« April 2017 | Main

May 22, 2017

Making 3000 Memoirs, One Post at a Time

Memoir ProjectFive years ago I was entering into memoirs territory. I had a decent start on my own memoir, Stealing Home: The Road to the Perfect Game. It was time for the 3000 community to have its memoirs, too. A few of the community's leaders shared stories, each a memory, of how the 3000 changed their life.

It was a simple and heartfelt formula I believed might be a book. What happened to the HP 3000 Memoir Project was that it became a dynamic story. Instead of being compiled into pages, the 3000's memoirs are in the History section of the blog. There are nearly 400 stories in there.

Three times a month, a history article gives us insights. I call these Wayback Wednesdays, or Fallback Fridays. Each memory is designed to supply meaning and insight. We can't change what happened to us. We might alter how we perceive it, though, as well as change the direction it propels us toward.

Everyone goes into every life situation with specific expectations. History shapes those expectations. We all try to make sense of what's happening to us; prior events give us context. We imagine how what we're doing in this moment will impact us in the future. Memoirs give us a guide to see how things might work out. Maybe most importantly, we draw on memories to evaluate what's happening and see what to do next.

So when Rob'n T Lewis of South Seattle College asked today, "Is the HP 3000 Memoir Project finished?" I said no. Perhaps it will never be, if there are stories remaining to tell. We told the first of them on this blog in 2007. We're always going to be evaluating everything for meaning, always drawing conclusions—not concluding the storytelling.

The Computer History Museum has an Oral History website section. It includes accounts from Alfredo Rego and Marty Browne of ASK. We're continuing that tradition for the 3000 founders, because everybody wants the last word.

11:26 PM in History | Permalink | Comments (0)

May 19, 2017

Friday Fallback: The White House's 3000

White houseThere's a torrent of news coming out of the White House this week, but there's also a bit of history noted for 3000 users and fans. Out on the FedTech website, editor Phil Goldstein created some history and reported on some more with a story about the HP 3000 being the first computer ever to support the White House. It was 1978 when the 3000 began to aid White House efforts like tracking the desires of Congress.

A nonprofit organization that's been telling stories about the White House since 1961, the White House Historical Association says that Jimmy Carter's first minicomputer was "assembling databases, tracking correspondence, developing a press release system, and compiling issues and concerns of Congress." Goldstein developed a high-points article about those heady days of undercutting IBM mainframes and the swift rise of the 3000. In 1979, for example, the 3000 accounted for 15 percent of all data systems revenues at HP. It was $150 million in orders, up from $50 million in 1976.

The article has its problems with history. The timelines suffer from either a 1984-style rewrite, or rushed research. By the accounting of FedTech (a website run by vendor CDW) the 3000's operating system only lasted until 1997, the computers first surfaced in 1972, and HP began to develop it in 1968. Wrong, inaccurate, and misunderstood, those dates are. These things happen when a story's subject is Old Tech. Who'd care if the facts aren't accurate. The 3000's dead, right?

Screen Shot 2017-05-20 at 3.08.37 PMThere's plenty to appreciate in the article. Appropriate links to resources like 579-page The HP Phenomenon, and the HP Computer Museum. There's a link to a book, Managing Multivendor Networks, that covers the 3000 and was written in 1997. Wait, that's supposed to be the same year as MPE was wrapping up, right? Geez, these details. The truth is that MPE is still working today, 20 years beyond the inaccurate sell-by date.

HP was only successful in selling some of the first working models in 1974 after buying back all the failed 1972 units. And the development begun in 1968 was to create the Omega Project. The System/3000 was a fall-back effort when Omega, a 32-bit revolutionary design, was killed by HP in 1971. The vendor's short-circuit of a game-changer started a history that ran right up to the 2001 pull-out notice from 3000 futures. That one killed the rewrite of MPE/iX for the Itanium IA-64 chips.

The HP Phenomenon has priceless accuracy and strong details about the 3000's roots, starting on page 159 with MPE—Rx for Business. Dave Packard's quote that "we're wasn't proud of the 3000" echoed the system's endgame at HP. It's a thankless task to stay current when the vendor relentlessly withholds funds for innovation. What is not noted in the history article is that the 3000 made HP a computer company with the biggest success it ever had by 1976. You read the HP Phenomenon to find that fact.

