Experiencing the HP 3000 First

Denys Beauchemin uses his firsts to lead a mission of preservation and migration for the HP 3000 community. The outgoing chairman of the Interex user group board of directors, Beauchemin is wrapping up five straight years of service on the Interex board, an unusually long tenure. He’s well-known as a prolific source of technical advice and passion on the 3000-L mailing list — going so far as to set up an annual luncheon for that virtual community’s members at HP World, where 3000 managers often met each other in person for the first time.

He’s posted his share of firsts for that community. The first time he saw an HP 3000 was in 1977, when “one of those pizza-oven type of systems” was running in a Montreal service bureau where he worked with the 3000’s predecessor, the HP 21MX. One year later as a freelance consultant, he replaced a pair of old accounting machines at a clothing factory with the first HP 3000 Series 33 sold in Canada, writing an application that is still running today.

Beauchemin was first on the scene during a crucial time for the 3000; he was one of the first to put a new PA-RISC HP 3000 system, then called Spectrum, into production. The experience he shared was important to stabilizing HP’s new investment in RISC computing. For more than five years from his post at Northern Telecom, Beauchemin participated in numerous beta tests for HP 3000 systems and software, such as the first Native Mode port of TurboIMAGE, new LANs, as well as the earliest releases of MPE/XL, HP’s first port of MPE to 32-bit systems. In 1990 he used that early PA-RISC expertise to join Bradmark, redesigning and rewriting major functions of DBGeneral to take advantage of the PA-RISC architecture and provided dramatic performance improvements.

Denys, as he’s known in the community, moved to Texas for Bradmark, then joined Hicomp and soon became an expert on multi-platform backup software and strategies. In 2003 he left Hicomp to strike out on his own once again with migration services and network installation and support.

After his years of work as the program director of many Interex North American conferences, and as he stepped down from leading the Interex board, we wanted to ask him about what’s made the HP 3000 so special to him for close to three decades, and what Interex has in store for the 3000 community in the years to come. A man whose community voice often appears in writing, he replied to our interview questions by e-mail over the year-end holiday season and reviewed the edited version that follows.

What are your proudest accomplishments for the HP 3000 community?

I see my 28 years with the HP 3000 through three windows. The first is from the end-user point of view. I should say that my work on the Spectrum beta in the 1980s was a good contribution. The second window would be from the vendor’s point of view, and I am very proud of the groundbreaking work I did on DBGeneral, moving it into the PA-RISC world. The third window is my volunteer work with Interex over the years.

Speaking of Interex, you signed up for a second term on the Interex board. Not many do this. Why did you want to extend your service?

There is a very good reason why very few people sign up for a second term: the board consumes an awful lot of one’s time. The reasons I chose to run for a second term is that Interex was approaching a perilous time in its history, and I believed that I could be of assistance helping steer it through these times by bringing in my long-time user bias. It remains to be seen if this was the right thing to do.

What was the most fun during your early Spectrum experience, and the scariest challenge you overcame?

I was at the Software Evaluation and Migration Center at the HP Cupertino campus for a month, migrating Northern Telecom’s main COBOL application to Spectrum, MPE/XL and HPIMAGE. After that migration, I demonstrated to HP that HPIMAGE was not going to be suitable for us. I pleaded with them to migrate TurboIMAGE to native mode, explaining that HPIMAGE was a disaster in the making.

Later on, we received a Series 930 HP 3000 at Northern Telecom. I was in charge of the beta program for the entire company. We belonged to the patch of the week club. The pressure was enormous. Our Series 68s, which later grew into Series 70s had reached their limit. We had a dozen or so in our computer room and they ran flat out, all the time.

Management saw that MPE had reached a brick wall. There was a lot of talk about going to another platform. My team and I were caught in the middle. In 1987, HP told me they were going to migrate TurboIMAGE to native mode, in effect canning HPIMAGE. Sam Prather, head of CSY’s labs, asked if we wanted to beta-test TurboIMAGE when it was available, and we said, “Bring it on!”

