New CEO John Lambert on Virtutech Present and Future
New CEO John Lambert on Virtutech Present and Future
By Ann Ernst published: Wednesday, October 05 2005
Virtutech’s full system simulation technology provides the foundation to move software development from unwieldy and expensive hardware to a programmer-friendly simulation environment running on every programmer’s desktop.
Following their announcement of a new CEO, John Lambert, VSM spoke with Mr. Lambert and with Paul McLellan, VP of Marketing.
VSM: John, you’re new to Virtutech. What brought you there, and where do you see the company going?
JL: First and foremost, I was looking for some significant opportunity to change the way a market operates, and to materially and over an extended period of time - improve the working lives of our potential customers.
I saw that what Virtutech was doing had the potential to fundamentally change how our customers in both the defense and aerospace and the datacomm/telecomm industries go about building the embedded subsystems at the center of the solutions they sell to their end customers. That was very exciting.
I have an embedded background and can relate well to the challenges software developers face in those markets. I spent 15 years with Rational Software both in the UK and the US, running a mix of field sales organizations and product development. Part of that time I was in charge of the software development environment that Rational sold primarily to the defense and aerospace embedded community. One of the basic challenges faced by those engineers es that the tools they use on a daily basis are still very much the same today as they were back when I was writing assembler code in the early 80s.
Virtutech’s technology fundamentally breaks that mold, and that’s what’s getting us market traction. Our solution has delivered measurable value to customers over time, and has been engineered so that the customers we approach feel comfortable making the switch and coming to rely upon our technology. In both due diligence and meetings with customers since I’ve been here, feedback has been universally positive.
Another draw for me was the caliber and passion of the people involved. Peter Magnusson, the founder and CTO of Virtutech, has done an outstanding job of getting the company to the point where it was necessary to bring on some additional management to deal with the breadth and depth of business opportunity he had created.
Lastly, it seemed to be a place where I could fit in and contribute to the growth and development of the company and the team.
VSM: How has the transition been?
The first order of business was to dig in and understand in more detail that Virtutech market opportunity. We closed a very significant deal in the second quarter with IBM that helped me come up to speed very quickly. That was an excellent validation of the strength and depth the company has built up since founding, both on the technical side and within the sales organization.
As we move forward, a key question is, “how do we exploit that long-term opportunity to change the way complex software systems are developed?” We are beginning to pull that next piece of the strategy together and understand how the virtualization movement plays into what we want to do as a company. Our goal is to become an integral part of every software developer’s ability to build, ship and maintain these complex system environments.
VSM: How did the IBM opportunity come to Virtutech? What was the value IBM saw in your solutions?
PM: IBM has a problem: They are building their next-generation server mainframes around next-generation architecture, the Power6. They don’t have any silicon for that, they don’t have any hardware, but they have a whole lot of software that they have to get written, going from the firmware and the hypervisor, the lowest levels of the virtual operating system, all the way up to applications.
The only solution they have to do that, other than waiting for hardware to be available, which is unacceptable, is to use simulation. They had an internal simulation project they felt wasn’t good enough, so despite the fact that we are a relatively small company and that this is a mission-critical project for IBM, they chose to rely on Virtutech’s full system simulation as a way of pulling in their time to market.
The measure of time to market, and improving the productivity of the development process through simulation, gives you a lot of things you don’t get with real hardware. Hindsight is the most obvious the ability to run backward but you can also get a non-invasive look at lots of detail going on under the hood in a simulated environment.
VSM: It sounds like companies try to find a way to do this simulation work in-house because there haven’t been viable tools to work with embedded software.
PM: We like that because it means we don’t have to go through a lot of evangelism to have prospective customers understand why employing simulation at the full-system level makes sense.
In the defense and aerospace community, because of the nature of what they build, they have recognized for a long time that, ultimately, the risk inherent in building those systems is in the software and not the hardware.
