Why SCSI is in it for the long haul

“How can anyone govern a nation that has 246 different kinds of cheese?”  That is a fairly accurate translation of a comment French President Charles De Gaulle made about 30 years ago.  He wasn’t really talking about cheese of course, but about the fact it’s hard to control a population that demands lots of choices.

We have a much more limited set of choices when it comes to disk drive protocols, but another choice is almost upon us.  Not so tasty as good Brie perhaps, but worthy of some mention here.

Serial-attached SCSI (SAS) is clearly a technology that is coming over the horizon.  Whether that is a good or a bad thing depends, I suppose, on where your bigotries lie.  If you have historically bought into SCSI devices, perhaps in extreme cases since the mid-1980s, you may see this as a logical development.  On the other hand, if you have invested in Fibre Channel technology over the last few years you may be tempted to ask, “Why did they bother?”

System builders who have a history of building SCSI devices will like SAS because it preserves the industry’s software investment in SCSI command sets. The new standard (accepted last week by ANSI) means SAS products will be easier to design, install and maintain due to compatibility with earlier SCSI generations.  Look for a number of plugfests throughout the year as companies work to get their products ready for market.

Disk drive vendors will like it because, well… the need for more mass storage never diminishes and they are always happiest to provide for that need in a format with which they are familiar.  Dealing with familiar technology is a good step towards keeping costs down and helping to deal with the razor-thin margins those guys have had for the last eight years. 

The disk drive divisions at Fujitsu, Maxtor and Seagate are all working on SAS products. And to support their development, chip builder LSI Logic (http://www.lsil.com) has released a single-chip SAS controller IC that transfers data to existing SAS prototype drives at 3G bit/sec, and at half that speed to existing Serial ATA (SATA) drives.  LSI is also coming out with SAS host and RAID adapters, which provide for both SCSI and ATA connectivity.

This last issue, the ability of the controller to provide data to both SCSI and ATA drives, deserves comment.  As SAS and SATA use the same commands to move data on and off the disks, and as they are plug compatible, IT managers can pick and choose the devices in their drive bays based on price, performance, price-performance or whatever other set of criteria they my care to use.  Managers who invest in this technology should like the capability to play mix-and-match with the drives on their storage subsystems, particularly when systems age and get relegated to less important storage tasks.

The interoperability that exists between SAS and SATA should certainly be well-received by those of you in IT, but it may turn out to be less of a blessing for the SCSI vendors if they are not careful.  They will always have to be at least a bit nervous about those IT managers who decide that cheaper SATA devices will do the job that needs doing “well enough.”  “Well enough” will mean different things at different sites, and there will always be buyers who for their own reasons prefer the SAS “high-priced spread.”  But SCSI vendors may find they need a better answer than the knee-jerk responses about performance and reliability specifications.  After all, MBTF (mean time between failure) numbers become much less important when devices are placed in RAID systems that have been designed to compensate for the failure of a disk device.

So the road for SCSI seems to go ever onwards.  But it will be interesting to watch what happens when it intersects with SATA.