Epic Sponsor Message

May 2002

Net.digest summarizes helpful technical discussions on the HP 3000 Internet newsgroup and mailing list. Advice here is offered on a best-effort, Good Samaritan basis. Test these concepts for yourself before applying them to your HP 3000s.

Edited by John Burke

Unbelievable! One of the most popular topics of the 3000-L over the last six months, HP’s merger, is still undecided. As I write, it’s in a Delaware court. However, I continue to stand by my prediction: The only sure thing is that regardless of the outcome, everyone involved will come out of this diminished.

April brought the annual Solutions Symposium (in a little bit of gallows humor, it was noted that many of HP’s slides were labeled “Solution” Symposium, as if there was only one solution: Migrate). The first morning was devoted to HP presentations. On 3000-L in April I posted my report about that day, which follows. (I’m writing this three weeks after the Symposium, and little has changed.)

The fact that HP 3000 division (CSY) General Manager Winston Prather’s talk was reduced from the advertised 30 minutes to 10 pretty much tells the story. He spent much of his 10 minutes congratulating himself and HP because, according to their surveys, fully 85 percent (a number that is hard to find substantiation for) of the HP 3000 customers were taking HP’s advice and planning to move off the HP 3000. He seemed unconcerned about the fact that since HP has not given any support to any other alternative, customers really have no choice other than to say they are going to transition off the HP 3000. A more interesting question is why even using HP’s numbers, 15 percent of customers still refuse to go lock-step behind the only presented alternative?

He repeated his mantra about the eroding ecosystem leading to the decision about the end of life (EOL) of the HP 3000, continuing to blame everyone except the organization that created the ecosystem in the first place and then failed to nurture and develop it. Come on, guys, take some responsibility. It would be easier to swallow the final decision if we were not constantly told “we did everything we could but you people just didn’t buy enough product.” (I wonder when “ecosystem” will replace “paradigm” as the word most likely to be used in any HP presentation. The guy that spoke about HP-UX used “ecosystem” several times.)

Prather acknowledged that people have been asking HP to make some decisions about things such as OpenMPE Inc, Open Source, etc. but pleaded:

• That CSY was too focused and busy providing for those planning to migrate; and,

• There is no clear right answer to supporting other alternatives. But he wants us all to know that CSY is “invested in looking at all the answers.”

Memo to CSY: You do not have to find the one right answer, just do something and stop wringing your hands in indecision. Note that it appears to those of us outside HP that less time was spent deciding to EOL the HP 3000 than has already been spent deciding whether CSY will allow any alternative to transition and migration.

There continues to be no sense of urgency to come to terms with any such alternative. In an area where perception is everything, the perception that CSY is just trying to wait out issues like OpenMPE because they will eventually go away is rather damning.

There was no mention at all of Interex’s five “Issues Needing Resolution.” Yes, these just recently appeared on the Interex Web site; however, they are issues that have been openly discussed on 3000-L and elsewhere since at least December, so they should come as no surprise.

Next up was Loretta Li-Sevilla, CSY Product Marketing Manager. In fairness, she gave a very polished presentation under difficult circumstances — trying to convince people that HP is doing all these great things to try to save them money on the millions of dollars of hardware, software and services they did not plan to be buying in the first place.

She focused on three areas:

• Making the transition more cost effective

• The need for more guidance

• The need for help in “coming up to speed” on HP-UX

After saying that programs and incentives vary with region she went into describing a number of items that she claimed will deliver up to $150,000 of value per system.

On the guidance issue, she went into all the services available from HP (primarily for the Enterprise customers) and introduced the three North American Platinum Partners that will provide and coordinate migration services for the small to mid-size customers.

On the “coming up to speed” track, she announced invent9k (similar to the invent3k program), available in late April (see jazz.external.hp.com); a Web-based tool comparing MPE and HP-UX commands; an HP-UX loaner program (for qualified customers, i.e. those committed by contract to a migration); and a series of five free monthly Webcasts starting April 23.

Next up was R&D Manager Dave Wilde. He announced that MPE/iX 7.5 would be out in the second half of 2002 and would contain, among other things, native Fiber Channel support, VA7100 support, WebWise and Sendmail included in FOS, disks greater than 4Gb usable as LDEV 1, a shutdown CI command and increased user logging capacity.

Wilde mentioned that the SIB results are in and HP is looking at them to determine what they can do — again, I doubt the results were much of a surprise so I was disappointed he did not speak to them.

Since I was on the committee that created the SIB along with CSY representatives, I know that the SIB ballot questions were not only reviewed by HP, but in some cases altered at their request. As I write in late April, it has now been almost a month since the Solutions Symposium, and we have still not heard anything about the SIB from CSY.

