Online Extra No. 58

Coming soon: The Best Of Hidden Value & net.digest

Presses were ready to roll on the five-years'-full edition of The Best of Hidden Value & net.digest, a special book of tips and new 3000 techniques edited by the NewsWire's own John Burke, who authors the two columns each month. The 64-page, spiral-bound technical resource is brimming with more than 250 tips and technical discussions and will make its debut at the February conferences for 3000 customers, where it will be available for free. Subscribers to the NewsWire automatically get a copy of the book, too.

Extra-special Web tool, free: Analog/iX

We were excited when we got the news that Andreas Schmidt had ported the Analog Web tool to MPE, mostly because it proves the computer is becoming a real choice for Web services. Analog, you see, is the leading choice for tracking the contents of Web server logs, and you'd hardly need it if the HP 3000 wasn't racking up Web traffic. Schmidt said he's ported the software, absolutely free, because "For MPE/iX such tools may become more important in the future HPe3000 world. It is designed to be fast and to produce attractive statistics." The engineer who delivered the Ploticus graphics program he wrote about in September and October issues, Schmidt has a Web page at http://www.hillschmidt.de/gbr/analog.htm that provides e3000-ready code and installation instructions.

The program's creators have delivered a tool that "allows Webmasters to take an incomprehensible list of traffic data and create intelligent, comprehensible and very useful - from a site marketing and planning point of view - reports that anyone can understand," according to one online review. "Wading through such a file by hand is pointlessly time-consuming, which is why log file statistics packages exist." About one in four Webmasters use Analog, which takes raw log files and creates reports showing the number of unique visitors per day, the number of page impressions, which files were downloaded (and how many times), the type of browser and operating system used , CGI command arguments, host names and/or IP addresses and more. Reports can be tailored to suit the user, and created in HTML, ASCII or machine-readable output.

What's so super about Superdome?

When it was rolled out at HP World amid smoke and corporate fanfare, Superdome seemed like a pot of gold at the end of the HP server rainbow. HP 3000 fans complained that their system wasn't mentioned among the supercomputer's platforms. And speculation on the show floor ran wild that the computer really would run MPE/iX, while the CSY brass said the clustered and partitioned server was a possibility for future years' 3000 lineups. The first review of the iron showed more cold, hard rain than rainbow, however.

Negative comments from a Merrill Lynch analyst poured doubt on the ability of the system to raise HP's horsepower in the Unix marketplace. "We saw benchmarks for HP's new Superdome server,'' Merrill Lynch analyst Thomas Kraemer said, referring to tests that determine the performance of the machines. "The benchmarks for Superdome were tepid." HP has said that Superdome server will be the catalyst to re-energize its Unix business. But Kraemer said the new servers fell short in performance and price when compared with Unix systems from IBM and Sun. HP, IBM and Sun Micro are vying for the business of moving corporate operations online.

"It did not beat the high-end servers that IBM introduced a year ago, and it was more expensive than Sun Micro's on a price performance basis," Kraemer said. "As such, we do not think that this product will be a strong catalyst to produce a turn in HP's Unix business." If so, then the 3000 division will look wise for focusing on the N-Class rollouts set for early February.

HP's Unix marketing manager Mark Hudson told CNET that the computer's marks will improve, even though IBM did beat it in the first TPC-C benchmark, 221,000 transactions per second to Superdome's 197,000. The IBM S80 server had a cost per transaction that was 71 percent of HP's price. HP said it beat Sun and IBM on a real-world benchmark of SAP's software, and Hudson predicts that TPC-C benchmark will increase to about 300,000 this summer.

CNET reported that Hudson said the Superdome benchmark was set using "only" 48 processors; the computer accommodates as many as 64. HP also wanted to stress how benchmarks get better all the time. We know about getting better all the time; it's the whole reason new computers are designed and released, to leapfrog competition. What we heard about Superdome at its smoky introduction was that it was supposed to leapfrog competitors from day one.

Conferences galore in 3000-land

By now many made plans for going to a 3000 conference in the next few weeks. HP's chipping in a sales conference for the e3000 in a Black Forest retreat in Germany, starting this weekend. After the Sunday brunch, resellers from Europe and elsewhere will hear "magical and exciting news regarding the HP e3000 product family." We think the letters N and A will be predominant in that presentation. The next two days include details on roadmaps, Web-enabling tools, ordering, configuring and upgrades for what's being announced on Jan. 21.