I attended that year’s 1987 Interex conference in Las Vegas closely accompanied by my HP-SE, because I was under deep non-disclosure. I had been told we would be receiving HP’s very first Series 950 to replace our Series 930, and I was told to keep my mouth shut about it. Imagine my surprise as I heard HP’s Bill Murphy announce in opening keynote that HP was finally shipping the very first Series 950, and it was coming to Northern Telecom. Talk about secret!

When we finally migrated our application to the 950, the system had some performance issues. The next night we migrated back to the Series 70 and the recriminations began. A co-worker and I flew to Cupertino to have it out with HP management. We flew back a few days later with two sets of Series 955 boards in our carry-on luggage. A few weeks later we migrated again, but to a 955 maxed out at 256MB of RAM. Under the watchful eyes of many CSY techies present in our computer room, the migration went smoothly. Then we had to migrate back, because we discovered a nasty bug in our application. The next time we migrated a week later, it was a non-event.

This entire episode was a series of challenges with unthinkable consequences as a “reward” for failure. We were exhausted after years of fighting with everybody and the systems. It is all a blur now, but in those days, the blood pressure ran high.

You did all that work when Spectrum was fighting for respect. Is Itanium going to make the same kind of stand at being HP’s small-to-medium business platform in a couple of years, or is it going to struggle?

I think the big difference is that Spectrum brought in several new things at one time: PA-RISC, MPE/XL and HP-UX; HPIMAGE, TurboIMAGE/XL and Allbase/XL databases; and HP’s optimizing compilers. On MPE, there was also the Transaction Manager. In other words, Spectrum was a massive sea change at HP in all aspects. Success was accomplished in about a lustrum, say 1983-88. By 1988, Spectrum was well on its way both on HP-UX and on MPE. I understand that a lot of the work came from HP’s cancelled Vision program, so you might want to add a couple of years to that. All this to say that it was well under a decade for Spectrum.

It is my belief that Itanium is a CPU chip, the follow-on to PA-RISC, if you will. Apart from the compilers, there is no major sea change in the rest of the environment, as far as I know. I think Itanium is very long in coming, but I also know people who are using it now simply love it for its performance. Beyond this my crystal ball is fogged up. The issues between HP and Intel probably complicated and lengthened the process, perhaps needlessly, but that is not for me to say.

Should the HP 3000 community keep holding out for an emulator, or a post-2006 MPE license?

Personally, I believe that a pure HP 3000 emulator will probably never see the light of day — not that it’s unfeasible, but rather that it may not be financially viable.

The board and I have been talking with vCSY management about [emulator and licensing] issues since 11/14/2001. So far, there has been virtually no movement in any direction. Interex has proposed to HP that Interex be a repository for the MPE source code, alongside the erstwhile RTE and HP-UX, when its time comes. Interex is not in the business of supporting software; this repository would only be for historical and perhaps hobbyist purposes.

Some say that MPE will be just fine for many companies as-is, frozen at today’s functionality. Are there a few emerging technologies that MPE had better get pretty soon?

IPV6 is about the only thing that comes to mind. Then again there are probably routers and network devices that you can hang in front of a legacy system.

I seriously doubt that anything new will be added to MPE now. My take on the situation is if you are going to successfully homestead on the 3000, you will eschew any new technology. The main value for homesteaders is the lack of change itself. Any new development for MPE is counter-productive for homesteaders.

Does the Java on the HP 3000 have a chance of being useful in the version included with the latest FOS?

Not really. I tested it for a while and I found it very clumsy and heavy on resources. I would guess that it would probably run fine on high-end machines, but not with too many users. I just never saw MPE/iX as a Java-friendly operating system. That doesn’t mean there aren’t some people out there doing very well with it. I just haven’t heard of any.

Does Interex have a plan to work more closely with OpenMPE?

As a matter of fact, Interex started talking with OpenMPE from the very beginning. Sometimes we didn’t talk for many months, but Interex is supportive of OpenMPE. Yes, Interex is very interested in being a repository for MPE, but just for historical and hobbyist purposes.

