Coincidently, immediately after my post on ‘What the hell is Private Cloud?’, I happen to talk to one of my best friend, ex-CEO of a startup, now CDO (Chief Delivery Officer), in an IT Services firm @ KY, US – Midwest region. After my discussion with him I realized the need for a ‘scale out’ cloud. But, their company is dealing with Insurance customers and no body in that industry (just like every other industry), is yet ready to put their data out in the ‘Public Clouds’. Obviously the word ‘Private Cloud’ was lot more comforting to him.

However, as I mentioned in my earlier post, they are NOT Pfizer or Unilever, neither their custmoers are, to invest in ‘Private Cloud’.  BTW, once could use ‘Cloud.com‘ and build and manage your own ‘Private Cloud’. We can argue on the definitions for ‘Private Clouds’ and till the cows come home, but companies need solutions for their current problems.

‘Virtual Private Cloud’ seem to be a more viable proposition for companies who can’t afford their own ‘Private Clouds’.

Image Credit : Amazon AWS

Virtual Private Clouds (VPC – not VPN), are also called as ‘Hybrid Clouds’.  In the VPC approach, customers and system integrators can have the basic set of deployment infrastructure in their own premise and have a secure VPN access to public cloud / IaaS providers like Amazon AWS or GoGrid and the likes. The diagram above is self describing. Also it is worth to note that, you can also create you own subnet with the VPC and manage it very similar to the way you would manage things in your data centre.

In the VPC model, you would keep your Database within your data center and use the servers on the VPC as compute resources for burst computing. Data does transiently goes out to perform computations but never gets stored outside your company network.

For the case it point (project) we were discussing, we could easily get upto 25 servers for 1 hour crunching every day for 20 days in a month and release them all back into the pool once we are done with crunching. All we pay is for just 1 x 25 CPU hours every day. If we use AWS Extra Large Instance (roughly equal to 2 cpu with Quadcore each i.e. 8 cores in per instance x 25 = 100 EC2 Compute Units) – that is really a lot of compute power. Still, that will cost only US$17 for the 100 Computer Units running for an hour for day, for a month (20 days) it will be US$340 and another $153 assuming we transfer 1 TB (1024 GB) of data over a month. In total it is under US$500 . Isn’t that Amazing!

Sounds like Win-Win to me. What do you say?

PS: At the time of writing this post, it is not clear whether we will go this route. However, I will come back and update this post based on how things pan out.

Filed under: Cloud Tagged: AWS, Cloud

Read the original blog entry...