also worth mentioning. It combines RAID 0 and 1 to stripe two drives, and then mirror those two drives on another two drives. Thus, a minimum of 4 drives is required. You get speed improvements, but also keep redundancy. The server you're viewing this web page on actually runs a RAID 01 array!
RAID 1.5 is a new RAID format that combines mirroring and striping with just two drives (which should provide good reliability and speed). The jury's still out on how well this works. I don't know much about this new format yet.

No comments:

Post a Comment