Why do you think HP wanted to mount its own technical conference for 2005?

The new management at HP is totally focused on representing HP in every aspect possible. They have decided to hold their own show and to invite the various users groups and association to attend and participate, after a fashion. We negotiated at length with HP before deciding to not participate in their event as they wanted us to do. HP will be in control of every facet of that conference, and thus Interex would lose its independence and its raison d’etre.

What’s the prospect for a merged user group for all HP users — folding in Encompass, ITUG and Interex?

Interex is unique among the list you mentioned. We do not depend on HP for financing; we are an independent company with our own offerings. We cherish the long-term HP relationship — 30 years is a long time, but we also strongly value our independence. That independence makes us unique among all the groups.

A merger with the other groups, with their dependency model, is something that should not be undertaken lightly. While most anything is possible, I do not know if merging is a good idea at this time.

What’s the one element you’ve seen in the 3000 community that is very hard to find in other places?

For me, the HP 3000 has been a love affair of almost three decades. I have used it from the Series II on MPE to an N4000/750 4-way on MPE/iX 7.5 and witnessed its demise. The road has been long and fun. The biggest element is the community surrounding the HP 3000. I have known many people in that community for many, many years. I have not seen a community like this one for Windows, Unix or anything else. I would guess that such communities exist for other proprietary platforms. But in a commodity world, I have not seen it.

I consider myself lucky to have been part of this community. It has been great, and I intend to continue to be a part of it as I help people migrate to other platforms or homestead.

Right now, for the ones who migrate, it is important that they do not lose their intellectual property in the form of the application software they have developed over the years and decades. That is what the team I work with is dedicated in doing: Preserving what the customer has and migrating it to another platform.

What’s the most significant technology to emerge and mature in mass storage/backup sector for 3000 shops? What do you recommend to a 3000 site today?

I am very excited with the disk arrays that are available now. I like the HP 7xxx series of arrays; they are fast and relatively affordable. I like the DLT drives, especially the 7000s. I was very impressed with the LTO drives and in fact I have a lot of experience with them. That was another “growing pain” issue, but they turned out real well. On an HP 3000 PCI system, especially an un-crippled one, such as a top-end N4000, these babies can scream.

My recommendation for anyone who wants to homestead for the long term: Get rid of DDS, then see if you can replace it with DLT (for Series 9xx) or LTO (on the N-Class), along with the right backup software.

You say HP crippled the 3000s with CPU throttling instituted in the A-Class and N-Class systems. Why do you call it crippling? What impact do you think it has had on the HP 3000?

CPU throttling is nothing new. HP and many others have been doing it for a long time, in an effort to maximize whatever they think gets maximized. But a throttled A-Class or N-Class is a crippled machine; it is defective. I purchased an A-Class when they first came out. I didn’t mind paying the 100 percent premium over the corresponding HP 9000 A400, because the A-Class had MPE. I knew that HP had put in some throttling: HP’s Dave Snow explained they had to slow down the CPU because the IO was not fast enough. (That was fantasy, by the way, misdirection.)

The code name for these A-Classes in the HP-UX world was Barracuda. HP was targeting that system toward applications that required speed as opposed to connectivity and disk capacity. On the MPE side, I called it “The Grouper,” a fish that is slow and ponderous. I was told to expect a throttling down to one-fourth the throughput of a Barracuda. In those days, I was deep into backup technology and I connected all manners of tape drives to my Grouper. It soon became apparent the crippling was a factor of 7 or 8, not 4 as I had expected. The machine I had was running at about 12-15 percent of its capacity; the throttling had had unexpected consequences and the system was running much slower than expected.

Just for the record, an HP 3000 N4000, un-crippled, is one fast machine. HP did a wonderful, albeit belated, job of moving to PCI with these systems. I even commissioned buttons that touted “MPE loves PCI” and passed them around at HP World 2002. The N-4000 would have been something to see on the new PCI-Express, but that will now never happen.

After that, I lost all illusion that HP was an engineering company. That was Hewlett-Packard. This is now The New HP.