The company insists that it has fixed the problem for all of the Shark DS8000 and DS6000 arrays it has sold to date. However the software flaw comes only months after what was the first ever major rework to the architecture of the disk array, and is an uncomfortable echo of difficulties the array suffered when it was first launched in the late nineties.
A problem with the microcode operating system on the disk arrays caused crashes, shutdowns and unpredictable behavior for Shark DS8000 and DS6000 devices after 49.7 days of uptime. IBM has admitted that some customers have had to take their hardware offline in order to apply a remedial software update.
One regional US bank told ComputerWire that IBM did not warn it of the problem until the very day that its hardware was due to reach the 49.7 day limit, and that it was then forced to take its DS8000 array offline for two hours to rectify the problem.
They came in and started unloading the cache. Then they re-booted the box, said a source at the bank who did not want to be named. Although the problem was purely the result of a software flaw, IBM returned a few days later with a replacement Shark, presumably in an effort to entirely reassure the customer that the problem had been vanquished.
IBM is adamant that every Shark it has sold has now been patched. Although the problem was still being dealt with in June, and the company could not say when the last Shark was patched, it attempted to play the issue down as old news.
How big a dent the bug has made in IBM’s reputation will depend on individual customers’ experiences. Arun Taneja, analyst at the Taneja Foundation, said: These boxes are supposed to provide five nines reliability. If the whole system goes down not just a volume that’s big news. This is absolutely something that matters to customers.
But IBM is not the first disk array maker to have suffered embarrassing bugs, according to both Diane McAdam, analyst at the Data Mobility Group, and Tony Asaro, analyst at the Enterprise Strategy Group. McAdam would not name the vendors, but said that within the last two years two high-end array suppliers have suffered problems with leap year and Halloween dates.
Greg Schulz, analyst at the Evaluator Group said: It’s a matter of how the vendor deals with the problem how pro-active they are about it.
The Shark suffered from poor performance when it was first launched in the last decade, and the DS8000 and DS6000 represented the first major change to its architecture since then. That puts some pressure on IBM to convince the market that it has not made the same under-development mistake twice.
Asaro said he had yet to speak to any customers suffering problems with the DS8000. But he added: If there are problems, we can’t just look at this from a high level. We need to look a lot deeper. It’s important for IBM to get this one right.
IBM launched the DS8000 late last year, but because of production problems only began shipping the device in volume in March. The company said yesterday that this was not related to the microcode problem.
