Number 61 (Update of Volume 6, Issue 5)

Enhydra moves into additional HP platforms

HP will be distributing Enhydra through a marketing agreement with Lutris as a solution for customers who want something less complex and less costly than the Bluestone application server, offering Enhydra for its Windows NT and Linux platforms in addition to the MPE/iX version for the e3000. Lutris said the solution will enable service providers, value added resellers, telcos, and individual and corporate developers to easily create and deploy Internet and wireless Web applications. The addition of the HP Netservers as a platform for Enhydra might make it more strategic to e3000 managers.

A report in this week's Infoworld included a comment from a Giga analyst who asserted the Enhydra server isn't something people would consider bulletproof or high-end. Lutris' Greg Schwartzman said the analyst wasn't all that familiar with Enhydra, but Lutris would be briefing him soon. From another quarter, HP's Java guru for the 3000 Mike Yawn has called Enhydra the killer app for Java on the platform. There are alternative Java application server alternatives for e3000 customers, such as the Web/iX package from Advanced Network Systems, Inc.

Jikes really does make Java compiles faster

After our Flash Paper report on the new Jikes compiler for Java applications surfaced, a Special Interest Group leader gave us some field test data to confirm the speed increase for 3000 users. Cortlandt Wilson of SIG-Consult said "Jikes! It's fast. Of course a lot of us develop and compile on our PCs. Jikes is only a compiler. The run time environment -- the Java VM (Virtual Machine) -- is slow to load, which is a problem for "stand-alone" programs. Web programs are fine because the listener program is a Java program that is already running in the background as one of the web server processes.

I'm starting to see that Java and the things that come with it -- Enhydra -- are more strategic than the Web servers. Some application environments automatically recompile changed programs the first time they are run. Jikes can help here."

Mike Yawn of HP's e3000 Java lab reported some specific times on using the freeware compiler, available from the LegacyJ Web site:

"I got a chance to download and try this out, and the performance is quite impressive. I should also mention that in SDK 1.3 there is a completely rewritten javac compiler that is quite a bit faster than the 1.2.2 javac compiler used previously. Using my sample NFL application from the Solutions Symposium as a benchmark, I found the following compilation times on my development system (the package contains 7 classes)

compile package using 1.2.2 javac: 1 min, 11.680 sec
compile package using 1.3 javac: 40.494 sec
compile package using jikes: 5.391 sec

"If you're doing a lot of Java development, you'll probably want to give this a try."

And Evan Vaala of Rockwell Automation reports that using Jikes on the e3000 makes it possible for an application to modify itself for speedier access:

"Jikes was created not only to speed development time, but more importantly to speed up the process of Java applications dynamically creating/compiling/instantiating classes on the fly. This is how JSP works, and also its forerunner JHTML.

"Jikes *greatly* reduces the startup times of application servers, such as Web/iX, Orion, Enhydra, Weblogic, etc... Not only the startup time, but also if on the fly changes are made to JSP pages, this compile time is reduced as well. "

Vaala notes that the standard compiler javac "is a Java application with a JNI wrapper. Jikes is a binary executable compiled for each platform, thus the speed increase. You can actually compile a program by executing the following at the command line:

java com.sun.tools.javac.Main MyJavaClass.java

I think that it is very cool that you can write a program that will alter itself while it is running. Very powerfull! This is one of the reasons that a J2EE application server does not need to be "shut down" to install, or reinstall an application. The application server will extract/deploy all classes/files out of the ..ear file, getting the .war file where the .jsp files are located. The .jsp files need to be dynamically compiled while the web module for the application is being initialized. This is where Jikes plays a very important role."

Some existing 3000 peripherals plug and play with new systems

Customers calculating the cost of moving up to the newest HP e3000s might be able to make existing disk and tape devices carry over. But the devices will have to be external, and there are some guidelines. We asked 3kworld.com Content Manager Chris Gauthier, who works down the hall from the Client Systems configuration bays shipping the new 3000s, to take a peek and see what can plug and play.

Regarding peripherals with Fast-Wide Differential SCSI interfaces, Gauthier said, "From what I understand right now, it's probably best to connect any existing FW/Differential SCSI devices to either the single-port (A4800A) or dual-port (A5159A) FW/SCSI add-on card. Supported peripherals currently include the Smart Storage, Jamaica, AutoRAID, Mod 10/20, EMC and XP256.

For existing Single-Ended SCSI devices, Gauthier reports, "Any supported SE-SCSI devices can be connected to the Ultra2 SCSI LVD/SE port on the MFIO board. (The port will auto-range from the low-voltage to the standard high voltage devices through detection.) That would include DDS and 7980S/SX tape drives. However, I wouldn't mix standard SE-SCSI devices with any true LVD/SE devices on that same port, otherwise the port will get confused and not work at all."

Kevin Cooper of HP's 3000 labs weighed in on performance considerations using peripherals with the new systems in an Internet message. He said, "Disk IO performance can be much better on an N-class system, but it will depend on your current utilization. If disk is not a bottleneck, it may not matter which type of disks you use. There are certainly many better options than a Nike in RAID-5 if performance is your primary concern.

"As to LVD or HVD, the rated speeds of an LVD card is 80 Mb/second, and an HVD with F/W SCSI is 20 Mb/second. But that may not make a bit of difference if you are not even approaching these limits. The only LVD drives supported on MPE/iX 7.0 are the drives internal to the system. For external drives, HVD is only option right now on [base] 7.0."

