It’s been almost two years since we posted our article on Google’s server power for cloud computing. It appears that the term “cloud” has received a much larger interest, like a fad, which means that people are interested in what cloud computing and related services can do for them, whether it’s virtualization or scalability, the cloud idea has it. There is a sacrifice to this though, and that is that your data is stored and served in a sea of others. I think the term “cloud” has a interesting connotation to it, in that when you are in it, you can’t see if anyone else is. However remote, there is a risk that your data stored in this cloud is freely available to those who are hosting it.
Time has passed and Google obviously has either contributed to the cloud frenzy or have sufficient knowledge from trend data that this is what people are looking for, even though their “cloud” services have been around since gmail and beyond. Whether they are calling it cloud or networked computing or whatever, the point is their computing power is very likely way beyond what is projected two years ago.
The Wikipedia page1 notes criticism behind the private cloud but it makes no mention that these privately owned and managed systems not only provide them with all their services like communications, storage, etc but it also provides them with an added bonus of computing power that can be used for things like real time modeling, statistical analysis, financial projections, and more. These corporations are not interested in the “economic model”, if anything it saved them a few million by not having to purchase a supercomputer. Don’t think for a second that this massive computing power is not being used, and given that computers are getting less expensive over time, that these cloud systems are getting more powerful. Powerful enough that in the wrong hands, can do some serious damage. The fact that the community ignores this or hasn’t made the connection is intriguing.
Someone needs to get Google’s cloud computing power to go up against the fastest supercomputer (the k-computer at the time of writing). The k-computer is estimated at 10 PFLOPS, but Google can easily reach 10-20 times that with these cloud systems. Interestingly the k-computer is similar to a cloud type system in that it houses 88,000 processors. The only disadvantage in this case is that the supercomputers have a much more efficient way of crunching data, however the cloud systems can easily compensate for that in pure computing power. Put an estimated 200 PFLOPS against 10 PFLOPS and the efficiency of data crunching is irrelevant. Soon, cloud systems will be indistinguishable from supercomputers.
1 – http://en.wikipedia.org/wiki/Cloud_computing