Dave Wilde
R&D Section Manager
Growth and Datacenters
Commercial Systems Division

Dave Wilde is making decisions at the crossroad for the HP 3000. As the section manager of the 3000 R&D labs who’s managing growth projects, Wilde is overseeing some of the most forward-looking work for the system. His teams are making the transition from today’s PA-RISC processors toward the systems of the next decade, computers that will be ready to accept IA-64 designs.

Wilde works at this turning point for the 3000’s next generation using a combination of software experience, hardware sensibility, and a little good fortune. Like his manager R&D chief Winston Prather, Wilde first encountered the system as a high school student in the 70s — in Wilde’s case, working at Crawford’s Department Store in Chicago doing data entry on the system. He took his bachelor’s degree in Computer Science west to California in 1984 and did his first two years at HP in the Test and Measurement Organization, where his EE college work was helpful on a project to create a VLSI IC tester. He made the transition to software as the R&D manager of the Allbase core and interoperability project. And eight years ago he entered the HP 3000 division, when Allbase made the move into the CSY labs with the more pervasive HP database, IMAGE. Project management and more R&D with some of HP’s biggest 3000 accounts led him to his most recent appointment.

Eighteen months ago Wilde was given divisional R&D section manager duties when the 3000 labs started to grow again, the kind of job that involves helping very bright engineers stay focused on creating the next generation of HP 3000. Hearing him describe his work is to listen to the sound of man who likes making decisions, and you can feel a willingness to take chances on change. Wilde tells of story of how he arrived at the California part of HP because of his alma mater’s good fortune in a football game.“I wound up here because of a football play,” he said of his work in HP’s California labs. Illinois had a highly-ranked football team in Wilde’s senior year, and he described a key play in a showdown game to determine a bowl berth.

On fourth down the other school’s running back battled toward the scrimmage line. “I remember their back sweeping toward the near side where I was sitting,” he says. “We had an All-American safety who came out of the backfield to nail the runner. We took the ball, returned it for a touchdown and won the game — and wound up going to the Rose Bowl. I came out for the Rose Bowl, and didn’t have California as one of the places I wanted to live. If I hadn’t fallen in love with California and Lake Tahoe on that trip, I probably would have been routed to Colorado somewhere.” The affable manager’s presence in Cupertino’s 3000 labs is a good turn for the platform as well. “You make some decisions that are pivotal, but hope that it usually doesn’t turn on a football play,” he jokes.

Wilde’s ability to identify details in such turning points will serve him and his team well in the months to come, as they make decisions that turn the HP 3000 into its third generation. After seeing him share previews of coming technologies at this spring’s IPROF conference, we asked him for more details on the system’s upcoming turning points.

At IPROF you mentioned active projects in CSY to meet up with the rest of HP’s platforms. Will that mean other HP platforms will pilot the implementation of those designs?

In the PA-8500 systems, HP Unix will ship earlier. There’s a few parts of what we’re doing to help bring that out, like the performance scaling work. The biggest piece of work is to move our IO system to a PCI environment. This is a very large, one-time change that’s needed for moving to the new platforms.

As we re-architect from NIO to PCI, we are for the first time trying to leverage some of the Unix IO driver technology. If it goes well, we may be able to do more of this in the future in other areas. Down the road we have high hopes for standardized IO. We might be able to leverage what we’re learning here to future projects such as [bringing the 3000] Fibre Channel.

How do you standardize IO?

IO stacks on Unix and MPE are different, in terms of the way they are modularized. By having well-defined layering, the way you might find networking in the OSI model, there’s a lot of effort in I2O to standardize those — so you could have more off the shelf logic that you can just plug into your IO systems.

You talked about drafting off HP’s investments in peripherals. How can that help the 3000 get some of its shortcomings in those areas resolved, like DLT library support?

The more we align with HP’s overall direction, the better off we are, and the better off HP’s customers are. We can deliver a lot more to our customers faster, better and cheaper the more we can leverage. It applies to everything, from hardware to middleware — and even to manufacturing and marketing.

Sometimes there are trade-offs to consider. Some of the MPE products add value beyond what’s available in other platforms, and our customers have gotten used to that value, like networking stacks and features we’ve built in for reliability. If we went to a more standardized architecture customers might lose something. There’s a trade off between standardization, time to market and cost with functionality. We always have to look at those tradeoffs.

What impact are PA-RISC 2.0 versions of compilers going to have on the platform’s growth plans? While last year’s growth plans pointed to these versions as bringing more performance, this spring HP said IA-64 versions were now the primary goal.

Those charts reflect our plans and expectations for where the increases will be coming from. As we actually prototype and build the different parts of our roadmap, we’re measuring the results versus the expectation. In the case of using the PA 2.0 back-end for our compilers, we found the return on investment is pretty small. We currently do not plan to make that specific change.

The benefits for some of the other performance work — like in the area of changes in the memory manager’s MMP disable enhancement, and changes to the dispatcher, and changes in how we use the Translation Lookaside Buffer (TLB) — actually have a much higher return.

Do you believe the bulk of the growth work is being done for the 3000 installed base as a primary objective? How does something like memory capacity expansion to 16Gb help a shop with a modest-sized 3000?

Most of the work we’re doing in the growth area has to do with performance and scalability and capacity. That work directly benefits customers at the high end. That’s true of installed base customers and new customers, platform buyers and solution buyers.

Indirectly it benefits all customers very profoundly. It guarantees a scalable platform, which results in more sales, and that in turn results in more investment we can make, which benefits everybody. Those enhancements are necessary to make some of the investments in the new platforms pay off. A lot of the enhancements we’re making right now are removing bottlenecks that have started to creep in. Those changes are necessary to see any benefit from the new processors.