As is often the case, the coda written to FedTech's 3000 story is rushed to a total demise. The wrap-up misses the work that the system does today, asking instead, "Why Did The HP 3000 Die Off?" Reports of its total demise are Fake News, something on the mind of the current White House occupant.

The report Goldstein offers tends to focus on the 1980s. The drastic transition from a Democratic President to Republican, from Carter to Reagan, is echoed in the White House's latest migration. A Series 33 was replaced by PCs during the Reagan era. (Oops, the 3000 artwork in the article showed a Series 70, a system not shipped until 1986.) The White House left such servers behind a few years before PA-RISC technology, a groundbreaker, revived them in the 3000 and HP 9000 lines.

Goldstein figured in his article that technology gaps killed the HP 3000. "The HP 3000, like many minicomputers of its era, was eventually supplanted by newer, faster and more capable machines, and by the widespread adoption of PCs in the late 1980s and early ’90s." Those elements did erase some opportunities. Hewlett-Packard played the biggest role in putting the 3000 to the sidelines, however, when the vendor's saleforce preferred Unix over MPE and then finally figured out how to sell PCs after stumbles like the HP Integra and the 3000's shadow terminal, the HP 150. HP really didn't know how to sell a 3000 for the first five years of the computer's life, either.

“President Ronald Reagan’s staff expanded the uses of computer office technology,” according to the Historical Association. Goldstein says the White House "soon adopted word processors with the advent of PCs in the 1980s." The Series 33 probably had HP Word, perhaps HP Deskmanager. A White House computer system is now a relic, no matter what system you choose.

Reagan had the Carter administration’s Xerox Alto removed from the Oval Office after he was elected, according to the Computer History Museum. No president since Carter has had a dedicated computer in the Oval Office, according to Slate. Looking to digital innovation, former President Barack Obama adopted "a fleet of computer-equipped staffers sitting directly outside his office doors," according to Slate. "President Bush sometimes used the computers of these personal aides to check news reports or sports scores. (He also had a personal computer at his Crawford ranch, which he used for limited personal surfing.)

The current White House is now operating under a new executive order which "could be tied back to budgeting for IT modernization, since agencies will need to decrease their security risk by investing in new technologies," said another FedTech report. IT directors like White House CIO Margie Graves have been told to update things to make them safer. The capital costs of that change, plus the operating expense of revising programs and training, would struggle to get past the DC cost-cutting of today. That's an example of history repeating itself.

Screen Shot 2017-05-20 at 5.18.51 PMThere's one powerful link to the 3000's history in the FedTech article. Computerworld reviewed minicomputers from IBM (AS/400) and Data General against the 3000 in 1992. This was the year when HP shifted its allegiance to Unix for business customers. At the time the 3000 beat both IBM and DG servers. The shortcomings of the 3000 ran well beyond the tech HP didn't pay to improve. The vendor's sales intentions kept the 3000 from holding its term in offices.

01:27 PM in History | Permalink | Comments (0)

May 17, 2017

Beyond emulations might lie migrations

Crm-data-migration-steps-cloudAs another webinar demo unspooled today for HP 3000 data migration products, the strategy of hold on or move onward demoed another facet. A 3000 might be a candidate for de-commissioning simply because the system has been too successful in the past. The next server will be different, but there's no guarantee the replacement will be better in significant aspects. Waiting for something better is not as easy as moving to something different.

Take COBOL compilers, for example. At the investment firm Fayez Sarofim, the HP 3000 was being evaluated for replacement. One element of the eval was finding a COBOL compiler compatible with the code running on the 3000. The company had to choose a way forward that was mostly different. Better was another phase.

"We chose AcuCOBOL over Micro Focus at the time of our migration because AcuCOBOL better handled the packed HP Floating Point without losing significant decimals," said George Willis. "It also had a more powerful set of debugging tools that were easy to use." Protecting decimal data was the priority. Getting a superior debugger was the improvement.

Time moved onward for the Sarofim strategy though, shifting away from apps and toward software services (SaaS). HP's Unix systems—an HP-UX 4466 Rx using AMXW, Cognos, Micro Focus COBOL, Suprtool and Warehouse—eventually got the boot, just like a 3000 did. The shift to services erased a department at the company. There's no emulation that can oppose that kind of sea change in strategy: "We don't even need our own servers, we'll access an app instead."

