A Firm Solution for 3000 Emulation

Willard West has spent a career completing circuits, and now he’s turning his company toward connecting the HP 3000 homesteader with a supply of emulated computing power. The founder of Strobe Data, along with his wife Vera, West popped up his hand in an OpenMPE session at HP World this summer and gave the HP 3000 faithful a new future to ponder about emulation of the computer. West spoke up after more than a year of discussion about software-based emulators, including an understanding on OpenMPE’s part that Strobe would be building only a software-based 3000 emulator. That software solution will also be in the company’s future, but the company will first develop from its most seasoned expertise: hardware. West said at the meeting that he wants his company to be the first to market with a hardware-based emulator card that will make a server-class PC pretend it’s an HP 3000. Strobe is working on a deal with HP to develop such a device, a product that will use the same processor now working at the heart of the newest generation of HP 3000s.

The idea of hardware-based emulation using the vendor’s chips might have been new to those in the 3000 community assembled at the OpenMPE meeting, but the concept has been long-proven at Strobe. The company was founded on its ability to emulate hardware vendors have stopped building: first Data General, then Digital and finally HP minicomputers, as the DG Nova and Eclipse, DEC PDP-11 and HP 1000 communities found new life for their applications, even after DG, DEC and HP stopped selling the systems.

Strobe has an existing relationship with HP on several fronts, offering both Digital system emulation and the HP 1000 real-time emulation card, the Kestrel. Since the HP 1000 community’s experience is being held up as a model for how HP might help the HP 3000 community, the entry of a company that has succeeded in extending the life of HP 1000 applications with an emulator looked like good news for 3000 homesteaders. We spoke with West a few weeks after that HP World conference, as he filled us in on what it’s been like to extend a computer’s life beyond the maker’s schedule, and how Strobe has overcome the challenges in such a task.

Why go to HP to get PA-RISC chips for an emulator?

It’s the easiest, quickest way to the market for a product. If we can get the chips, we can get into the market fairly quickly. We’ll have to write some software to emulate the peripherals using PC peripherals, but that’s almost standard operating procedure to us.

At the same time, we can get started writing an actual software emulation for PA-RISC.

Do you expect to be able to write that software without much MPE experience? Or is the PA-RISC IO bus so well documented you don’t feel like you need that experience?

That’s exactly right. The documentation is readily available. In general the IO bus is in the public domain, and if not, it certainly exists somewhere.

Have you used HP’s chips to create your other emulation products?

No, but we did use the Fairchild F9445, a 40-pin single chip implementation of the Data General Nova architecture that could glow red-hot. The original Falcon card did not have an IO bus, and we enhanced the structure of the design so we could do the DG Eclipse instruction set.

Have you ever had a chip discontinued on you? Nobody’s sure how long HP will be offering PA-RISC.

When NSC acquired Fairchild, that F9445 chip was discontinued. We designed and successfully fabbed, in cooperation with IC Designs, our own 1.25-micron CMOS version of the F9445, the ICD9445, yielding almost twice the performance of the original.

Do you still pursue your own chip designs, or are there broader sources to buy chips from? We’ve heard the PA-RISC processors are available from Hitachi.

We have considered going down that path. What we do today with Osprey, our PDP-11 emulator, and the HP 1000 emulator, are both implemented on the exact same hardware. It’s a PCI card that has a couple of Xylinx programmable gate arrays with 4Mb of main memory.

Do you often have outside certification of your products’ ability to emulate the original hardware?

What we’re going to be asking HP for is the hardware diagnostics. With that, we’ll certify the emulation.

How can you protect and provide for HPSUSAN and HPCPUNAME data that third party software companies need for their copy protection? Does this require HP to give you access to its SS-CONFIG software, or can you engineer something that protects the software vendors but leaves HP out of the loop?

First of all, this will not be an expanding market. Where we run into a situation where we had to protect software vendors, we’ve done a special version of an emulator that was unique to them. That’s one possibility. The other possibility is that every Osprey emulator goes out of here with an embedded serial number. We can license our add-on software by reading that serial number, and we can make that capability available to the third parties. In the past, if the third party software supplier can tell us what [number] they are looking for, we can emulate that conditional on finding a specific serial number. There has to be a payback for that kind of custom engineering. Modifying their software so they can read our serial number would be free.

Have you seen any reservation from HP so far about enabling customers to buy an emulator card built upon HP’s PA-RISC chips?

