| Front Page | News Headlines | Technical Headlines | Planning Features | Advanced Search |
Click for Quest Software Sponsor Message News Icon

July 2000

Number 52 (Update of Volume 5, Issue 9)

HP's busy cutting patches for 6.5

Like any major operating system release, MPE/iX 6.5 has some bugs in its initial version. The version was written to satisfy the needs of some of the most performance-hungry customers in the HP 3000 community, so its flaws are getting some prompt attention. Within a month or so of release, 6.5 already has patches, items that will be collected at some point for a PowerPatch of the release. Healthcare entities like HMOs are finding the problems, since they're among the first to install 6.5.

One illuminating interchange over the Internet showed a handful of patches either in place or being written to squash bugs that came up once 6.5 started shipping. Rich Holloway of Providence Health Plans reported on four bugs, and Bill Cadier of the Commercial Systems Division Escalation Team reported on the patches available to fix them -- some still in development.

"We would find the system would hang when doing fairly heavy processing," Holloway said. "This was caused by two memory management issues. Both issues have been resolved and a general patch has been released. One patch resolves both." Cadier said that MPELX75 (which supersedes and includes MPELX66) addresses this.

Holloway said while using Robelle's Suprtool to create a self describing file under 6.5, "we found the file end of limit would be a negative number. This was caused by a file I/O error. This too has been resolved and a general patch released." HP's Cadier said patch MPELX71 fixes the problem.

Cadier had just begun work on another patch, MPELX86, to fix another problem, "when an FCONTROL 2 and FCONTROL 6 that basically tells the system to take the information in memory and write it to disk." Holloway reported. "What it's supposed to do is take a block of records, write it to disk, then get the next block. The problem is that the start for the block always goes back to the start. Example, let's say there are 8 blocks of 100 records to be written. It takes the first 100 records and writes them, then it takes the first 100 and the second 100 and writes them, then it takes the first 100, the second 100 and third 100 and writes them. By the time it gets down the first 100 has been written 8 times."

Holloway also outlined a fourth bug that deals with disk mirroring and number of spindles. "What happens is there is a PQUAD table, or table in PQUAD, or at least something to that affect that is used by mirroring. The more spindles you have and the more writes you have causes the use of this table to be increased to its limit. If the limit is hit gracefully, all that happens is mirrored partners are lost. If the limit is not hit gracefully, it will cause a system abort with an Abort 1101. We currently have about 250 spindles and are testing the patch now." HP said the official patch ID for the PQUAD problem will be MPELX84.

Holloway also said his site "had system halts when we installed 6.5, and increased our 997 to 12 processors from 8 and 16Gb of memory from 4Gb. The added processors and memory were removed but 6.5 left on, and the box is running our production fine. HP has come up with different memory and processors and will attempt the hardware upgrade this evening."

Cadier summarized the patch activity on the 6.5 release by saying that "MPELX66, MPELX71 and MPELX75 are all general released now, and I believe will also be on 6.5 PowerPatch 1. MPELX84 and MPELX86 will are yet to be done, so I cannot say with certainty that they will be general released in time to get on to PP1. If not then certainly PP2."

HP doesn't have an announced date for the 6.5 PowerPatch 1, although a few customers report they've been told it will be available in the fall. Individual patches which are in general release are available immediately through the HP Response Center.

Customers await 6.5 PowerPatch, but not HBOC

HP 3000 managers are a cautious lot, applying new releases of the operating system only when a first round of patches have been collected into what's known as a PowerPatch. But early reports show that one of the leading application suppliers for the 3000 is prodding customers into early adoption of the latest 6.5 MPE/iX. HBOC's Payor Solutions Group sent letters to all Amisys customers that said the company would not be supporting the 6.0 release of the operating system for its sites beyond October 1, according to several customers. Officially, Amisys hasn't even been released for 6.0.

