Home > Buying Tips > Full Story

RAID your network

Are you serious about your data. Investing in a RAID system can help keep your data safe and sound. by Suryakant Naik

'Server extremely low on disk space'! Network administrators cringe when these words pop up on the screen. Perhaps they already knew they were running out of storage. And now, to upgrade/install additional storage capacity, the server has to be shut down creating service blackouts for employees and customers. In a 24x7 service world, where downtime means lost productivity or business, an event like this is unacceptable.

To avoid a situation like this, some companies anticipate their storage needs well. That often results in overspending on equipment. This attitude and the increasing storage appetite of corporate networks has led to a boom in the data storage market. Sales of storage disk systems will climb from $39 billion this year to $53 billion in 2004, according to market research firm IDC.

DAS
DAS (Direct Attached Storage) is mainly concerned with the storage that is attached directly to a server. Today, more than 90 percent of all computer storage devices like disk drives and disk arrays are directly attached to a client computer through various adapters with standardized protocols like SCSI (Small Computer Serial Interface) and FC (Fiber Channel). This type of storage is also known as sever attached storage.

In PC configuration, storage is usually integrated in the same cabinet within the processor system. In mainframes and large servers, the storage (disk subsystem) is typically located in a separate unit at some distance from hosts. This subsystem may be configured either as separate and independent disks, typically called JBOD (Just a Bunch OF Disks), or as a fault tolerant arrays of disks, popularly known as RAID (Redundant Array of Independent Disks) subsystem. This plays an important role for DAS selection.

RAID arrays for data redundancy

RAID—initially called Redun-dant Array of Inexpensive

Disks—was originally conceived by a research team at the University of California Berkeley in 1988. By combining a collection of disks into a single logical array, an administrator can store redundant data across these multiple disks in order to provide fault tolerance by enabling quick recovery from a disk failure.

To choose a RAID system

Should one go for a SCSI RAID array or FC? Here are five main categories to be considered while choosing a RAID system.

Performance: The highest available SCSI standard is wide Ultra2 SCSI whose bandwidth is 80 Mbps, whereas FC's bandwidth is 100 Mbps. For example, if you are configuring a RAID system that has no more than ten drives, that is nine data drives and one parity drive, then the performance benefit of FC's high bandwidth is not fully utilized. Although ten drives will swamp the bandwidth of Ultra2 SCSI, they will not use the full available bandwidth of FC.

Reliability: It is important to know that FC was designed for fault tolerance and high reliability. The following list describes some of the in-built FC standards which make the technology more reliable than SCSI.

• FC supports dual ports and dual loops while wide Ultra2 SCSI does not.
• FC guarantees data delivery.
• FC has less than 10x12 bit error rate far exceeding wide Ultra2 SCSI.
• FC standard contains 32-bit CRC (Cyclic Redundancy Check) error detection while wide Ultra SCSI has only parity.
• FC with copper interconnects uses only four wires between each FC enclosure. Ultra SCSI uses 68 wires per SCSI channel between enclosures.

FC has a higher level of resilience and reliability. This becomes more important when looking at higher bandwidth and greater distance requirements.

Cost: Wide Ultra2 SCSI is more cost effective per MB for RAID systems up to 500 GB on PCI (Peripheral Component Intercon-nect)-based platforms. FC becomes more cost effective per MB for RAID systems over 500 GB.

Cost certainly needs to be balanced against performance in any RAID system. As seen in the above discussion regarding performance, in RAID systems with more than ten drives, FC takes the high performance lead. And, when multiple enclosures are cascaded together into large storage systems, FC becomes more cost effective than Ultra SCSI.

Installation: With single enclosure RAID systems, there is little difference in installation issues between wide Ultra SCSI and FC. When your system expands to multiple RAID enclosure configurations, the installation advantages of FC become readily apparent.

During installation, SCSI requires an ID be set up for each disk drive/device in the RAID. FC devices have a WWN (World Wide Name), like Ethernet, which eliminates the ID set up. SCSI requires term power and termination to be applied correctly. FC has no such requirements. All of these factors make an FC installation faster,

simpler, and more reliable than Ultra SCSI.

Flexibility: FC is undoubtedly a more flexible technology than Ultra2 SCSI. The following list describes some of the reasons for this conclusion:

• FC-AL (Fibre Channel-Arbitrated Loop) supports up to 126 devices per loop. Ultra SCSI supports 15 per channel. Essentially, one FC controller is equivalent to eight parallel SCSI controllers.
• FC-AL storage devices, can be located up to thirty meters apart when using four-wire copper cable. This distance can be increased up to ten kilometres when using fiber optic cable, which greatly facilitates off site expansion. Ultra SCSI can be 3 meters apart with standard connections and 12 meters apart with differential connections.
• FC drives are standard with dual loop capability. This allows complete fault tolerance to be implemented in a RAID subsystem. Ultra SCSI has a single bus connection. Therefore, in Ultra SCSI RAID systems, mirroring drives need to be implemented to attain complete fault tolerance. This increases costs significantly.
• FC RAID systems are simple to scale to Terabytes of storage. With Ultra SCSI, scaling to such capacities is extremely difficult.

Both FC and Ultra SCSI has its place in the RAID system market. Ultra SCSI is a good solution for single enclosure and server embedded RAID systems with low to medium performance requirements. For large capacity, high performance systems with requirements for high fault tolerance and data integrity, FC is a superior solution. j

RAID glossary

Mirroring

Disk mirroring involves the simultaneous writing of the same data over one RAID controller on two separate hard disks

Duplexing

Similar to mirroring, disk duplexing is the simultaneous writing of the data over two RAID controllers on two separate disks

Striping

This involves breaking data into small pieces and distributing it across multiple disks. Depending on the RAID implementation, data is broken up into bits, bytes, and blocks

Parity

Parity is logical information about the data you're storing, and it's use to re-create lost data in the event of a disk failure. It is used in combination with striping and usually involves at least three disks

Suryakant Naik can be reached at surya@lanbitmail.com