While making its move to HP's Unix, Willis said "We did not want to go through another riskier migration until we were stabilized. We are certainly stable now, but the firm has decided to move a different direction." So onward it went to SaaS. Emulation never got a fighting chance.

There are other places where emulation gets its shot. Once in awhile it comes up short, even after yeoman work to fit the performance needs. Veritiv Corp. runs four of the largest HP 3000s, N-Class servers loaded with RAM and HP's fastest PA-RISC processors. This profile of user needs to believe that emulation is a good long-term goal. Hardware for this top-end N-Class level emulation must be specified with an eye to a long-term play. Value for money gets amplified when you're replacing HP's 3000 hardware for a long run on an emulator.  "We don't need to ignore the issue of hardware," said Randy Stanfield this week, while investigating migration partners. "We need to put together a better long term plan than staying on the HP 3000 for more than 10 years."

That's searching for something different, that talk of needing to be off the system before MPE/iX stops date-keeping in 2028. Ten-plus years is a long time, enough to enable the magic of making CALENDAR work in 2028 a reality, perhaps. It's not impossible, although someone has to do the work to salvage MPE's date capability for 2028 use.

The silver lining for the 3000 community in any migration story is that the business often goes to a vendor who's been in the market a long time. MB Foster is one player like that, demonstrating its roots with a demo like today's of UDACentral. MB Foster celebrated its 40th anniversary in the 3000 market this week. 'The HP3000 market is our home market," Birket Foster said in an email today, "and we are grateful for the support, suggestions and collaboration with us."

Stanfield is looking for customer stories about migrating with Fresche Legacy, which earned its 3000 reputation as Speedware.

He had specific questions and would appreciate an email in reply.

1. What system did you convert to (Unix/Windows/Linux)?
2. What system did you convert from(HP3000 A-class/N-class?) and how busy was the system? Number of users?
3. Are you still running that system?
4. Did you convert to using the Eloquence DB?
5. Performance after conversion: good or bad?
6. Any Do's or Don't's?
7. Primary Code base (Speedware/Powerhouse/Cobol/Fortran)? Amount of code converted?

05:45 PM in Homesteading, Migration | Permalink | Comments (0)

May 15, 2017

3000 Cloud Doings: Are, Might, and Never

Flight-simulator-cloudsThe latest news about cloud computing for HP 3000s came from Stromasys. The company selling the Charon virtualizer (many think of it as an emulator) announced a new bundled offer as well as announcing that any public cloud can run Charon. Sites that employ the Oracle Cloud to host their virtualization systems get un-metered cloud services as part of that deal with Stromasys. Oracle Cloud is one of the newer players in the cloud market. There's no place to go but up in market share for Oracle Cloud, carving out its business among providers dominated by Amazon's Web Services.

Emulating HP 3000 servers, however, is a job that's not often suited for a shared Intel-based server. There are exceptions, like light-duty 3000s or those in archival mode. Those are the best profiles for 3000s in the cloud running Charon, according to the Stromasys HP Product Manager Doug Smith. 3000 A-Class systems — Stromasys calls this Charon model the A520 — can be run from the cloud.

Many of the cloud's typical servers make memory and CPUs available on an as-needed basis, swapping processor power and RAM in and out. This is in contrast to dedicating a highly-threaded CPU and all available memory to a task like emulation. "Charon requires dedicated resources," Smith said. "If I say I need a 3.5 GHz CPU response, then I need that 3.5 GHz in the host itself, not being shared among other virtual machines."

It's safe to say there are 3000s in our community that are good candidates for a cloud profile. A-Class systems running the one last MPE application, some app still critical to a datacenter, for example. Better to have this sort of foolproof hardware service chain using virtualization, instead of stocking redundant 3000 memory sticks. (The better option to stay with the 3000 hardware from HP is an independent support company.)

The cloud — a term that doesn't have much traction for classic 3000 pros like Smith — might evolve to the point where dedicated CPU performance at any level could become affordable. Not even Hewlett-Packard knew how to price and sell its HP Cloud so its Unix customers could host datacenters in the cloud. Integrity chips were the next generation of PA-RISC, so emulating any chipset with that pedigree is no small matter. Smith, like any other analyst in IT, considers dedicated performance from a public cloud as cost-prohibitive.