Displaced 9x7s might be fewer in number

For the thousands of 9x7 HP 3000 server owners out there, the next 12 months could be the most interesting time of ownership. HP's shutting off the system from support and preventing it from running the latest operating system, and in some places that means the end of the 3000's service. Our March article on the choices for 9x7 owners included a story from a Lockheed Martin division in Orlando, Fla. where the system manager expected this would happen, even though the 957 there was outperforming a host of HP 9000s.

We thought we'd add a bit more information on the situation at Lockheed Martin. The Unix systems which are in line to step in for the 3000 are HP's. This is an example of the kind of attrition that is acceptable to HP, since HP's Unix systems are stepping in for the 3000. Whether it's acceptable to the customer is a good question, since the IS coordinator has a different opinion than his management.

Lockheed Martin is a big company, with lots of operations and many more HP 3000s running elsewhere. The point of our article was to show how 9x7 managers were reacting to the A-Class offerings being positioned as replacements for their systems. In this case, it looks to us like any 3000 offering that required a capital outlay would have been cause to decommission this particular 957, since it's got an April 2002 end of supported life -- like every other 9x7.

The upgrading could well be a savings move, according to one HP authorized reseller. "Most of the midrange 9x7s out there can be migrated to A class servers," he said. "If you evaluate three-year cost of ownership of the current 9x7 and compare it to the cost and support of the A-Class, you will find that, in many instances, it is cheaper to upgrade than keep the 9x7."

There are other ways to go with 9x7s. Several third party firms are promising to support these boxes beyond April of next year. We've already reported on such services being offered by Beechglen Development which reports it is already assisting 9x7 sites as an HP support alternative.

Another source of 9x7 support is at TERiX Computer, a Silicon Valley firm working with InnerVu Corp. to offer 9x7 owners viable maintenance alternatives to fill the HP maintenance void. TERiX says it serves 106 US cities with over 500 field engineers and also offers maintenance programs in Canada, Europe and Australia. InnerVu is the professional services "sister company" to Computech Systems Corp., which bills itself as the nation's oldest and largest remanufacturer of HP 3000 servers. InnerVu points to "16 years of MPE hardware and software experience and is staffed with full-time and on-call HP trained MPE engineers."

In December's FlashPaper we quoted HP sources as predicting that "As the end of support for the HP 3000 9x7 servers approaches, parts will be in shorter supply and support prices are likely to rise." InnerVu asserts that keeping the 10-year-old servers on the job is a real option.

"This is an economic business decision designed to solely benefit Hewlett-Packard," stated Frank Zaccari, President of InnerVu. "HP's business decision to End-Of-Life the 9x7 family does not mean that an organization must change its plans to match HP's."

Here at the NewsWire we are among those who believe such systems essentially disappear from the 3000 community's revenue channel, except for the support fees paid to those third parties. A customer which can't get funding to replace a system may have to scratch hard to get budget to buy software tools, new applications or services. But we'd be glad to hear otherwise. Independent consultant Cecile Chi dropped us a note to say she had a client who "was still getting excellent third-party support on a Series 70 in 1999 when they finally upgraded to a Precision Architecture box." In some cases, it might be true that a functioning older HP 3000 is better than one replaced by a Unix system. And not all of those replacing computers will have an HP logo on them.

Free IMAGE/SQL, a benefit revived by HP

On the other hand, moving away from a 9x7, no matter how well it's performing, could be a revenue-neutral transaction for some customers. The older computers surfaced in the era when IMAGE/SQL was just being released by HP, and some sites missed the opportunity to get the connectivity for SQL offered through the ODBC drivers included with the system. Back in the 1990s, customers could opt-in to get their TurboIMAGE turned into IMAGE/SQL, but only for a limited time with a support fee increase. Now that IMAGE/SQL is part of every new e3000, that's an increase that has disappeared inside the support for the system, which is less costly than its 9x7 ancestors. Rene Woc of Adager explained it to us:

"My understanding is that IMAGE/SQL is again bundled with the A-class and N-class systems," he said. "It will be a nice break for those users who did not buy into the support-fee increase to obtain IMAGE/SQL at zero cost -- and have been reluctant to pay the upgrade fee to convert their TurboIMAGE license to IMAGE/SQL. It also means all the new e3000s have ODBC connectivity via IMAGE/SQL." The increases in support for IMAGE/SQL weren't all trivial, ranging from $120 a year at rock bottom to $7,200 yearly at the top end. Now those charges will appear as virtual discounts to companies still paying them while running on older machines -- if they move to the new systems.

The end comes sooner for some 9x9 systems

The end of support life, that is. HP has notified customers who own Series 939 and 959 HP 3000s. March 31, 2003, less than two years from now, is the final date for HP to support the 939s and 959s. We haven't heard of other dates for support life termination among the 9x9 and 9x8 servers, but HP won't be selling any of those systems by the end of this year.

Correction: they're large files, not big ones

In our report from the March FlashPaper on HP's halt on shipping the 6.0 version of MPE/iX (it stopped at the end of March), we called something big when it should be large. The nuance is more than just a play on English. Stan Sieler noted that in the article, "the last two words should be "Large files." not "big files." The "Large file" is any file larger than 4Gb. (We'd have called them "jumbo", but then they'd be confused with jumbo datasets, which are logically larger than 4 Gb, but are composed of non-Large files of up to 4 Gb each."