By John Burke
After reading Jim Hawkins’ reply to my SCSI is SCSI article, I was reminded about HP’s 4Gb disk drive fiasco. These branded drives had a nasty habit of failing after being powered off after they’d been running for a while. The problems were not limited to the HP 3000 versions, either.
At one point we got so frustrated we just replaced all 4Gb drives with the much more reliable 9Gb drives. I never blamed HP for these failures, or the failures of the 4Gb drives on my HP 3000 — even though all were purchased from HP, and had HP stamped all over them. The failures were the fault of the manufacturer, and no amount of certification testing would likely have shown the problem. But the failures made me wonder: What does HP certification and HP branding mean?
In Hawkins’ reply, he puts great emphasis on the statement that “In the SCSI peripheral market, Industry Standard is really defined as ‘works on a PC.’ Unfortunately, the requirements for single-user PCs are not always in alignment with those of multi-user servers.” Maybe inside HP the desktops look different, but I have never seen a company use SCSI peripherals as a standard for desktop Wintel systems.
At my last employer, we had approximately 1,200 desktops, and not a single one had a SCSI disk drive. SCSI disks are used primarily in the multi-user server market, not the desktop market. While Hawkins says some interesting things in the rest of his article, these two sentences tend to prejudice the reader against everything else he says.
Unfortunately, Hawkins’ best argument came out in private correspondence: “Putting newer disks inside a 9x7, 9x8 or 9x9 may overtax the power supply and/or ‘cook’ your CPU or memory.” However, most of us outside HP have been advising against using internal drives in production machines for many years because of the obvious maintenance headaches. It still amazes me how many people believe you have to have at least one internal drive in an HP 3000.
The debate seems like it highlights at least four things going on.
2. Does the listing of a drive in IODFAULT.PUB.SYS imply its certification by HP? I think it is reasonable for a customer to look at IODFAULT, pick out a “supported” drive, and think he can buy it from whoever will give him the combination of price and service that meets his needs. For anything but the newer systems, you only need two generic drives listed in IODFAULT (one SE and one FWD). So what is a customer to make of the drives listed in IODFAULT.PUB.SYS?
3. Does HP branding imply HP-specific firmware? So why then do the HP drives almost always report the original manufacturer’s model number instead of HP’s? Again, this leads to the customer assuming he can buy a STxxxxxx and it is the same as the drive he buys from HP.
4. Are all HP-branded drives equal? I had an HP-branded drive that I pulled from a server that I got happily spinning in my 9x7 at one time. And, yes, of course the duty cycle in this 3000 was extremely light. But I am also relatively sure the HP part number (as opposed to drive, whose reported model number is in IODFAULT) was never sold as compatible with the HP 3000. It sounds like this HP-branded drive is just as risky as any non-HP branded drive.
It was never my intention in the original article to bash HP, or the fine people who continue to be associated with the 3000 division. Perhaps I should have reworded the title to read, “If you feel abandoned by your vendor, then take comfort in the fact that in most cases, SCSI is SCSI.” But that doesn’t exactly roll off the old tongue. Yet, in reality, this is what most of HP’s argument has been about: those few cases when “SCSI is SCSI” is not true. It should also be noted that my original article was aimed at those HP 3000 sites planning to homestead for some period of time.
After considering Hawkins’ response to my original article and numerous private messages, my position can now be stated like this: With the exception of the “hot” drive issue, any name-brand manufacturer SCSI drive you can electrically connect to your HP 3000 will likely work. If it survives your own testing (mount it as a separate user volume, and bang on it for awhile before moving it into production) then you should have little to worry about.