Never-say-neverAny company can arrange to use an offsite, networked host for MPE/iX apps. This seems more like timesharing to the 3000 pros than Infrastructure as a Service. Cloud computing is supposed to reduce costs, and it does so by sharing resources. Sharing is not a great match for emulation at multiple levels. When you use a VMware host to create the Linux cradle on one level, which then virtualizes PA-RISC with Charon, that's a more intense CPU requirement than public clouds can handle. Pull out the VMware and you're fine for a smaller datacenter.

Cloud computing users definitely are shifting their expenditures from capital expenses to operating expenses. OpEx can be easier to place in a budget than CapEx, especially for legacy systems like the 3000. We'll never see a day when there's no more CapEx spending in datacenters like those in the 3000 world. OpEx is on the rise, but like the Paperless Office of the 1980s strategies, CapEx will always have some benefits. One is the constantly dropping cost of HP's hardware, if you can arrange for enough backup components and parts.

OpEx, however, gives vendors and customers a way to tune up a services agreement. At Stromasys, for example, the Oracle Cloud already has advantages for some Charon users. "There is definitely, for example, an added benefit for [Sun Sparc] SSP users," said Marketing Director Sarah Hoysa. "By emulating their SPARC instances on Oracle Cloud, they have an additional way of continuing their close relationship with Oracle."

"The big thing is that customers now have a lot of choice," Hoysa added. "We know people are moving to a wide range of public clouds. We're making our solution on all of those public clouds." Dave Clements of Stromasys said the company's got an insurance firm running Charon in the Microsoft's Azure cloud, for example. It's not a 3000 site. The cloud is all potential for 3000s today.

There are all the software and license arrangements needed to put a 3000 onto any of those public clouds using Charon. Stromasys went to a software-based licensing arrangement two years ago, so the need for a USB stick with HPSUSAN data has been swept aside. The 3000 customers using N-Class systems might have an interest in cloud computing in the future. For now, however, Smith said the security, control, and command of on-premise hardware is preferred by larger manufacturers. The interest has been from smaller manufacturing companies.

It's safe to say—given the competition for customers among a growing rank of cloud companies, we will never see a future with zero HP 3000 cloud computers. It's coming, and companies like Oracle will drive down pricing in ways we've never imagined. The 3000 datacenters will hang on long enough to see that day, because you can never say never when it comes to failures of hardware that's 14 years old and aging.

 

07:06 PM in Homesteading, Migration | Permalink | Comments (0)

May 12, 2017

How to Step Through a CSLT Reinstall

Stepping-stonesAssume you've seen your Series 918 crash to its bones. You replace the 3000 and need to reinstall from CSLT. Many a 3000 site hasn't done this in a long time. This is when you're getting in touch with your independent support company for advice and walking the steps to recovery.

Oh. You don't have a support company to call. One more thing that's been dropped from the budget. You could ask on the 3000 newsgroup for help, so long as your downtime isn't serious. This support strategy is one way to go, and James Byrne got lucky this week. One expert walked him through the critical steps.

"I am working my way through the check-lists for reinstalling from a CSLT," Byrne said, "and I have come to the conclusion this stuff was written more to obscure than to illuminate." The HP documentation advised him to boot from disk, something he couldn't do. "Fortunately this is a backup 3000, and nothing too bad can happen yet."

Gilles Schipper, our esteemed homesteading contributor, provided the answers. The problem lay in a bad boot drive, but how do you discover that's true? We'll get to that in a bit. First, the CSLT reinstall.

"Assuming you have no user volume sets:

1. Mount CSLT in appropriate drive and boot from alternate path
2. From prompt, type INSTALL
3. At completion, boot from primary path and START NORECOVERY (this is your second "boot".)
4. Mount whichever tape contains your directory (could be the CSLT or your latest full backup - if directory on both, use whichever is more current)
5. Log on as MANAGER.SYS and restore directory (:file t;dev=?;restore *t;;directory - note 2 consecutive;)
6. Use VOLUTIL to add additional discs to MPEXL_SYSTEM_VOLUME_SET.
7. Mount backup tape and:
:restore *t;/;keep;create;olddate;partdb

Reboot with START NORECOVERY

Ah, but what to do if you have user volume sets?

