ASP and IDC: Sides of the same coin

The ASP business model relies on a number of key aspects. The first is suitability of application for business needs; second, architecture of the application; third, access to the application and fourth, reliability of the access. While the first two aspects hinge on domain knowledge and software development expertise, the other two depend on public and private networking infrastructure.

A key technology enabler in the possible success of the ASP business model is the emergence of Internet data centers (IDCs). While IDCs also support private networks, their primary business dynamics are built around traffic from the public internet cloud. An important player in IDCs in the country is Net4India, a part of the UK-based Sawhney Group of companies. The company holds a category 'A' ISP license and has launched IDC services in Delhi, Mumbai, Pune, Bangalore, Chennai and Hyderabad. The IDC services cover hosting and co-location of servers with an array of 24x7x365 gaurantees.

Steering the technology direction of the company is Iqbal Gandham, Chief Technology Officer. He is 27 years old and has a masters degree in spacecraft engineering and communications and a bachelors in physics and astrophysics. He has previously worked with various UK-based ISPs, including World Careers Network, Prestel On-line, Onet and the Virgin Group of Companies. In an extensive interview with Arun Shankar, executive editor, Jasubhai Digital Media, Gandham expounds his perceptions on the role that data centers can play on delivery of ASP services in the country.

You have set up multiple data centres across the country. How does this help in network response and service differentiation?

Okay, why invest in six data centres of 2500 sq feet each, rather than one, which is 20,000 sq feet? The simple reason with 20,000 sq feet, you are looking at five to seven years, long term occupancy. After seven years, I don't know where India is going to be. I don't think anybody knows in the Internet age -it's very hard to predict these things. So we went for a slightly smaller model and said, okay if I fill these in a year and a half, two years, I can always migrate to a larger one. It doesn't take that much time to upgrade these things. Rather than outlaying a lot of money and saying I have a big data centre and don't know what to do with it.

While you are covering virtually the whole country with these data centers, aren't customers asking for mirroring of their Web sites or co-located servers abroad?

We have access to a data center in the UK, where we have a few racks available. You have to understand, I can colocate a customer in the US, but how do I mirror their Web site. See, if he has a database application it's not a straight forward thing.

If I am an ASP, would it make sense to host my application on all the six data centers, if I am looking at benefits of response time?

I think in the short run--yes, in the long term--no. With the amount of fibre, bandwidth and inter-city connect going on in India today, the benefit will not be there in about six months to one year. Our Mumbai to Delhi connection response is 30 milli seconds. Obviously there is the benefit of redundancy and off-site backs ups, but that is a different thing.

The idea of six data centers is localisation. If I have a large data center in Mumbai for example, and I ask a customer in Delhi to co-locate in my data center in Mumbai, he will probably say why don't you do it in the US. At the end of the day, the distance is still a bomb. So we decided to have six smaller ones because now the customer in Delhi can actually walk into my data center and touch his server. Otherwise I may as well just co-locate in America.

Many global ASPs are offering their services across the web versus the corporate VPN. With the current connectivity infrastructure in the country and the various available data center offerings, do you think application delivery across the web is feasible in the country today?

Well I think it depends on the application they are using. Obviously they would go into the private cloud if they are looking for security. The public cloud makes it easier to access from elsewhere and not necessarily from India. VPN across countries is not the simplest thing to do. You need to have relationships, server capacity and so on. Again it depends on the applications.

I think the crucial mistake, which most people seem to have made is they were trying to convert applications, which were meant for the desktop into an ASP delivery. For example--Office Suites. Why wont that model work? Simple, if that model did work then in a LAN environment we would be doing that already--say on an NT server--but it doesn't work. The only applications which will work are those which have been developed specifically for the purpose of ASP delivery or the internet. You can't just take something, which is standalone on your desktop in a LAN environment, load if off an ASP server and expect it to work. You have things like security, server performance and a number of issues for which the software was not designed. There are certain accounting applications, which were designed for this model and we believe these could be successful.

Suppose the applications were suitably selected and colocated at an appropriate data center. Would ASP offerings be viable within the country?

I think it is viable now. The only slow segment is the last mile dial-up connection...

Suppose further that the corporate ISP account and ASP services were available from the same vendor. Would that speed up the last-mile access time?

...Then it is obviously quicker. That's the only stumbling block apart from selling the concept to an individual.

On a more pragmatic approach suppose these offerings came from different vendors. Is there any way to speed up inter-ISP access, responsible for last-mile and overall latency?

That's the complicated issue right now. In the internet environment world-wide, ISPs have always worked together. In India there is a fundamental dislike of collaboration, which goes down to ISP peering points. Now without effective peering points in India you are not going to get effective traffic routing between ISPs. Now ISPs will not get into peering points with each other because they want a third party to manage those points. There are companies, which are doing that now but again it's going to take time. This has to happen, otherwise bandwidth is just going to be wasted through ineffective routing and peering. It will only happen if the top ten ISPs sit down across a roundtable and say, okay guys, let's get this thing working. This is going to benefit everybody, not just Net4india.

I believe there is some amount of peering of ISP traffic at VSNL...

I think you can call it peering, to a point. I won't say it's effective peering, but--yes, peering is occurring. Since they are the largest carrier right now and everybody has bought bandwidth from them. But you have to understand that VSNLs prime concern is not peering-it is not their business. So you need somebody who is going to dedicate themselves to peering and say let's get these routing tables in place. I reckon you need about five to ten CTOs to sit down at a round table...

And VSNL has to be one of them?

I think VSNL for the time being has to be one of them, since they have the maximum number of dial-up accounts.

