Anatomy of Innovation: Exodus Founder B.V. Jagadeesh (Part 8)

SM: Essentially what you are doing at 3Leaf is allowing cumulative resource pooling of servers?

BVJ: Exactly. If you go back to the mainframe days you will realize it was absolutely a successful concept. You ran a single operating system on a big computer. The cost of running a mainframe was prohibitive for mass deployment. Exodus 6 is cheap, can run applications independently, and is very flexible. As you need computing power you can just invest $2,500. What you lost out on was the ability to have that expansion: when applications needed more computing power, they would not get it, and it then becomes inefficient. You end up doing the job in a very different way.

We take all the servers and we will give you whatever capacity you want. If an application wants 10 CPUs and 20 Gigabytes of memory, we can now construct that, custom for the application, from the pool of available resources. Sometimes there is a peak demand; the reason you have unused resources and computing power is that the system has to accommodate the peak. You use those resources and once the peak is over, you just return the resources to the pool.

SM: From a technical point of view, where does the technology sit? At what level is it doing the optimization?

BVJ: It sits on every server. It is a combination of hardware and software. It is actually a chip that sits on every server, and there is software that drives the server.

SM: Is it compiler-level code?

BVJ: It is at the processor level and is interacting with the processor. That is why the inefficiency goes away, because of the level of operations.

SM: I did not think that was possible.

BVJ: It is a very hard problem to solve. We are the only company in the world that has solved the problem this way. Several companies tried to solve it entirely in software, but if you do that you end up making the applications run very slowly. That resulted in companies just giving up.

SM: Does the operating system interact with your controller or with the processor?

BVJ: The technology evolution plays a key role here. AMD came up with a hypertransport interface to the processors. That allows us to take advantage of communication of the processor at that level.

SM: There is an API of sorts at the hardware level that allows this to take place?

BVJ: Exactly. Intel followed that with their own interface called QPI. Intel is an investor in 3Leaf.

SM: Who are you putting out of business because of this solution?

BVJ: We will create new opportunities and expansion of the market into applications that cannot run in today’s virtualized environment, or run very inefficiently in the previous architecture.

SM: Can you give me an example?

BVJ: Applications such as engineering applications like CAD and entertainment applications. They are going to be applications which require a lot of memory. We can give any amount of memory that an application wants. A lot of applications cannot run in today’s virtualized environment, or they run very inefficiently. That is why our first approach is to position in a way to expand the virtualization market.

SM: Are you already selling?

BVJ: We are selling the first phase of the solution which is the I/O phase which solves connectivity. We have a customer for that phase. The second phase is the memory and CPU virtualization.

SM: What is the timeline for that?

BVJ: We are looking at releasing that sometime next year.

SM: Is this a company that you have funded yourself?

BVJ: No, it has been funded by Storm Venture Capital, Ally Ventures, and Intel Capital. We just closed a $35 million Series C. From a financing point of view it requires a large upfront investment. There is a large initial investment required.

SM: Congratulations, this has been an interesting story that I will follow further.

This segment is part 8 in the series : Anatomy of Innovation: Exodus Founder B.V. Jagadeesh
1 2 3 4 5 6 7 8