"So," Schipper said, "with user volume sets:

Modify step 6 to read as follows:
6. Use VOLUTIL to add additional discs to MPEXL_SYSTEM_VOLUME_SET, as well as any additional user volumesets that you have.
The approprite VOLUTIL commands you will need are:
1. NEWVOL command to add additional volumes to the existing MPEXL_SYSTEM_VOLUME_SET
2. NEWSET command to add first (master) volume of user-volumeset1, say
3. NEWVOL command to add additional volumes to user-volumeset1
4. repeat 2 and 3 for each additional user volumeset

Then, prior to step 7, add step 6A, which is the repeat of step 5, which is to AGAIN restore the directories from the tape containing the directories so that they are restored to the master volumes of their respective user volumesets.

And then step 7 remains as before.

The bad LDEV1 discovery? You can always swap out a known good drive. Byrne did this and noted he had to restock his cache of backup drives. "Only three left," he said. "Time to order more, perhaps." Advice on checking for bad drives before a boot:

"Instead of INSTALL, type ODE," Schipper said. "Then run Mapper. That should show you all of your devices including hard drives. Although that may even show drives that cannot be INSTALL'ed to. Also, check all of your SCSI terminators."

Stan Sieler added this to the remedies for a bad disk. "Try doing:

  start norecovery message

"That 'message' option will cause 'start' to generate a **LOT** of output. Hopefully, the last few dozen lines might provide a clue to the problem."

Do all the backup that you need when protecting that HP 3000. The success of its backups falls in the lap of the system's managers. (Running CHECKSLT from the TELESUP account will verify if a CSLT is still good enough to boot your system. You'll want to check that CSLT to ensure it'll run on any tape drive, not just the one it usually runs on. Alignment issues kick DDS drives out of service regularly.) Don't forget about keeping your CSLT healthy.

 

 

03:53 PM in Hidden Value, Homesteading | Permalink | Comments (0)

May 10, 2017

Wayback Wed: 3000s Needed More Time

In this era of cloud computing, the roots of the original HP 3000s rise up. Clouds are the ultimate shared computers, systems so fluid they use hardware that can be provisioned with a set of entries on a webpage. Forty-five years ago this month the first computer that created our community wasn't making its way to its first loading dock. HP called this system a server for multi-programming, designed with the full intention of enabling people to use it from remote locations. The product couldn't bridge the miles between California and Connecticut, unable to ship from the HP factory location to a customer facility on time. It was the beginning of a black eye the vendor wore for nearly two years.

First-HP-3000-Sale-DelayHalting starts have been in many a successful product's history. In May of 1972 the HP 3000 was already running late, beset with hardware problems. The archives in the NewsWire offices include a letter to the first customer to order an HP 3000. The initial shipments of HP 3000s only fulfilled Bill Hewlett and Dave Packard's doubts about being in the commercial computing business. Their H-P was stuck with a product which started as a disaster. It was up to another Bill to break the news in the letter (click for details).

HP put its best face on this first delay, telling Yale-New Haven Hospital that "When a first order comes from a hospital such as Yale-New Haven and from [Dr. David Seligson] a person with an international reputation in the field of laboratory automation, we are doubly flattered." But this HP 3000 system was going to ship late to New Haven.

"Although our development is remarkably close to the targets we set over a year ago, we find that we must slip our shipments to insure that our customers receive a computer system with the built-in reliability that HP is known for," read Bill Terry's letter to Seligson. "Your system will be the first shipped outside the immediate Cupertino area and is scheduled for December, 1972."

The letter arrived in May, seven months before HP would finally allow the first 3000s outside of California. It was a simpler time with crude technology. HP offered the hospital a bonus for enduring the delay. "We would like to donate an additional 8K words of core memory (part 3006A, $8,000.00) to your HP 3000 system. Additionally, our intention is definitely to continue with plans for the training of your people, both in Cupertino and New Haven, as soon as possible." The 3000 entered the world as an ASAP project.

Things didn't get much better once the computer finally arrived at customer sites. MPE crashed, the "golden saddle on a jackass," as one account put it. Eventually the 3000 was withdrawn from the market and HP proceeded to buy back all of them. In late 1974, the server poked its head above the surface of market waters.

Even with that very first order of the HP 3000, the vendor was delivering its product by way of "intention" rather than guarantees. HP's founders had made a fortune with a practice of under-promising and then over-delivering by 1972. Conservative to its core, the company nonetheless would ship a system so crippled it had to be returned for a do-over, two years later.