Wilde also said that CSY had not decided yet if there would be a 8.0 release. It was under consideration.

I suppose the big news was an official blessing of HP Eloquence for those customers who feel the need to have IMAGE as a database on their chosen transition platform.

What then followed were presentations by representatives of HP-UX, HP NetServers and Linux.

Mark Hudson, Worldwide Director of Marketing for HP’s Business Systems and Technology organization, talked about HP-UX. While he did not explicitly refer to all the speculation that HP-UX would follow the path of MPE in only a couple years, he did several times emphasize that HP-UX is key to HP’s overall business strategy and would remain so. Of course, he also boasted about HP’s cluster technology even though most things I’ve read suggest that the one thing HP-UX will take from the merger with Compaq is their (the old DEC’s) “superior” cluster technology.

This guy was a real dynamo. The really depressing thing is he just happened to start his HP career in CSY. In fact, he made a point of saying his most enjoyable years with HP were those years spent in CSY. Imagine if he had stayed to head up marketing in CSY, instead of the succession of bland apologists we’ve had? (Question: Can you name even two?)

Oh, and the guy who spoke about HP’s Linux Strategy, Mike Balma, also began his HP career at CSY. Do you see a pattern?

I’ve been pretty hard on HP through this, and I’m sure there will be those who disagree with some of the things I’ve written. However, just as HP’s announcement was not personal — even though there are times when it seems that way — but just business; so too are my comments not personal, just business.

Even though the merger and the future of MPE continued to dominate 3000-L, there were still hundreds of postings where people graciously shared technical information. We report on some of them here.

As always, I would like to hear from readers of net.digest and Hidden Value. Even negative comments are welcome. If you think I’m full of it or goofed, or a horse’s behind, let me know. You can reach me at john@burke-consulting.com.

FTP and invalid IPs

You might have missed this one but if you use FTP, you’ve undoubtedly been bitten. Occasionally, when using FTP, an “open” is issued against an invalid IP address and FTP appears to hang. James Hofmeister provided an analysis and a promise of a fix:

Just a quick note for those folks who mentioned that FTP does not time out on a bogus IP address: ftp 99.99.99.99

What I found was it actually was timing out, but after 10 iterations of the configured TCP timers.

Path: NETXPORT.GPROT.TCP

[4 ] Retransmission Interval Lower Bound (Secs)

[180 ] Maximum Time to Wait For Remote Response (Sec)

[5 ] Initial Retransmission Interval (Secs)

[4 ] Maximum Retransmissions per Packet

So in the example of the default TCP timers,

Initial SYN

5 seconds Retransmit SYN

10 seconds Retransmit SYN

20 seconds Retransmit SYN

40 seconds Retransmit SYN

equals 75 seconds, and this sequence was performed 10 times for 750 seconds or 12 minutes 30 seconds and resulted in the 50 Sin’s /?;~) (50 SYN packets sent out).

If your timers are configured different from the default, then YMWV.

This problem is identified as SR 8606231617. I will post back to the 3000-L newsgroup when the FTP client is repaired and patches are available at the HP-RC.

Hate vi? Then this may be for you.

Another one you may have missed. Generally, most long-time MPE users find vi difficult to learn to use. But what are you to do if you need to edit bytestream files? I’ve long recommended installing Samba/iX because you can then use your favorite Windows editor. Mark Bixby announced another option:

“Okay, all of you vi haters will be pleased to learn that there is now an HPEDIT beta patch available that adds support for HFS-syntax filenames and bytestream format.

The patch IDs are: MPE/iX 6.5 - EDTMX71(A); and for 7.0 - EDTMX71(B).

There does not appear to be a patch for 6.0. These patches are hot off the press and I haven’t had time to personally try them. I will try to get the 7.0 version installed on invent3k in the near future.”

How much memory is ideal?

A memory vendor is pushing one formula and frequently the advice on 3000-L has been “How much can you afford.” As we’ve seen in previous months, it may actually be possible to have too much memory. So what does HP say? Guy Paul, an MIS manager of 3000s for HP, contributed the following:

We use the formula:

1.5Mb X #processes + 100Mb

You can get the number of processes running by doing a

SHOWPROC PIN=1;tree;system > temp

LISTFTEMP TEMP,2

(The EOF-2 is the number of processes.)

You also want to get your number of processes during your systems peak period. This is not an arbitrary formula but is derived from the formula used by the dump-to-disk calculations: dtodsiz.hp36375.telesup ‘calc’

The statement “buy as much as you can afford” has some merits but if you were to say that to your non-technical manager I think you know what their reaction would be. I find presenting a factual number is more likely to open their eyes.