When you look at other verticals such as datacomm/telecomm, there isn’t that same level of fundamental familiarity with these kinds of technologies. Systems companies like IBM recognize what full system simulation means because typically they have tried to do it themselves. They very quickly latch on to the benefits of a fully commercial, fully developed product over whatever internal efforts they’ve had.
VSM: Simics models from the silicon up?
PM: Simics models at the software-hardware boundary. We model the hardware at the level that the software binary runs unchanged, device drivers and low-level stuff. The software can’t tell the difference; there’s no way that the IBM hypervisor can tell whether it’s running on real hardware or on Simics. Even though the simulation may run at a different speed, the clocks are simulated too, so the time will get simulated exactly as it would on the real hardware.
We model at that level, which is the optimal level of detail. If you model at a higher level, if you try to simulate at the operating system boundary for application programs, then that’s tremendously difficult to make accurate and to maintain. At the lower level, there’s so much detail that makes modeling and performance issues a challenge. The instruction set architecture, which is where we’re modeling, doesn’t change very often and, therefore, that’s an easy place for us to model.
Once you’ve done it, it remains valid for a long time. It runs fast, not necessarily as fast as the real hardware, but fast enough that IBM can get development done. Every software engineer can use it as part of their edit-compile-debug cycle. It moves onto every engineer’s desktop.
Even if they had hardware, the engineers would have had to share relatively little of it amongst a huge group. But it will be another year or so before they have any silicon, let alone have any systems. So it’s a timeto-market improvement, it’s a cost improvement, and it’s a productivity improvement for the engineers.
JL: Our core constituency is the software developer. Our mission is to bring the benefits of virtualizing those hardware elements onto a software developer’s desktop and use that to fundamentally improve the way they develop software. Nothing of the revolution in the standard desktop Microsoft development environment seems to have filtered down to any of the more complex, real-world systems such as airplanes, large-scale servers or other domains that we target.
VSM: As I recall, this sort of simulation historically hasn’t been widely used by software systems development companies like IBM. It’s mostly been in the defense industries, airlines, cell phones. So this must be an important step for Virtutech?
JL: Absolutely. The opportunity for us lives in the realization that full system simulation brings a new level of software development capability that’s of value to an entire industry. The challenge with instruction set simulation in the past has been that a software developer couldn’t get the results they needed to feel comfortable using a virtual replica of the hardware on their desktop; they had to wait until the real box was available to do the significant test integration debug activities. This was because previously instruction set simulation was slow, it was too low-level and it didn’t really meet the needs of what is, in practice, a highly-interactive activity: software development. Virtutech Simics has rewritten that script by providing a solution that is so fast that now every software developer can use it to aid their day-to-day development work.
PM: If you go back further, instruction set simulation was primarily focused on validating the hardware using a bit of the software. Even though it was slow, it was much faster than the simulation that the hardware people were using for the rest of the system.
People didn’t try using it for software development because it was off by a factor of maybe a hundred times too slow. It is a big difference for a software developer if something takes five minutes, which you can do as part of your regular debugging; or if it takes 100 times as long 500 minutes which means that you set it going before you go home and it still hasn’t finished when you come back in the morning.
Clearly that’s not going to be something you use on an hourly basis during the day. But our technology - plus the technology that Intel developed that improves the performance of the basic underlying PC hardware on which we’re running - have improved price performance. Ten years ago, to run simulated minutes of a program on this IBM mainframe internally at IBM would have taken so many computers that it would have cost $1,000. Now it’s about $0.001. It’s that kind of dramatic difference.
Performance in instruction set simulation has traditionally run at about 1 MIPS. Our technology runs in the hundreds of MIPS or even billions of MIPS, billions of simulated instructions per second 100 to 1,000 times faster.
VSM: I assume you will work with IBM to track, over the course of the contract, the savings in terms of time and dollars?
JL: Yes. In fact, many of our customers create and track metrics and measurements they use internally to assess ROI. Indeed, we encourage them to do so. Along those lines, we view this as an initial engagement with IBM, and by proving our value on this first project we’ll earn the right to exploit all of the significant opportunity we see beyond this group and out into the broader IBM community.