And those 8K words of memory, at a cost of $8,000, are so small today that 125,000 of them are available for almost free. Not core memory, specific to only one computer, but a 1GB memory stick can be used in 100 million computers, and millions more cameras, printers and phones. A small contribution indeed, HP offered, in the face of a delay. It was significant for the time, though.

Bill Terry was doing his best with what HP had for the nascent 3000 community. He would survive the debacle of the first HP 3000 models to see himself and other HP computer founders honored with a documentary film, screened when HP's restored Palo Alto garage reopened.

07:10 PM in History | Permalink | Comments (0)

May 08, 2017

Turnkey cloud services mirror 3000 roots

Tree rootsRe-hosting HP 3000 applications keeps getting less complex. Managers who homestead once had to locate an HP 3000 which had similar specs, one that could preserve their application license levels. Because it was an HP box, their auditors wanted them to execute an MPE license transfer with HP. More paperwork, like finding a 20-year-old bill of sale.

When Stromasys virtualization showed up, the mandate for specialized MPE Hewlett-Packard servers fell out of the equation. Configuration was still required, though, sometimes because the Intel servers in the strategy were already otherwise engaged in a customer’s datacenter.

In each of those formulas, companies continue their local management of hardware, storage and networking. It’s an On Premises choice, called On Prem in some planning sessions. Cloud computing is the opposite of On Prem. It is changing the need for hardware in the datacenter. The strategy now has an official role in virtualized 3000 practices.

Stromasys recently introduced Oracle’s cloud services as an option for Charon HPA, the emulator that transforms an Intel server into a PA-RISC system. The company also issued a notice that Charon works with all other cloud computing options.

Charon has been ready for the cloud for several years. The new element is a packaged set of software, support services, and provisioned cloud computing without a meter for usage. Director of HP 3000 Business Development Doug Smith at Stromasys said the cloud equation best fits archival installations and smaller, A-Class-grade production shops.

Extending offsite strategies

Stromasys completed testing for its cloud-based Charon in November of 2014. During the next year the company took an agnostic approach to cloud hosting. Whatever service a customer preferred was worked into Charon installations. Stromasys would also recommend a provider.

The newest formula is an end-to-end bundle that includes unlimited cloud capacity, making it more like a turnkey, locally-provisioned MPE/iX host. HP 3000 customers liked turnkey datacenters. Find a customer from the 1980s and the 1990s and you’ll uncover a manager who wanted more homogenous computing. Virtualization in the cloud is a way to release On Prem operations to trusted third parties.

The unmetered option for Charon on the cloud comes in the Oracle Cloud offering. The more dug-in suppliers of cloud Infrastructure as a Service require customers to manage separate billing, usage rate metering and support. A company looking to migrate off HP’s 3000 hardware might have existing relationships with Amazon Web Services, or Azure or Datapipe or Rackspace. All of these will work with Charon HPA.

The novelty of Oracle Cloud plus Charon is its turnkey nature. Turnkey has a resonance with the 3000 customer, especially those who found that heterogeneous IT had downsides. The required amount of management increased as IT got more heterogeneous.

Most 3000 sites have lean management resources. Getting a single-call arrangement for legacy apps and hosting, while being able to leave On Prem behind, should interest some sites. Especially those who might be finding that even a DR-status 3000, powered down for many years while it keeps the MPE apps available, is a strategy with measurable risks.
 

03:18 PM in Homesteading | Permalink | Comments (0)

May 05, 2017

Newer 3000s come at low cost vs. downtime

A-Class vs. Series 900 performanceEarlier this week a 3000 site worked through a system halt by replacing Series 918 memory sticks. Ultimately the problem was resolved with newer memory, but a full replacement of the server might have been just as easily obtained. The relative performance of 3000s sold across the years becomes a factor in this sort of support equation. A manager might find themselves poring over a chart like the one at left (click for details of the A-Class comparisons to older 3000s.)

HP was never much motivated to benchmark its 3000 line against the rest of the world's business servers. However, customers about to upgrade were able to define the relationship between the boxes in the lineup. HP used a 1.3 HP 3000 Performance Unit rating for its Series 42 HP 3000 classics, the pre-RISC range of models before MPE/XL. Most customers consider a Series 917 to be the bottom of the PA-RISC line, and that server earns a Performance Unit score of 10.