We first reported on HBOC's haste in getting onto the 6.5 release in our April FlashPaper, noting the company wanted to move all of its customers to the MPE version that many in HP were calling "The Amisys Release." 6.5 benefits from the expansion of many system limits, especially those that limited files to 4Gb. But the backwards compatibility of COBOL applications between MPE 5.5 -- the version most Amisys sites are using -- and 6.5 is in question. The compiler was updated in 6.5 to generate code that takes advantage of new 64-bit millicode routines.

While a few 3000 sites are already running successfully on 6.5, nearly all of the customer base is waiting, as usual, for the first PowerPatch release of the operating system. The software is a by-request, or pull release, meaning HP's not sending it automatically to its entire installed base to give it the widest test base. Amisys customers have been told their supplier won't be supporting the 6.0 version of MPE/iX, and the sites are expected to migrate from 5.5 directly to a 6.5 release.

At least one customer noted that in the past HBOC took months to catch its application up to the 5.5 release for its customer base -- and now appears to be on the other end of the adoption curve with 6.5.

Configure Apache easier with a GUI

Web administrators using the Apache server on HP 3000s can look forward to easier administration of Apache, using Windows software called Comanche. The Apache Web server has a good deal of its power invested in its rich configurations, but keeping them tuned can be challenging for users new to Web server administration. HP's Mark Bixby, the engineer who ported the Apache server to MPE, reports that Comanche looks like a good candidate for helping out on the HP 3000:

"Based on what I read about Comanche at http://www.comanche.org, you should be able to install Comanche under MSWin and then access your e3000 Apache config files via Samba," Bixby posted. "I haven't personally tried this, though." Comanche is free software supported via the open source community, the same kind of support structure that Apache enjoyed on its way to HP-supported status on the 3000.

HP World meetings for Maestro and Web surface

Although the conference is still weeks away, HP World volunteers are already announcing meetings to be held around the many roundtables, keynotes and seminars set for Sept. 9-14. Maestro users running systems under MPE/iX (and now supported by ROC Software) are among those invited to the The Maestro Birds of a Feather (BOF) meeting Sept. 12 from noon to 1 PM. As it's a lunchtime meeting, Interex will be providing "a tasty box lunch," according to the BOF organizer Keith Olson of Advanced Computing environments. "The Maestro BOF is an opportunity for users of Maestro (aka Tivoli Workload Scheduler) to get together and discuss job scheduling techniques, bugs, enhancements, releases, and vendor relationships. The BOF is an informal technical users meeting. In that spirit, representatives of the vendors, Tivoli Systems and ROC Software, are welcome. This will be the third year in a row for a Maestro BOF at HPWorld, and we are looking forward to getting together again."

Another lunchtime meeting takes place on the following day as the SIGWeb special interest group gathers from noon to 2 PM. SIGWeb co-chair Michael Gueterman of Easy Does It Technologies says that the meeting will include "'four or five ten minute talks' Some potential topics that have been [considered for discussion] are Web Servers: Apache/iX, WebWise; Web Application Servers: PHP, AutoBahn II, Powerhouse Web, Web Dimension, ASP and ColdFusion; JSP Java Servers; Scripting and Integration through perl, CGI and XML; and simple steps for getting a Web server environment up and going."

No, 6.5 doesn't need 4Gb of memory

Over at e3000 North American distributor Client Systems, "Geek at Large" Chris Gauthier has weighed in with more information about what a site needs to get the latest operating system up and running. In his "Support Call of the Day" posting at the 3kworld.com site, Gauthier elaborates on a topic first brought up on the 3000-L mailing list, and covered in John Burke's net.digest column -- how much memory 6.5 demands. Gauthier, who supports HP's resellers, said:

"Some of my customers have been led to believe that you need at least 4Gb of memory to run MPE/iX 6.5. This is absolutely not true. Most of the current installed base of HP e3000 machines will not even recognize beyond 3.84 Gb of memory (the absolute mathematical limit of a 32-bit processor). Just because your current machine can't go beyond 3.840Gb of memory does NOT mean that you can't take advantage of moving to MPE/iX 6.5."

