IBM Shipping Seagate Drives As iSeries Disk Problems Persist

by Timothy Prickett Morgan

I just so happen to be looking at one of the new 35.2 GB disk drives that IBM announced a few weeks ago and it says that it is indeed a Seagate disk. IBM may be using a mix of its own UltraStar disks and Seagate Cheetah 73LP disks; the company isn't saying. Maybe none of the brand-new 10K RPM disk features that IBM is shipping to iSeries and AS/400 customers have IBM's drives in them; they could all be Seagate drives. IBM isn't saying, and under normal circumstances I would say it was not important so long as IBM cut prices radically and gave OS/400 shops a good deal on disk drives like owners of Unix, Wintel, and Lintel servers enjoy. But my instinct is telling me that this is still not a normal situation, and my advice to you is to watch your machines very carefully and find out exactly what you are getting from IBM when you buy disk capacity. You'll want to know if they are IBM or Seagate drives because these devices, though adhering to UltraSCSI standards, have to work somewhat differently in conjunction with OS/400 to provide the same features, functions, and performance. I've been told by people familiar with the situation that IBM is using the Seagate drives not only in the 6717 (8.58 GB), 6718 (17.54 GB), and 6719 (35.2 GB)disk units of iSeries and AS/400 servers using SPD disk arrays (which I told you about back in January), but also in the equivalent feature 4317, 4318, and 4319 versions of those features for HSL disk arrays and the feature 6817 (8.58 GB) and feature 6818 (17.54 GB) disks for PCI disk arrays. I have not been able to independently verify this information from IBM or Seagate, neither of which want to talk about this.

Aside from being alerted to continuing problems by readers of this newsletter, I am also concerned about that phantom DASD Availability Fixpack we told you about in last week's issue. IBM put out an updated DASD Fixpack on February 25--one week after beginning shipments of the new 35.2 GB disk drive--and then three weeks later removed it from circulation. One customer who emailed me said that they had requested several months ago that IBM remove all of the disks in its Model 830, which has four partitions, and replace them with new units after four drives in the machine failed. IBM did not do this, but rather said to apply a bunch of PTFs and activate Service Director. The customer did this. Then last week, a disk associated with the primary OS/400 partition started to fail. An IBM tech was called in, and he trucked out to the customer with a new disk. When he was installing that disk, he found two disks had failed, and then a third one went--in the same parity set as the first one. This caused the whole machine to crash mightily. This is one unhappy customer.

Let's review the situation. IBM's UltraStar 36Z15 15K RPM disk drives were delayed by more than a year because of serious engineering problems, which apparently have been corrected and IBM is beginning to ship them in limited quantities to customers. Big Blue's UltraStar 36LZX and 73LZX drives, both 10K RPM units that are used in IBM's OS/400 servers, had all sorts of problems that were causing them to drop off the SCSI buses, which was causing much consternation among the OS/400 customer base throughout 2001. IBM Technology Group, presumably after much heat from IBM chairman and CEO Louis Gerstner, IBM's OEM customers (who flocked to Seagate in droves), and the divisional managers within IBM's Server Group, got to work and stopped blaming the problem on the various server operating systems that were using these disks and got around to making substantial microcode changes to the UltraStar 36LZX and 73LZX 10K RPM drives. The confidential document supplied to IBM's disk buyers (but not its customers, who are directly affected by the issue), indicated that IBM put out an all-encompassing disk drive microcode patch at the end of November 2001--the kind of patch that usually is announced for a new disk line just as they start shipping (and which was indeed announced for the UltraStar LZX disks in October 1999 when they started shipping). This report from Technology Group showed that the LZX drives had seriously flawed microcode. The disks would hang after overwriting vital data and the microcode had bugs in pieces of the code that allowed a drive to reset after a failure. The error-recovery code and write cache microcode had bugs that caused it to hang. The code-controlling arms of the drive could cause the drive head to go off track, causing errors. As I said before, IBM has not said that this microcode has been rolled out to the OS/400 customer base that is still using these disk drives. This might have been the core components of the phantom February 25 DASD Fixpack. However, something appears to have gone wrong.

The rumor I have heard is that the hardware data compression algorithms that IBM has been offering to OS/400 customers for a couple of years may not work properly with the Seagate disk drives. People who have used the new disk say that the data compression algorithms--which are controlled by OS/400 but implemented in the disk IOP controllers and the disk drive features--do not work with the new 35.2 GB features when they use Seagate drives. If it doesn't work for one, then it probably doesn't work for the others. Whether or not there are other issues with the Seagate drives, I do not know. If you are still having problems with your iSeries and AS/400 disk drives, let me know at tpm@itjungle.com.

Incidentally, alternative OS/400 disk array vendor BCC Technologies still stands by its statement that the UltraStar 36LZX drives work reliably when updated with the new microcode from IBM's Technology Group. The company says that the new UltraStar 73LZX 10K RPM and 36Z15 15K RPM disk units are also reliable after IBM made substantial engineering changes to them. BCC sells all three IBM units in its disk arrays, and has stopped using disks from Seagate or Fujitsu in its OS/400-compatible arrays.

While this is a very serious matter, there is one hilarious irony to the problems IBM has been having with the iSeries (and presumably other server families). If IBM's MidMarket Server division is indeed using Seagate disks exclusively in new features, then if you want non-IBM disks, you have to buy from Big Blue, but if you want IBM disks, you have to buy from BCC, a third party supplier.

One last thing: IBM, give the iSeries and AS/400 base a break. Seagate is selling the Cheetah 73LP drives for around $1,000 in single unit quantities. (These are apparently being formatted down to 35.2 GB capacities in the iSeries line.) Raw 36 GB UltraSCSI Cheetah and UltraStar drives Seagate and IBM respectively sell for about $500. But IBM is charging a stupefying $3,200 for a 35.2 GB drive when modified for an iSeries or OS/400 server. While I concede that IBM adds valuable electronics over and above the disk drive that make them (always in theory and mostly in practice) more reliable--and therefore worth more money--this is a bit much. You want more new iSeries customers, IBM? Stop gouging your existing ones. The iSeries is a great machine, and more people would want to come to it if you would start charging reasonable prices for memory, disk, and other commodities that are part and parcel of an iSeries server.