VSM: The last new product-related news that I heard from Virtutech was about Hindsight.
PM: We anticipate the full release in the third quarter. It’s currently in beta test with a number of customers where it’s going really well. We had a customer who had three people working for three weeks to track down why one router wouldn’t boot. We brought it up in Hindsight and understood what the problem was in 10 minutes.
The other point worth mentioning is that virtualization itself opens up a whole new vista of opportunities to test at the system level, even beyond the value that it brings to the sub-system level that engineers haven’t been able to approach in the past.
Once you can virtualize an element to this system, a board or a particular device, it’s very straightforward to be able to build up a system simulation based on multiple instances of these things connected across either real or virtualized networks. Then you can run real end-to-end use cases that are material for end-user functionality.
Historically, that’s difficult in these complex system situations. Trying to have even one instance of the entire end-to-end system hardware was very difficult for any sizable project.
VSM: Looking at data center virtualization from a systems view, while point by point there may not be many comparisons, at an overall level you could compare looking at a data center system to looking at full system simulation for these embedded platforms.
JL: Is there evidence of people beginning to do that now in that space?
VSM: What we see is that once you virtualize storage, servers and networking, you need management tools to look at that as a system, looking at provisioning, service-oriented architectures and those kinds of things in a more systemic, data center-centric way than looking at each piece individually. It’s a system with pieces that get virtualized and then managed at another level. So in that sense it does fit the open system virtualization model.
JL: I hadn’t thought about it in that way before. I first came across virtualization at the box level, as part of a software tool development group. There, we began using VMware to multi-boot a particular box so that you could run regression tests on different versions of a product running on a range of different OS variants.
In my last company, we saw virtualization being picked up in the data center as a way of maximizing the performance available on any one box, rather than buying a new server every time a new app came along. Then VMware began to add management-level capabilities such that the technology moved away from being focused on a single entity to managing a community, a system that had to operate in consort to deliver data center functionality.
VSM: You will be releasing Hindsight and you will be focused on IBM and building other relationships, but in terms of a product, what does the future look like?
PM: We’re not announcing any other products at the moment, and our roadmap is still being defined. One area in which we expect to announce is modeling. It is a challenge in full system simulation to get the software people actually developing on virtual models of the system, because someone’s got to create that virtual model.
Customers will always do some of that themselves; we can’t model every platform in the world. But in order to make it effective for customers, we have to make it easy for them to do. My vision is that semiconductor vendors who create components will eventually be forced by software developers to supply models of those components.
We will be announcing a strategy around a Device Modeling Language.
VSM: How does Virtutech compare to VMware?
PM:What we do is directed at engineering, making it possible for people to develop code with a lot of instrumentation, so you can look at what’s going on under the hood. We are less focused on reliability and security, and whether you can run 24/7 for six months in a row. If Simics crashes, it’s just inconvenient.
For VMware, who are focused on production virtualization, it’s very difficult to look under the hood and see what’s going on. But that’s not what they do. They make it easy for you to take 20 servers and run them on three computers. Then when that’s not enough you bring in a fourth, and seamlessly move all that stuff around, and it never crashes.
Those are two different markets, and although some of the technology being used is the same, they’re different as to how it’s delivered, and to what the customer cares about.
JL: The power base in the computing industry is quickly moving away from being focused on the hardware element to reflecting that it’s really now a software industry, because ultimately that’s the persisting and enduring value of the work that gets done in this space to create end-user, application-level functionality. Virtutech is very well positioned in our markets to facilitate that shift at the software developer level. In addition, we’ll be working hard with the semiconductor providers as they move to multi-core architectures, which is their way of supporting the ever-increasing need for more compute performance, itself driven by growing end-user application requirements. In short, and to recap where we started the discussion, Virtutech has an outstanding opportunity to help shape an industry, and there’s nothing more exciting than that.
*****
For more information about Virtutech, Simics and Hindsight, visit www.virtutech.com.