Coming back to the original question--suppose the suitability of applications for ASP delivery, data centers, network infrastructure and ISP peering have been sorted out. Is the ASP model viable in the country today?

I would say yes, depending on the applications, routing and bandwidth--there is no problem using the ASP model. I would say no, questioning whether the fundamental ASP model is correct. You have to understand, with inter-city bandwidth and bandwidth into offices increasing, do people really want to outsource their applications or keep them local. Only the large applications are going to work on an ASP and reduce cost.

What are the system platforms that Net4india has standardised on for operations?

We have gone for Windows NT and Red Hat Linux.

Unix boxes are generally an important part of hosting and ISP infrastructure. Any reason why you haven't included them in your investments?

For one I am a great fan of Red Hat and two customers do co-locate Unix servers. But if I am going to provide hosting I just rely on Linux. If someone comes and tells me that a Sun machine is going to provide better performance for the amount of load and hostings I have on my service, I would say no it doesn't. For most of the customers over here, it's not going to benefit one way or the other. You have to understand for Web or mail servers the software you are going to use on the Sun or Linux is the same. Now what is different is obviously the processor. For 900 Mhz or 1 Ghz two-way processor--the dial-up connection and the bandwidth is never going to let them see the difference. To this day I have never had a problem with performance. You have to fine tune the kernels of even a Sun machine to actually get the maximum performance and I would much rather go with Linux.

Suppose a customer asked you to recommend a system platform for their web-based applications, how would you approach it?

If I was to recommend something I would need certain statistics from them. And when it comes to managed services I will offer them a service level agreement with suitable performance. Now whether I do it on an SGI or Linux machine, I will give that performance. I once moved email from a Linux machine to a Sun machine that had been incorrectly configured and it just collapsed. It's all down to how you fine tune the performance and clean up.

We have gone with Unicenter TNG from Computer Associates. TNG seemed to offer me a more complete suite of applications and I felt I could sell that to a customer much easier. Then you have to see how much support and experience is available in India as well. When I intend to buy software, I tend to not just look at the software, but the company and the people who are representing the company. If someone comes in and says it's going to do everything for you--on the whole I don't go with it.

Looking at the current dynamics of Internet and Web services, how is India different from the rest of the developed world?

The most crucial thing we learnt simply because we came from England and we have an ISP background, is that India is not America and never will be no matter how much anybody wants it to be. Customers here want to have the handshake and cup of tea, kind of 'no formalities.' They want to be more friends with you--whereas in America, its business. You come in, pay the money, post the server and it's done with. Here they want the local thing. They want to be able to speak to the CTO. They want to be able to send an email to the CEO with their own query and everything.

They are not looking for a space shuttle--especially for the data centre. They are not looking for biometrics or scanning or anything else. They want their server to be secure to a point, beyond that it's all frills. Most of the customers say, I would much rather have a guard outside and card access and that's it. Why do I need an eye scanner? I think that market may develop two years down the line, but its not here right now. There are some institutions in India who want that added level of security but on the whole the mass market is not interested.

Are you monitoring infrastructure and policy announcements to ensure that your technology investments are heading in the right direction?

We do monitor certain licensing policies but we take everything with a pinch of salt in India. A simple example is voice over IP--everybody in the world is doing it. I think, India is heavily based on hype. Everybody went into dialups, then everyone into dotcoms, then everyone rushed into IDCs, then wireless and broadbands. But then the hype dies down, there are one or two companies left doing it and the rest have got burned badly. Anything that happens in the US happens here as a knee jerk reaction, from what I have seen. If there is a dotcom bust over there it happens over here. If IDCs start closing down the same will happen here. And I am sure there are a few IDCs in the Mumbai region who possibly in the next three to six months may think of moving off or saying this is not for us. There is a quote somewhere on Forrester that says by 2003 there will be 30% extra capacity on IDCs worldwide. So I think a few people are going to slow down their expansion plans.

Your title is CTO. Can you explain what is your job role and what is the value that you provide to your organisation as a CTO?

I think the days when a CTO was just a technical person are long gone. Just talking technical all the time is not going to help in the long run. Our CEO for instance knows something about technology. I, in return have to know something about marketing and the business side of things. And I need to be able to negotiate. The days when CTOs were given, Rs 2 crore per quarter of funding are long gone. Before I buy something, I need to work out how am I going to sell it. So when I go down into the technology level, I try to understand how to sell it first before I even ask for it.

In addition I have to worry about HR. In an IT based industry, your staff is the most important thing. The days when a CTO could have layers of management between them and the people doing the work are over. For the simple reason that people like to report directly to the senior person in a company. If they can do that they are happier. And two, customer support and the systems staff who talk to customers give you the best feedback about how your company is doing, more than anybody else. So if I lock myself in an office and don't talk to them I kind of have the wrong impression of my company. The CTO role is one of the most crucial in the company because it is linking technology with business and if you can't get that right you are kind of going to go in this direction and that direction.

Is there any aspect of this role that is critical for you or that gives you sleepless nights?

I actually worry about everything from basic dial-up services, networking services like VPN to hosting, co-location, data centers, bandwidth, all the way across the board. If I provide an end to end solution to a customer, then I have to look after my entire operations, end-to-end, as well. It's not something that I can say--okay, I am just looking after the data center and the hosting will take care of itself. The hosting and the data center are linked, just as my networking operations are also linked. So, day to day I don't have to configure a router, but I have to keep hands on, which is not a simple thing because there is new technology everyday. I have to keep upto date and try to run the business as well. It's not simple!