I was pleased that HP was standing firm on not making any concessions to sell hardware after Oct. 31. What HP has told me is that after October they will not sell the 3000s, and they will not allow HP 9000s to be converted to HP 3000s. So far I have not heard even the tiniest squeak of concern from HP over us helping customers to keep MPE applications.

On Strobe’s side of this, doing the product with the PA-RISC chip means the product will go away when the PA-RISC chip goes away. We fully intend to do a software emulation, but using the chip is the quickest way to the market. From everything I’ve heard from HP, they are earnestly working to satisfy the customers who are going to stay. It’s my impression they are aware that if a product comes to market, then more customers will stay than will migrate. But so be it is their opinion, I guess.

Are you talking with the HP people who are only in charge of PA-RISC chips, or the HP officials in the HP 3000 business as well?

Our meetings with HP have been pretty much preliminary so far. We had a meeting in Cupertino, and I was able to meet with Dave Wilde, Ross McDonald and Mike Paivinen in Atlanta. At the moment we’re working through them to get access to PA-RISC. I wouldn’t be tempted to go around them. We already have some history with HP on the HP 1000 and with Digital, now part of HP.

HP has asked that the emulation run on HP hardware. I asked how do we assure that. They came up blank on that, and then I asked how do I enforce it? They agreed that I should put on my best face and say it should be on HP server-class hardware. We came back and decided there would be a rebate to the user if they could show it runs on HP hardware. We propose that HP tell us what class or model of hardware, and we’ll negotiate if they get too unreasonable. But we want server-class hardware with ECC memory. We don’t want to run the product on a cheap PC. We’ve had VARs and OEMs make that mistake. We’ll come up with a rebate scheme, more of a good faith kind of thing.

Has Strobe done software-based emulation before?

Yes and no. We’ve done software-based emulation for the Data General Eclipse and Nova series, which went absolutely nowhere. We wrote the emulation in 1996, and it had about half the performance of the hardware solution. We recently fired up that software product on a Pentium 4, and now it outruns our hardware quite substantially.

Isn’t this the promise of software-based emulation? That sooner or later, a fast enough processor will emerge that can provide performance as fast as the original, native hardware?

Yes, sooner or later. The dilemma is: can your legacy environment limp along until that level of processor appears? We’re currently doing a PDP-11 software emulation, and I want a software emulation that lets you unplug the hardware solution. The team is in the process of marrying the new PDP-11 emulation to the Windows NT peripherals virtualization.

Do you think you need some outside assistance from Allegro Consultants on the MPE aspects of your product?

We’ve assumed that all along. How much is going to be the question, and we’ve had several discussions about what level of cooperation and how to implement it.

Jon Backus of OpenMPE has expressed concern that several emulation products would pose a prospect of failure for all of them. Is competition from other emulator makers a risk?

With our Data General products we had a couple of competing products, and we trounced them. In the PDP world, there’s a competitor in SRI, the same company that has announced it’s interested in creating an HP 3000 emulator.

Digital decided to do us in with chip availability years ago. Then they came to the realization that it wouldn’t hurt to have something on the horizon to keep people in that PDP-11 world. So Digital was the originator of SRI’s PDP product, and now SRI has gone off and done a VAX emulation, also.

In SRI’s case we exist in two different worlds. We have two keys to our success with our emulators. One is to be able to throttle our product so it doesn’t break the real time timing loops. Two, supporting the actual IO bus.

Since you’ve created an HP 1000 emulator, do you have to support a new bus to create an HP 3000 emulator?

No, it’s only that this IO bus will be in a commercial environment.
What advantage will you provide over other emulation projects that have been announced for the HP 3000 community?

SRI has existed in the commercial segment of the market, and we’ve existed in real-time process and control. This is our first venture into the commercial side of the market. We expect to be successful, and hope to be able to put some kind of deal together with Allegro that helps them help us, without them going into competition with us.

How fast do you believe the HP 3000 emulator market will take off after Oct. 31? Can you arrive too early, while there’s too much 3000 used hardware to compete against?

At the time we hit the market with our PDP emulator, Digital was still selling PDP-11s. We’re clearly not going to interfere with HP in that respect. At best we’re a year away from a working prototype.

Our major competition has been the used market. We’ve out-survived that. Yes, we often lose sales opportunities to people buying in the PDP-11 world. When people take those offline, they’re often worthless. Eventually you will need used equipment fixed or repaired, and we expect to outlive that situation.

When will you need to have operating environment experts working with Strobe to do a hardware-based emulation?

Our first benchmark will be to emulate the hardware religiously. With any luck, then the software will just come up and run. We’ve done that again and again.