The tables and rating sheets were printed by HP until 2004. For example, here's one of the last, a page from the HP e3000 Business Servers Configuration Guide. The ultimate generation A-Class servers start at a 17 and make their way up to an 84. The N-Class computers start at 100 and build up to 768. In general, an A-Class with a 200 at its model tail end will be faster than what is being replaced from anything that's not an A- or N-Class.

Even in 2017, these numbers can matter. A Series 918 will be approaching age 23 by now, first manufactured in 1994. An A-Class server is at most 15 years old. Replacing  9x8s with A-Class servers can be a way to delay replacing HP's 3000 gear altogether. Rejuvenation like this is not a long-term solution, but a manager might be in between the rock of aging iron and the hard place of frozen software licenses.

Used A-Class or N-Class servers can add reliability for customers who must homestead. Even the ones selling for $2,000 are a lower cost to avoid downtime than a search for 25-year-old memory components.

The numbers in HP's Relative Performance Chart are more than just a reflection of the model numbers. A Series 928, for example, is less powerful than a Series 918. HP explained the methodology. "Performance Relative to HP e3000 Performance Unit" is a general guideline for OLTP performance, since the factors influencing an application's performance vary widely."

AICS Research, the software company that developed the QueryCalc 3000 app and the QCTerm freeware terminal emulator, created the most direct comparison tool back in the middle of the 2000s. Even though AICS has closed its website, the data lives on across the 3000 universe's webpages. 3k Ranger owner Keven Miller posted a link to his copy of the AICS data.

What HP's 3000 performance rating means can be debated. At HP World in 2002, HP announced the final new 3000 systems, all based upon the PA-8700 processors. At the high end HP announced a new N-Class system based upon the 750 MHz PA-8700 processor. The new N4000-400-750 was the first HP e3000 to achieve an MPE/iX Relative Performance Units (MRPU) rating of 100; the Series 918 has an MRPU of 1.

HP contends that the MRPU is the only valid way to measure the relative performance of MPE systems. In particular, they maintain that the MHz rating is not a valid measure of relative performance, though they continued to use virtual MHz numbers for systems with software-crippled processors. For example, there are no 380 MHz or 500 MHz PA-RISC processors. Unfortunately, the MRPU does not allow for the comparison of the HP e3000 with other HP systems, even the HP 9000s and Integrity servers.

HP has changed the way it rates systems three times over the life of the HP 3000. During the middle years, the Series 918 was the standard with a rating of 1. In 1998, HP devised a new measurement standard for the systems it was introducing that no longer had the Series 918 at 1. It is under this new system that the N4000-400-750 is rated at 100. Applying a correction factor, AICS Research has rated the N4000-400-750 at 76.8 relative to the Series 918’s rating of 1.

10:57 PM in Homesteading | Permalink | Comments (0)

May 03, 2017

More than ever, old sticks trigger backups

Memory-cageRegular and frequent backups still hold their spot as keystones in a stable HP 3000 datacenter. The backups are even more essential this year. 2017 is the 14th year and counting since any HP 3000 components have been manufactured. Excepting some third party disk solutions, the average age of Hewlett-Packard's MPE/iX servers has more than doubled since HP stopped building the boxes in 2003.

In 2003 a manager might be daring enough to run a shop with a server built in 1991, the first year the 9x7 servers were manufactured. Systems in the first wave of PA-RISC design were still in service, but a Series 950 was a rare box by the time HP stopped building them. That oldest 3000 server at the time was still only 15 years old. That made the average age of a 3000 about 8-10 years.

Add 14 years to that lifespan and it's easy to locate a 3000 and its components which are more than 16 years old. The Series 9x8 systems turn 25 this year. The numbers came up in a recent emergency repair discussion out in public. A Series 918 at Harte & Lyne ground to a halt with a bad memory component, and even a pair of replacement sticks were duds on this 23-year-old system. The 918-928 servers are still among the most frequently used servers in the community.

The manager at Harte & Lyne keeps this hardware high-wire act going because of Powerhouse licensing problems. The repairs of his 918 coincided with very recent backups, but it could have been up to 30 days behind. Loss of company data was well within reach at this logistics company. The backups prevented the calamity that's now started to hit spare parts, too.

Two failed memory sticks, plus two replacement failures, triggered the system halt. The problem was a defective card.

"I am doing a full backup just in case," said James Byrne, "but as it happens we did our monthly CSLT and SYSINFO jobs yesterday. Once the new backup is complete I will add two more sticks and see if they work, and repeat until we encounter the problem of getting the memory cage filled."