Expect more of the same at the Solutions Symposium starting Feb. 7, when the four days of training kicks off with a keynote speech from CSY general manager Winston Prather. The agenda is thick with new 3000 capabilities, all hosted in California, where the expected announcements might help provide power to an electron-starved state. Details are at <http://www.interex.org/conference/hpe3000solutions2001/> and conference organizer Interex has extended the Early Bird Rate through Monday, January 22. Inaugurate your new year with a tour of the 3000's new capabilities.

Or wait just a little while longer and get better marks for your performance. The Florida Regional
User Group puts on its fifth annual Performance Conference starting on Valentine's Day, with talks about the 3000's performance, database speed, networking and even "I/O Performance: Living with 400 I/Os Per Second" from Robelle's Neil Armstrong. How you get that many IOs remains the mystery we expect to be solved after the February introductions. Get more details on the conference in the land of recounts at the FLORUG Web site, <http://www.florug.net/NEWSLETTERS/5thperf/5thperf.html>

Not to forget the SIG3000 meeting that overlaps the FLORUG show, either. At $100 it's value-priced, and it's returning to its roots at the HP labs this year with meetings in the Oak Room in Cupertino. Rub elbows with HP engineers and get in on the ground floor of planning for 3000 products from HP. The SIG meetings will provide "detailed technical updates from Hewlett Packard's SIG liaisons. The agenda includes a Languages Day, all about UI/VPLUS, JAVA, RAPID, COBOL, and FORTRAN; aSystems Day covering Client Server (C/S), MPE, SYSMAN, Consulting, and SIG-Web, and a Database Day about IMAGE/SQL and OMNIDEX . Interex is also promising "a special HP Webcast" on Feb. 14. We'll look for the familiar voice of HP's George Stachnik on that day, and look for you at the SIG meetings. Register at the Interex Web site for SIG3000 at <http://www.interex.org/advocacy/sig3000/>

When does PowerPatch 2 grow cold?

MPE/iX 6.5 has earned its second set of patches, being called an Express release because they involve some subsystem products as well as the operating system. But users have noticed they've got a Feb. 28 deadline to order the patch -- or at least that's what the documentation seemed to say. Patch expert Jon Cohen of the 3000 labs explained the software won't be pulled off the shelves on that date, but it's a lot easier to order until then:

"6.5 Express 2 is indeed an Express Release, since it consists of a new SUBSYS tape (which now contains those new HA products you mentioned), but it also consists of a new Powerpatch tape, which contains all the GR'd (Generally Released) patches that were available at the time of the Powerpatch build. Think of the Express Release {SUBSYS + Powerpatch} as a superset of the {Powerpatch} tape. So, there is a 6.5 Powerpatch 2 tape, which is a component of the 6.5 Express 2 release. Customers who don't have any of the new products on the 6.5 Express 2 SUBSYS tape may still be interested in the 6.5 Powerpatch 2 tape on its own."

"The note on "last date to request this is 2/28/2001" is also a bit complicated. The reader's digest answer is that 6.5 Express 2 will remain available and orderable until the first customer ship date of 6.5 Express/PP 3 or the "end of support (EOS) date for 6.5", whichever is earlier. Before you ask, we haven't determined when either of those dates (6.5 Exp 3 or 6.5 EOS) will happen."

"The "last date to request" is associated with how HP's manufacturing group manages these releases. When we ship a release, the manufacturing group (called Electronic Software & Information or ESI) actually generates and holds onto customer orders for every customer with a support contract with us, and we provide each customer with a fax-back card containing order information. ESI holds onto this mass of generated orders until it receives the fax-back card, at which time they process the order. If they haven't received the fax-back card, they discard the unfulfilled orders after a period of time. For 6.5 Express 2, ESI will discard the left over orders after February 28th. Using the fax-back methodology has efficient for ESI and also has the customer benefit of quick turn-around time and ease-of-ordering. After 2/28/01, it takes ESI a bit longer to fulfill customer orders, and the effort for customers to order it is a bit more."

Our thanks to Cohen for explaining how HP manages its patches for the 3000. Customers who want even more patch information can get it at the Solutions Symposium, where our net.digest and Hidden Value editor John Burke is giving a tutorial on how to patch a 3000 system. He'll also have free copies of his upcoming book, "The Best of Hidden Value and net.digest."