How will variable sized pages help improve HP 3000 performance?

They don’t necessarily, by themselves. It turns out we had an issue we needed to deal with: over time the TLB size on the chip is getting smaller, and so we needed to change the way we used the TLB so we don’t run into performance bottlenecks there later.

We needed to be able to keep the same amount or more information in the TLB, used to keep track of frequently used and recently used pages. The changes we’ve made are allowing us to make effective use of the reduced TLB space.

Are those redesigned page sizes going to mesh into the architecture for IA-64?

This is something we needed to do to make sure we could scale as the new chips came out.

Are these page size changes the kind of re-engineering that is going to tune the 997s later this year to retake the lead in the product performance line?

A little bit, but there’s a lot of different aspects that go into performance capacity and scaling. We’re trying to make sure that our systems in the midrange and high end scale well. We’re still gathering data on what the performance characteristics will be.

What growth projects can you elaborate upon for the 3000’s low end? At IPROF you mentioned that PCI backplanes would be needed to refresh the 3000’s entry-level line. Won’t that push the refresh further out into the future?

All the new systems coming out from here on use PCI as the IO backplane technology. The change is pretty fundamental and essential. We expect to roll to the new PCI-based systems on the low end as well. It isn’t an optional thing — it’s required to get the new platforms. It brings us more in line with HP, but we’re doing it because we need to.

At IPROF you said IA-64 wasn’t a big area of focus for CSY. Did you mean for the coming year or so, or in general?

We always need to make sure we’re balancing immediate customer needs with investments for the future. For the current calendar year, CSY’s focus is not on IA-64. Our customers are telling us this is really the right balance. Customers want a long term roadmap that includes IA-64 while focusing our current R&D on the most critical issues.

We plan to begin looking at the long lead time items like compilers and object code translation during the next 12 months. We’re planning to get a grasp on what’s needed for those.

Will Modcal, the root language of MPE/iX, get full support under IA-64?

Supporting MODCAL is one of those long lead time items for IA-64.

There’s been talk about HP having the safety net of PA-RISC to cover any delays or disappointments in the IA-64 effort. Does PA-RISC have enough gas left to carry a customer well beyond the 2003 intro date for the 3000’s IA-64?

I think it does. HP has learned from back in the Spectrum days to evolve into a new design. The PA-RISC roadmap is always evolving. My personal belief is that these platforms and chips we have planned should carry our customers well beyond 2003. As we start looking at the long lead time items for IA-64, we’ll continue to update the platform roadmap and the timelines.

Why do you think there have been few 3000 customers and developers requesting information about IA-64?

Customers and developers are really running businesses, and they have a lot of issues to deal with. IA-64 is really a future for most people that they’re monitoring. But they have a lot more pressing issues right now, like Y2K and the strong economy is leading to a lot of growth they have to accommodate. I think people have their hands full now. The idea is to plan long, but execute short.

What’s the latest word on the decision to support Classic 3000 16-bit code in 64-bit HP 3000s?

We have looked at this a little bit, but it’s much too early to say anything conclusive about it. We need to figure out what the customer needs are and what makes sense technically as we plan our roadmap.

Some of your group’s work won’t be seen for years. How do you keep a team of engineers close to the customers when results won’t be tested in the field for some time?

A lot of the work we’re doing is going to be rolled out in the next six to 24 months. Some of the work in the peripherals area is even closer. The bigger challenge is keeping close to customer when we’re doing so much complex and time-consuming work. Engineers and managers like myself are consumed with executing, and that’s the bigger challenge. It’s not like the platforms won’t be delivered for years — it’s making the time to stay close to the customers.

We’ve been pretty successful in attending IPROF, managing the Customer Advisory Council and HP World. We try to invite customers to some of the coffee talks we have in the division. We have engineers and managers do surveys, and we have some consultants in the organization who work with enterprise customers and partners to connect Solution Teams to customers — to discuss needs directly.

It’s been a problem for me. I have so much going on that I don’t have as much time as I’d like to visit customers lately. We haven’t had a round of customer-focused visits, where we go out and experience a day in the life of a customer lately. I found those to be the most useful and insightful. I hope we can do more of this soon.

As a technical R&D manager, how do you see the shift in the new customer’s concerns away from technical details toward solution goals?

We think there are things that we’ll need to do differently for those sorts of customers. We’re in the process of trying to understand that better right now. The adjustments will help us target our R&D investment appropriately.

How does the feel of the engineering team for HP’s newest 3000 growth platform compare to the Spectrum project of the middle 1980s? Can you make any comparisons?

The teams then and today are very excited about the projects they’re able to launch. There’s a feeling of momentum, energy and excitement. We’ve hired a lot of people into CSY in the last 12 months. There was a lot of hiring going on then, too. We brought a lot of new equipment into the lab. All of this rapid change has been very energizing and exciting. And it’s been very challenging and exhausting. We’re taking a more of a phased and building block approach, as opposed to a complete rewrite of the operating system and everything else back then. We have a larger installed base now, so we need to make sure it’s more evolutionary.

We have a lot of experienced managers and engineers that have lived through the Spectrum experience. We’re really blessed with people who have been through this and know what to do. This time we’re taking a pragmatic approach with people who have been through this kind of change.

And there will be more change, now that Harry Sterling is taking on other duties. Have you felt some of them shift toward you?

I participate in our CSY HP 3000 Business Team, and I certainly feel like we’re being given a lot of responsibilities and a lot of opportunities to make decisions. It feels very empowering and interesting. The HP 3000 business has been very healthy, and is much more on the radar screen of HP now. I’ve been spending a lot more time with other partner organizations, like peripherals and diagnostics. Other HP organizations have become more supportive of CSY since it’s become clear that we’re really serious about the HP 3000.