"For your own reference, the following machines are supported with these absolute minimums with MPE/iX 6.5. (While these minimums are not recommended for performance reasons, the system will physically function with them nonetheless.)

9x7: 32-48Mb
9x8: 32-48Mb (32Mb is the minimum required as of MPE/iX 5.0)
Original 939: 64 Mb (orig. 939 sold with this minimum.)
Newer 939s, 929s, 959s & later: 128 Mb (orig. sold with this minimum.)
989: 256Mb (orig. sold with this minimum.)
990/992: 128 Mb (orig. sold with this minimum.)
991/995: 256 Mb (orig. sold with this minimum.)
996: 256 Mb (orig. sold with this minimum.)
997: 256 Mb

Gauthier added that the only HP e3000 systems supporting more than the 3.84Gb of memory as of July 2000 were the 929/030s, 939/030s, all 979s, 989s and 997s. Gauthier regularly posts such high technical wonders on the 3kworld.com site, like general memory module loading rules for Series 9x7 boxes.

PowerHouse Web demands more open security

After many moons of waiting for Cognos to release its Web development suite, PowerHouse customers are finding out what software writers have known for years: Posix implementation on the e3000 can have its quirks. In one message posted to the PowerHouse mailing list, a customer complained that his Posix files always seemed to have security problems when working with PowerHouse Web. Dennis Barnes reported:

"I am experimenting with PowerHouse Web 2.29D on MPE/iX and constantly have to modify the security on files in the Posix name space. Every time I compile a QUICK screen, or FTP a file from my PC to the HP 3000, the file security is such that the PowerHouse Web Dispatcher is denied access to the file. The Dispatcher is not running from the account where my application resides.

"Being brand new to the Posix side of things, I don't know if this is possible, but the question is: Can I somehow avoid these constant security problems with my Posix files?"

Later the same day, Cognos product manager Bob Deskin offered an explanation about why the PowerHouse Dispatcher was complaining:

"PowerHouse Web on MPE/iX is an interesting mix. The Dispatcher and PHCGI programs are Posix, but the PowerHouse Web Server is MPE/iX. QUICK screens (actually PH Web screens) are MPE/iX files. If the user and account from which you run the Dispatcher (and therefore the PowerHouse Web Server) has access to these files, there should not be a problem. This in itself has nothing to do with Posix. This would be the same as any user getting access to files in another account. You'll have to open up your MPE security to those files."

"As for FTPing a file, it is not uncommon to have to change the permissions once the file arrives. I'm not a Posix expert, but it may depend whether the file is pulled to the HP 3000 or pushed from the PC."

No need to wait for IA-64 systems

In early July HP began to brief some journalists about its plans to migrate from the current PA-RISC architecture to the newer IA-64 designs, but the activity looked like it was still years away from reality. In one story that appeared on the CNET.com Web site, HP's Duane Zitzner was quoted as saying HP expects its PA-RISC systems to outsell IA-64 systems for the next five years.

If Zitzner was quoted accurately, then his comments explain why few people in the HP 3000 division seem to think of IA-64 as anything but a future. In one interview after another, lab experts and general managers praised the new architecture, but pointed out that it had little to do with meeting customer demands for performance. Now we seem to know why: the stuff won't be ready for datacenter-level performance for years.

While one analyst thought these delays might be a problem, we think they're a blessing in disguise. There's nothing so broken in PA-RISC that it needs to be replaced. And if its successor is still on the drawing board, that lets the 3000 lab focus. Considering how tough it is to staff development labs, nobody's engineering effort needs the distraction of having to build more than one version of a system at a time. IA-64 looks like it's going to have about a 10-year history of being a future at HP, considering that it was first announced in 1994. (Of course, back then, HP was calling it Tahoe, and then Merced, and so on.) Since HP has four more generations of processors in the wings for the PA-RISC line after the PA-8500 rolls out next spring, it looks like IA-64 might have more impact on PowerPoint slides than in any HP 3000 for the next five years. But like HP, we're just guessing on when it will be ready.


Copyright The 3000 NewsWire. All rights reserved.