But wasn't OCP an alternative route and a validation of the fact that a company does not have to limit itself to buying traditional gear from the likes of HP, Dell or IBM?
Perkins: It is and it isn't. The open compute project was initially started off by Facebook. They 'open sourced' their infrastructure. People like Rackspace and others have since then proved what Facebook started off—the idea of passive cooling in data center. The most important thing here is that our data center designs are open source based. It thus makes it possible for all sorts of companies to use it. Whether you build it yourself or whether you approach a company that is already an expert at building infrastructure—is secondary.
In fact, I feel it's better to use best-of-breed hardware by those who know how to build things to that (open) standard than to try and build those things by yourself. The key thing here is the fact that the 'standards' are open. So buying from one vendor does not lock you into a proprietary design from them in the future. So it is very easy to mix and match different players, because it's all built to a standardized open source design. Just like how we adopted Open Stack, the open source cloud operating system, we now have open compute, which is the open source infrastructure platform. So factors like who does the software design or who does the hardware manufacturing are actually secondary.
How relevant is the 'open compute' discussion to an end customer?
Perkins: It depends on the customer. If they have a corporate conscience around green IT, or if they want to market a green message, then it's very relevant for them. The savings in open computer environment are quite significant in terms of power consumption. Typically, there is no air conditioning being used. Most of such data centers use the air conditioning only one or two days a year. The usage of metal is up to 40 percent lesser than proprietary designs.
Do you think the recent cloud outages have shaken the customer confidence in public cloud?
Perkin: As a CIO, who has had a chance to use many of these technologies, let me tell you that this can happen to any computing system. The whole notion of cloud, public cloud particularly, is built around the fact that 'computers are designed to fail'. The beauty of cloud is, it enables you to design your architecture in such a way that computers can come up and take the load if their 'peers' go down. Cloud providers need to build their system keeping this factor in mind. At the customer place, the buck still stops with the CIO. That's why SLAs are very important now. We give our customers monthly SLAs now, rather than annual ones. It's much easier in an annual SLA for a vendor to get away with a big outage because it's spread across the year.
Sign up for CIO Asia eNewsletters.