Byrne has access to independent support for his 918, but used the 3000-L to diagnose the problem at first. "One of the two replacement cards I initially used was also defective," he reported after he got the 3000 back online. "It took some time for our support people from Commerx and myself to finally realize what was going on, as we pulled cards and swapped locations trying to get a full bank of memory reinstalled."

Fortunately I still have about six two-card sets kept in a vault as spares, together with sundry other spare parts. No doubt there are other duds lurking therein but now I am alive to the possibility. Thanks for all the help.

And yes, the reason for all this dancing around to keep a 23 year-old piece of hardware running is Powerhouse and the intransigent present owners of the 'Intellectual Property Rights' thereto. To describe these people and their representatives as unethical grants them too much dignity for what they are. And what they are I shall leave unsaid.

These aging 3000 servers sometimes have been retired out of production duty, serving out their remaining days in archival mode. But even archives must be updated from time to time. Production data demands a more serious backup regimen, one getting more serious every year.

Solutions to avoid this risk run a wide gamut. Migration onto other environments reduces the age of replacement parts. A stout support contract for MPE/iX hardware ensures component replacement. Good backups are still crucial there. Beyond these remedies, there's one way to get newer hardware without migrating. PA-RISC virtualization software doesn't make backups an optional task. It can make them less likely to be needed, though.

06:41 PM in Homesteading, User Reports | Permalink | Comments (0)

May 01, 2017

No obit for your OS, but not so for hardware

ObitThere's a new documentary in theaters about writing obituaries. The film Obit covers the work of obituary writers at the New York Times. No matter how you feel about mainstream media, the Times is one of just a handful of media outlets still telling stories about the lives of people who've just died. Obits were part of my reporting tasks when I broke into journalism, but I never wrote anything about someone famous while I worked at the Williamson County Sun. My very next job was Editor of the HP Chronicle, where people were famous within a modest community. I didn't write an obit at the Chronicle. I was eight years into the work at the Newswire when Danny Compton, a co-founder of ROC Software, died so very young at 40.

Compton and his wife Wendy were fun and important to reviving the Maestro user community. Why care about them now, you might ask, or even a legend like Fred White, whose life story I wrote up when he died in 2014? An obit is written for a death of something material, a person with a body, or an object which can be scrapped or destroyed. Something of value which made a mark on the world. The years continue to pull out HP's 3000's hardware from service. No one's making new gear with PA-RISC chips. Someday nearly everything that's an HP 3000 will go to dust.

The same cannot be said about MPE/iX. An obituary of the OS might be hard to prepare. The demise of HP's hardware could be written in advance. Advance obits, sitting up on the cloud by now, are a practice of those obit journalists. Obit writing tells about a life, not a death. Even as a celebrity continues to celebrate birthdays, their clock is ticking. Maybe that MPE/iX demise is in 2028. It's easier to see an obit arriving earlier for HP's hardware, though. It might read HP 3000, durable business server, dies at 51.

Hewlett-Packard's HP 3000, the first minicomputer which included a database wired into a file system, passed away on November 14, 2023 when the last CPU board failed to boot the server's operating system. The hardware, whose design was revised from the heyday of mainframes through the boom of the Web economy, carried commerce and data among entities as varied as aerospace makers and police departments. It is survived by the software written for the MPE/iX operating system as well as the database IMAGE/SQL.

HP's 3000 gear grew from a system that demanded raised flooring and specialized cooling to systems that could be carried under an arm. The hardware once read data from paper tapes and gained its ultimate IO abilities processing Internet data from standardized networks.

The obit for MPE/iX will be harder to write in advance. The OS is still going places.

The inability to know what day it is won't keep a painter from rendering visions, for example. In the same way a calendar will help that artist, some unforeseen code can give the OS a life that could outlast our own. The reason anyone would want to cut and test such code is virtualization.

Plenty of the surviving HP 3000 shops are in archival mode now. The low-processing needs of archival fit the glove of cloud computing. MPE/iX is a vessel to carry the ideas of application software. Like a ship's hull that can be cared for across generations, your OS has been a vessel that's gained a second generation with virtualization. Putting that virtualized hardware into the cloud lets the light-duty of archiving stay on course.

08:26 PM in Homesteading | Permalink | Comments (0)