Ejasent goes head-to-head with outsourcers

by Michael Vizard

April 2, 2001 —

AS CTO AND founder of Ejasent, Rajeev Bharadhwaj envisions a new paradigm of computing based on a utility model in which customers only pay for what they use.

In an interview with InfoWorld Editor in Chief Michael Vizard, Bharadhwaj explains why computing utilities are inevitable and how his start-up company will compete, beginning in February after completing eight pilot programs, with rivals such as Loudcloud, LogicTier, Mimecom, and a host of traditional outsourcing companies.

InfoWorld: What makes you think that computing utilities are the next big thing?

Bharadhwaj: We think that over a period of time everything is going to be bought in bulk and sold in pieces. This has been true for the traditional utilities, such as water and electricity. It is now true for bandwidth, it is true for content, and the last and final thing that it will be true for is applications.

People are evaluating the model right now for content and bandwidth, but that is the least expensive thing to buy in terms of the total amount of money people spend. The amount of money people spend on bandwidth and content is substantially less than what they're willing to spend on applications. But it's very difficult to do applications as a utility. It's a long and hard process, and a good portion of our company's life has been spent on building infrastructure to provide that.

InfoWorld: Among potential customers, where do you see the most resistance to your approach?

Bharadhwaj: Today, the people whom we find are initially resistant are the Web operations people. The issue is not owning the equipment or the data; it's just that they feel they can't control the execution of things.

But this is where we have an extremely good value proposition. We actually provide huge amounts of control to the customer. You can get all the monitoring you want and you can make all the changes you want on policies and decide where you want to place things. Whatever changes you want to make to your application, you're free to do it because it's just going to get propagated on our network.

We're saying that we give you the response time that you want at the price that you want to pay, and you get the same response time regardless of your data load or geographic location. We say that you're able to control your costs without having to do capacity planning or incurring big outlays of capital.

InfoWorld: People have been talking about shared infrastructure for years. First we had time-sharing, and today we have ASPs [application service providers]. How far along are we in terms of creating a true utility model?

Bharadhwaj: It's basically the same wheel that has come around for one more spin. You had time-sharing back in the 1960s. At that time you could actually buy time on a machine and you could go run jobs.

Now we've come full circle, and computing as a utility is really about time-sharing in many ways. It's not just time-sharing of machines but it's also time-sharing of knowledge. The reason why time-sharing went away is because at some point machines became inexpensive. But the Web changed everything. Demand is unpredictable and you need [access to Web applications] immediately. Otherwise you're going to lose something. That's why it has become clear that computing as a utility is going to become very important.

InfoWorld: What makes Ejasent different from companies such as Loudcloud, LogicTier, Mimeecom, and other managed service providers?

Rajeev Bharadhwaj, Ejasent

Age: 38

Title: CTO

Biggest success: Coming up with and sticking to the Ejasent idea and business plan even when Linux and dot-coms were the fad among venture capitalists

Key challenge at Ejasent: Staffing and integrating a 100-person organization with a wide range of expertise

Personal notes: Enjoys Indian classical music and is married with one child

Bharadhwaj: We actually do not compete directly with the managed service providers. The managed service providers are Web operators who create complete solutions for some customers. Because they have a stovepipelike approach, every new customer requires a whole set of machines to be set up for them. There's no sharing in that infrastructure. In a computing utility, sharing is essential.

InfoWorld: How do you take an application running at a customer's data center and migrate it to your network?

Bharadhwaj: We work with any application running on Solaris.

The first phase is site capture. We go to the customer's site and take a snapshot of their whole application, the whole stack, or the portion of the stack they want to outsource. After we do that, we prepare this site for deployment into our network. That consists of a phase called site-splitting. So the customer always has the main site, and anything you want to move from there to us can be done using links.

InfoWorld: So can just about anybody who is in the outsourcing business today set themselves up as a utility tomorrow?

Bharadhwaj: There are two important things you have to solve before you can become a computing utility, and they all [revolve] around sharing.

The first problem is application switching. A box that is running a certain application stack needs to quickly switch from running for customer A to customer B. The switching is essential to the process. If you don't have switching, then you have the stovepipe model, and you're going to spend a lot of money on hardware and support.

Second, we have separated applications from machines. We think, in the future, more companies will leverage this kind of infrastructure. The magic takes place between the application and the kernel. It's in that layer between the two. We have a very strong relationship with Sun, and we basically put this application switching in the operating system.

InfoWorld: So ultimately will this be the dominant computing model in use in corporate environments, and how long will it take evolve?

Bharadhwaj: I think it'll be a 10-year kind of thing. There are a lot of people who feel that, in the future, hardware vendors are only going to sell to a set of service providers. In fact, Scott McNealy [the CEO of Sun Microsystems] said there'll be only five companies that will actually buy hardware from him.

InfoWorld: But if it gets that big, won't you then find yourselves competing with companies such as EDS?

Bharadhwaj: At this point, everybody views us as a partner. We haven't found any competitors. But as you said, this industry is a little bit cannibalistic. But that's how innovation happens.

InfoWorld