Another consideration on memory is related to MPE/iX 7.0. With 7.0 it now scales your MAXPIN (SHOWVAR HPMAXPIN) based on your system’s memory using the following algorithm:

mem <= 64mb MAXPIN = 400

mem <= 256mb MAXPIN = 1000

mem <= 512mb MAXPIN = 2000

mem <= 1gb MAXPIN = 4000

mem <= 2gb MAXPIN = 8190

mem > 2gb + BIGPIN enabled + PP1 MAXPIN = 12000

If you have a 6.0 system with 512mb memory running 2,500 processes, you will not be able to run those processes after upgrading to 7.0.

An interesting warning

Faster is better, right? Perhaps not always. Someone was asking about putting a 100Mb card in a Series 918. While technically this can be done, John Wolff added this word of caution:

The proper answer to your question is don’t bother to install the 100Mb card in your 918. About 18 months ago or so we installed one in our 928. It turns out that the design of this card is essentially bit serial (software driven) and totally a function of the CPU speed. The 928 has a rating of 48 MHz, which is 40 percent more than the 918. Upon testing we found that the “so-called” 100Mb rate was no better (and in some cases worse) than the 10Mb NIC we replaced. After questioning HP and much lab research they admitted that this was all true. They said the documentation for the product should have “Not recommended” it for the low-end 9x8 series. This is a poorly designed throwback interface that should not have been offered on the 9x8 boxes. The 9x9 boxes have a different IO and NIC design and are okay.

Someone is still developing for the HP e3000

Here is a nice little piece of work from Lars Appel. There have been periodic complaints in the past about the problems encountered debugging VPlus without a DTC terminal. In fact this has even led to requests for enhancements to MPE/iX. Thanks to Lars, it looks like at least one of the following problems is solved:

• If you are trying to debug a VPlus program on MPE/iX and need to redirect the forms display to another session LDEV on the same or another 3000, which does not need to be a nailed DTC LDEV as needed with the classic “FILE termfile;DEV=nnn” approach.

• If you are trying to migrate a program to Unix or Linux, but want to keep the calls to the VPlus user interface during the first steps of porting or testing (by redirecting the VPlus dialog to a 3000).

Find a few more details and downloads at invent3k.external.hp.com/~LARS.APPEL under the heading Remote VPlus.

This is open source software and comes with no guarantees. It worked for me as far as I used it, but your mileage may vary. Contributions to this little project are highly welcome (e.g. source code patches or enhancements, additional tools, better documentation, example programs, etc.). I haven’t tried the client part on Solaris or Linux so far, but it does seem to work on MPE/iX and HP-UX (at least for what I used).

Oh, by the way, the intercepted intrinsics include VBLOCKWRITE and VBLOCKREAD, but I didn’t have test programs at hand, so there is yet another area where you could contribute (programs or test result feedback).

Another system abort warning

Here’s a nasty little gotcha (seems to be a theme this month), again from James Hofmeister:

I am working on a SA700 caused by a sockets application performing a “gethostbyname” after a network shutdown was in progress. The “:netcontrol stop” told NETCP.net.sys to shutdown all system network process and delete the network data structures, but a timing window exists where the procedures called from “gethostbyname” tried to read a data structure which NETCP “just” deleted. HP is working to fix this problem. But the moral to this story is to avoid this problem and all problems in general, HP Recommends as Standard Operating Procedure:

• Shutdown all socket programs

• Shutdown all network socket utilities like network spoolers or network data shadowing

• Shutdown/Abortjob all Telnet and VT sessions as well as DSCOPY FTP, etc.

• Stop the ARPA Services “:Abortjob JINETD” and the NS Services “:NSCONTROL STOP”

• Netcontrol Stop.

Following this procedure will avoid the aforementioned SA700 and other problems in general.

Plan B

I’d like to give a little nod in the direction of Wirt Atmar for his suggestions on how to avoid migrating. It makes for interesting reading:

“A number of people, including some at HP itself, have recently described staying on the HP 3000 indefinitely as being very risky. I completely disagree with that assessment. In fact, I believe that it may well be the least risky thing you can do, thus I’ve spent the day writing a general reply to that assertion. Because I wanted to include pictures in my response, illustrating what we do in our own circumstances, I’ve put the reply up as a Web page at aics-research.com/planb.html.

A primary moral that is implicit in this Web page is that you don’t have to make up your mind on what you’re going to do in the next six to 18 months. If you are a part of an organization where staying on the HP 3000 is a very real consideration and makes good sense, you realistically have 10 to 15 years to decide. There’s simply no reason to be rushed into your decision. A lot of things will disappear from the HP 3000’s ‘ecosystem,’ most notably HP itself. But many things will stay the same too.”

John Burke is the editor of the NewsWire’s HiddenValue and net.digest columns and has more than 20 years’ experience managing HP 3000s.