News Stay informed about the latest enterprise technology news and product updates.

Former Cassatt scientist Steve Oberlin explores cloud computing market

Steve Oberlin discusses where cloud computing is going, how history, from Cray to CA, has repeated itself, and more.

Steve Oberlin began building supercomputers in the 1980s as the chief architect for the Cray T3D and T3E systems. He moved into distributed computing and helped found and develop core technologies at Cassatt Corp., a pioneer in cloud computing. CA acquired key parts of Cassatt's technology in June, which Oberlin helped develop from 2003 onward. Oberlin spoke with about what's going on in the cloud computing market today.

Where is cloud computing in terms of the technology that underpins virtualization and automation?

For more on cloud computing and automation:
CA buys Cassatt to stand above cloud fray

IBM sheds light on cloud strategy

Cisco exec Marie Hattar talks network virtualization, cloud computing
Steve Oberlin: I think there's still a lot of maturity that needs to evolve and be implemented in rigorous ways. I think one of the things that's been most dramatic over the last year [to] 18 months has been how quickly people have come to assemble a worldview around cloud computing and what it means to them in various different ways. People are seeing that cloud computing is providing solutions in a lot of dimensions, but I think we're still in the early days of adoption.

What's ahead for these systems in how they're built and distributed and organized?

S.O.: You're seeing some vendors trying to come out with what they're calling "unified solutions" in one way or another; look at Cisco's recent offerings. I think you'll see more vendors trying to put together distributed computing clusters that are pre-configured with a certain set of capabilities. That way, people can do the "forklift upgrade," if you will. This will be one way that people can treat cloud computing in their environment as a complete appliance.

People are seeing that cloud computing is providing solutions … but I think we're still in the early days.
Steve Oberlin, former Cassatt Corp. engineer,
In addition to that, you're seeing more mature offerings coming out of the external cloud vendors as well. The capabilities that they're providing now are going beyond merely making platforms or virtualization available as an easily requested and returned commodity; they're actually providing mechanisms for having automated, load-based, elasticity management rather than having the users of cloud provide their own demand-based mechanisms for requesting and returning resources to the [public] cloud.

There's a land rush on in terms of cloud computing right now; everyone and their cousin is calling what they do "cloud". What are your current thoughts on the taxonomy question?

S.O.: I think the SPI [SaaS, PaaS, IaaS] taxonomy is winning out -- it's the clearest explanation that people have so far. Obviously there are a lot of subdivisions and combination offerings are definitely going to blur the lines. When you think about taxonomy, the whole purpose is to help people understand the relationships between the various members in the hierarchy, if you will.

In a 2001 abstract, you discussed the evolution of distributed computers as it replaced the supercomputers you cut your teeth on. Almost 10 years later, where is that phylogeny now?

S.O.: That's a good question. You know, from a high performance computing perspective, in some ways things have followed a "Moore's law" path almost exactly. What happened is something that a number of people were talking about quite a number of years ago; and that is we [have] reached a point where from a pure computing technology standpoint the best use of transistors on a processor chip is no longer to try to improve single processor performance. Now we have multicore processors where we're effectively starting to replay the history of server computing on a single chip.

If you look back a couple of decades, we had single monolithic large CPUs, then suddenly we started having symmetrical multiprocessors all sharing memory, and then at a system level we had an architectural break into the distributed computing path. We're almost seeing a direct analogue of that taking place on a single chip now. Really quickly, I think you're going to see the same efficiency break happen on microprocessor chips as well. People will move different elements of the system onto the chip instead of just putting more and more CPUs on there. Look for memory and networking, complete systems on a chip.

Why did CA want Cassatt's technology?

S.O.: One of the things that's been sort of surprising now that Cassatt is part of CA is really seeing what the depth and breadth of CA's products are: Their existing expertise and cloud offerings and comparing and contrasting those to what Cassatt had developed. In a way, you can look at what Cassatt's focus was in terms of product innovation and architecture innovation as a laser: extremely bright, very focused but also kind of narrow. I think in customer's minds there were valid questions left unanswered about how exactly this miracle occurs.

So they started from the end of governance/management/SOA/compliance tools that Cassatt was missing?

S.O.: Exactly. They started on the outside with the holistic picture and have been adding these various automation pieces. Cassatt spent a lot of money and a lot of time doing a lot of work to deal with heterogeneity in all the different ways there are to connect and to monitor and to control all the various types of resources that are out there and frankly, a lot of times we did as good a job as we could, but I can tell you it's nowhere near as complete or as professional a picture as what we're seeing CA already has in that monitoring capability.

Is Cassatt's technology going to replace the underpinning of CA's automation technology?

S.O.: CA already has the provisioning and the basic end effectors of that. Don Ferguson, (CA's chief architect) describes the automation flow as "monitoring, analysis, planning and execution" and clearly CA has very strong capabilities at the two ends of it. Where the Cassatt technologies are going to push the envelope for CA is on the analysis and planning, modeling the environmental choices and the optimization choices and making the best picks and speeding up that whole process.

Carl Brooks is a Technology Writer for Write to him at And check out our Troposphere blog.

Dig Deeper on PaaS and other cloud development platforms

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.