categories

HOT TOPICS

Subscribe to our Feed

Thought Leaders In Cloud Computing: Sankarson Banerjee, CIO Of IndiaInfoline (Part 3)

Posted on Thursday, Feb 17th 2011

By Sramana Mitra and guest author Shaloo Shalini

SM: What were the implications of the decision to move to cloud-based servers from a cost structure point of view? In general, what analysis did you do to make that decision? I am sure there is a convenience aspect, which is one thing that I am hearing from a lot of CIOs. CIOs don’t want to be in the data center management business, and historically they have been for their respective organizations. One of the key drivers for enterprises moving into the cloud is that CIOs are fed up with being in the data center business. With the advent of cloud computing, they are outsourcing it to data center vendors. What was your thought process when you moved into the cloud, and what was the impact on your cost structure?

SB: Well, at IndiaInfoline, we got out of managing our own data center after the floods in Mumbai in 2005. The floods were very disruptive, so at that time the decision was made to move wholesale to the Netmagic data center, which had the proper protection against all of these kinds of disruptions an disasters. So, we have been out of the data center business for a while now. That decision made it a bit easier for us to move into the cloud because we did not have any of those servers on our premises anyway. A big part of our network is managed directly out of the Netmagic data center. Everything converges there, and our entire central server, risk management servers, all of them run there and all our offices connect there.

SM: So, the configuration you moved from was already out of your premises; the data center was at Netmagic. But were you still managing that data center yourselves, or was it managed by Netmagic?

SB: Netmagic was managing that data center and our servers for us.

SM: So, for you the driver to move to the cloud was more going from older servers that you needed to replace and so forth. Aren’t those the kind of things that Netmagic deals with now, buying servers, or is it something that you are still doing?

SB: No, these are all our servers that we purchased, so the capital expenditure for the servers was all ours. But Netmagic used to host them for us, put them on their racks, manage the power and connectivity, manage the switching on and off, and so forth. So, as these servers were ending their lives and we were on the verge of replacing them, that is when we decided to replace them with virtualized or cloud-based ones.

SM: You knew you would face a major capital expenditure in upgrading your servers or revamping your server infrastructure at the data center, and at that point you decided that you would rather go to the cloud and not make that kind of capital expenditure, right?

SB: Yes, for a big contemporary brokerage firm, because the volume of activity goes up and down with the market, the question for us to deal with was how we avoid a large fixed cost base that does not go up and down with the volume of activity.

SM: Right!

SB: Because you know, if you have let’s say, 500 servers, and the market goes down, you still have hundreds of idle servers to manage. You can’t do away with them; there is almost no resale value to server investment. If the markets go up, they can go up quite quickly, but you cannot scale up just as quickly because it takes time to order these servers, install them, and get them up and running. By the time you have them set up and configured, the market could swing and the movement may be on the other end of the spectrum. So, for us, a large part of the past two years has been spent converting our fixed cost base into a variable cost base.

SM: You have been successful in getting the kind of flexibility you need in dealing with market and volume situations with the help of the cloud technology, the peaks and troughs flexibility in your infrastructure that you are now set up with using a cloud vendor, right?

SB: Yes, we have made significant progress so far. I wouldn’t say that we have gotten rid of all our physical servers yet, especially on the database side. The database servers are much harder to move into a cloud environment. We haven’t been doing that in a significant way yet, so that is still a challenge. There are 40-odd core servers that the entire company was very uncomfortable putting in the cloud. Well, the discomfort was not only putting them on cloud but putting them anywhere outside the premises. That is not something we are comfortable with. The philosophy was, They work fine so we don’t want to touch them. So that part remains with us, the core of about 40 or 50 servers out of the 318 servers we have in the data center. Those are still maintained as physical assets. The remaining ones we have, various servers that I know of, like our test and development servers, we are slowly bringing those into the cloud along with all our casual application services. We launched new lines of business which in the near term require servers. Until we understand how big the scale of those new businesses is, how stable the business is, we decided to keep them on cloud based servers. Later, at the point when we understand what the exact sizing of that new business is, and then we can make a decision whether we need to have those lines of businesses supported by cloud or physical servers. For our new business, we almost always put them on the cloud first.

SM: Okay. Are there any other major decisions about your cloud strategy that are worth discussing?

SB: The other big decision that we made for the cloud was to adopt Google Applications. We decided about a year ago that managing e-mail was not our cup of tea. We had run into severe capacity constraint problems and maintenance and management issues related to e-mail. In terms of our workload, e-mail varies widely. Sometimes we are out of e-mail, sometimes we have only Gmail working for us. Planning for peak capacity means you have a lot of wasted capacity, whereas planning for average capacity means there will be things not delivered sometimes. In addition to capacity issues, there are things such as spam which pose a big problem for businesses such as ours.

Spam particularly is a huge problem for us. If we ran a good spam filter server, it made the e-mail too slow for us. If we didn’t have a good spam filter, we got caught in a lot of junk e-mail. The other big problems we had were that we were running our e-mail service based on skill levels that we had hired earlier. Skill levels tend to stagnate, whereas e-mail technology had moved on from when we had installed it in 2005–2006. Since then, e-mail had moved on significantly to e-mail on mobile devices such as the iPad and all of that stuff. We were struggling to keep up to date with the e-mail management software; we had constantly software and hardware upgrade problems. So, we decided at some point this was not what we wanted to be doing in terms of managing our e-mail service, and we moved to Google Applications.

This segment is part 3 in the series : Thought Leaders In Cloud Computing: Sankarson Banerjee, CIO Of IndiaInfoline
1 2 3 4 5 6 7 8 9 10 11

Hacker News
() Comments

Featured Videos