A few notes on what's not installing in 6.5

In a word, Samba and Apache, but there appears to be a way around it. [We'd also like to note that in our Samba/iX item of last month's Extra, we slipped up and assumed the Samba 2.07 code was downloadable from the CSY Jazz Web page which describes the software. Not so fast on the download. You've got to get it as a beta-test Response Center patch, according to Samba expert Michael Gueterman of Easy Does It Technologies.]

Tracy Johnson reported that Samba doesn't seem to want to install by itself when putting 6.5 on the HP 3000. He reports:

"AUTOINST on my 6.5 upgrade from 6.0 fails on the Samba job (I00IAMBA.USL.SYS). I assume this is because I have SAMBA20 on my machine and AUTOINST is trying to put back the earlier version. Seems I cleaned up the old SMB POSIX subdirectories after I upgraded to SMB20. So the job tries to copy something that isn't there.

Seems the workaround is simply to edit I00IAMBA.USL.SYS so it does nothing (delete 2/LAST and add !EOJ on line 2) and AUTOINST will happily continue. Otherwise you get INSTERR 23 and next time you run AUTOINST you'll be prompted to continue with the old audit trail or start over."

Johnson's report prompted another from Doug Werth at Beechglen Development, who said that the Apache Web server failed to install for him:

"During a 6.5 update. the I-job for Apache (sorry, don't have the job name in front of me) was issuing a Posix 'chmod' command. It attempted to follow a symbolic link where the target of the link had been removed. The error was reported back to the CI and the job subsequently failed.

I did not see this as a problem worthy of canceling the entire job. Nonetheless it did abort leaving me with the options of

A) Modifying the job to put a !continue before the chmod step thus ignoring it
B) Modifying AUTOLOG to tell AUTOINST that the job had completed successfully
C) Cleaning up the initial cause of the problem by removing the symlink and restarting AUTOINST.

I was a little irritated at the time because I didn't think the job should have aborted based on the setup of the user files under Apache. But then I realized that this is the first release of Apache that is included with the standard update tapes, so all in all the process went very well."

Why you should use Critical Item Update

Database wizards more seasoned than most humans lobbied for Critical Item Update years ago on the HP 3000. The software hasn't really come into vogue, though it's hard to see why when the benefits get examined. Ken Sletten, chairman of the SIGIMAGE special interest group, did the examination recently on the Internet and reported the benefits:

"We made almost immediate pervasive use of it when it first came out many years ago. Besides the obvious and in most cases substantial performance improvement for doing just a DBUPDATE of one search or sort item instead of having to DBDELETE / DBPUT the entire record, there are other related benefits.... A few the pop to mind right now:

(1) QUERY becomes a *much* more useful tool for database maintenance; especially in conjunction with IMAGE B-Trees: You can slice and dice your search criteria, and update just the one or two or whatever items that you need to change in each of one or two or 10,000 records. Before CIUDPATE, for practical purposes you had to write a program to do many database maintenance operations that involved changing a critical item.

(2) Because you can now change individual critical items in a dataset with DBUPDATE, you are much freer to make additional items search items with your favorite database tool; and subsequently to add B-Trees to the masters those new search items are connected to. This opens us opportunities for more flexible, easy, and efficient reporting.

(3) For "extra credit" (some people will say that I have gotten carried away adding paths); and depending of course on your applications, load on your machine, disc space, etc.; you can (for example) do what we have done without paying any significant performance penalty (notice I did not say NO penalty; YMWV): We have a total of 66 paths from Details to AUTOs, for four of the most pervasive key identifiers in our system: DOCument number, PART number, part unique identifier, and PROCedure number. This allows essentially instantaneous checking of one or two AUTOs (two in some cases; for DSEM advantage and / or to avoid linking very small details to huge masters), to determine if a particular unique search value exists or does NOT exist ANYWHERE in our database...... go try that on some of your relational databases that do not have hashed keys; and see how long it takes to run a B-Tree search on maybe a dozen tables each with a few million records; just to determine if the value exists or not..... oh, yeah, I suppose: <smug_alert>.

Robelle warns about disk-eating log files

Robelle Solutions Technology began adding fresh items to its Web site on a daily basis during December, and the site has a become a regular stop for us in our Web browsing here at the NewsWire. Robelle has a staff as experienced in the 3000 as anybody's and more than most, so they research both new features and not-so-new gotchas. We liked this one from Neil Armstrong posted at the site: