storagesearch.com

• Backup SW
• NAS
• Solid state disks
• iSCSI
• Serial ATA
• Manufacturers
• US VARs
• UK VARs

When the Bean Counters Can't be Trusted Any More

As if we didn't have enough financial scandals already in the stock market. Enron, Worldcom and now we hear that blue chip pharmaceuticals company Merck may have counted $14B of revenue which it didn't really bill. Unfortunately the fallout from these accounting irregularities has hit all kinds of companies.

On July 8th when Intel announced that it was shipping its newest Itanium 2 processor in volume, which you would have thought was a good thing, its share price actually dropped 5%.

Ouch!

Now overstating things to make them look better only works until someone takes the time to look carefully. In my youth I can remember having many interesting discussions with some memory board suppliers whose bus interface speed was only about a third of what they claimed, and on the other hand being pleasantly surprised by some disk drives which were actually faster than suggested in their spec sheets. (It may have helped I was in a beta test site for a new very fast disk controller card at the time. Faster than the disk manufacturer had in mind when they specced their product.)

But here's a worrying thought... What would happen if the malaise which has infected the financial reporting community started to work its way down into the other markets, like gulp! - storage?

How do you know that the RAID system you've got really has got as much capacity as claimed? It should be easy enough to check by writing some simple code, but with everyone being so busy nowadays who really does have time to check all these little details?

How do you know that the UPS in your SAN really is rated for 4 hours, and that the manufacturer hasn't cut corners by leaving out a few batteries?

Last year Sun Microsystems got stung by applying some short cuts and omitting to put ECC checks into a cache controller for its high end processors. That problem cost them and their customers plenty. It shouldn't have mattered if everything worked as per the simulations, but it did matter in real-life because some of the memory devices had a susceptibility to alpha particles and sometimes flipped a bit. That old alpha particle problem hadn't been seen in the industry for about 20 years but nevertheless the ECC would have fixed it. The result? IBM sold a lot more mainframes than expected that year, and Sun lost its reputation for reliable crash proof systems.

As we move even further down the food chain what would happen if a manufacturer of GBICs were tempted to change their testing regime to increase the yield which passed noise and jitter tests? That could be a way to make more money. Users wouldn't lose out too much because their SAN protocols would just redeliver the data if it didn't look right first time. There might be a speed degradation. Not much to worry about there. Most systems don't run at full speed anyway. Who's to know?

Or a tape manufacturer, knowing that most people only do weekly backups, could specify a different coating from a cheaper source. Only problem... which wouldn't show for a year or so, is that the tape wouldn't last so long for the small percentage of users who do daily backups. By then who cares? The recession is over.

The scope for these kinds of bottom line "improvements" is massive in every field of life. Let's hope that the inflatable bean counters stay out of the storage market.