Souping up storage

Most servers — PC servers included — play too important a role to risk data

loss or downtime because of a hard drive failure. Unlike most of the parts

inside modern computers, hard disk drives have moving parts and therefore

suffer failures more often than the semiconductor components.

The most common way to guard against data loss from a disk failure is

to use Redundant Array of Independent Disks storage. But adding a RAID controller

to a PC server could cost customers another $800 to $2,000, said Subo Guha,

director of product marketing for Dell Computer Corp.'s enterprise systems

group. By integrating it onto the systemboard, vendors can add RAID reliability

for only about $300 more, Guha said.

Once RAID is in place to protect the data, servers still need to support

replacement of defective disk drives while the server is running so that

agencies can continue to do business while the server is repaired. "All

of our servers have the hot-swap drives," said Alvin Mitchell, contract

manager for the Navy Facilities Engineering Command contract. "They come

in handy, because you can just pop them out and put the new ones in."

That means that the server stays up not only when the drive fails, but

also while it is being replaced. "You do get a performance hit because it

has to rebuild the data on the new hard drive," Mitchell said.

Those hard drives are usually SCSI devices, so SCSI controllers are

also typically standard equipment on entry-level servers. Gateway Inc.,

aims to set its products apart with support for the speedy new Ultra SCSI

160 specification, the fastest version yet, said Chris Schlieter, Gateway's

marketing manager for advanced servers. Because the new standard is backward-compatible

with the existing specification, customers will still be able to use older

disk arrays, he said.