Saturday, June 27, 2009

Clouds In The Sky

After a long time in the industry has a technology come which people are so excited about – everyone trying to define it and everyone trying to explain what they think it is. It is surprising that still, most articles on Cloud Computing start with defining what the author considers constitutes cloud computing. 

Most however seem to agree that a cloud enabled infrastructure enables:
  • Dynamic provisioning
  • Scalability and flexibility
  • Universal access
  • Computing cost as opex with little or no capex expenses

What is not talked as much as the hype on the cloud but is bound to get relevant as people get into enabling it for their enterprise are aspects related to:
  • Security on the cloud ( yes, there are people already talking of it) is something which needs to be considered at all levels from application layer to infrastructure layer
  • Very high dependency on the WAN bandwidth for performance of the applications. Until now, for those enterprises with critical applications on computing infrastructure within their enterprise, they were mostly accessing it over a LAN for a large part of the population typically in the head office. Other users would access it over a corporate MPLS network. Now, everyone does the same and while it is possible to get dedicated access and get the same experience, the network charges are bound to go up
  • The big challenge of application migration where applications will need to move from existing infrastructure in-premise to cloud infrastructure. This will throw up a whole lot of issues where currently running applications for years will need to be unseated and moved to a new environment or even may be get a parallel application environment stood up in the new cloud infra. However as more critical applications start moving on this path, unless these have been recently deployed, challenges will come forth on getting the exactly same environment in the cloud. It is not unheard of, to have critical applications in large enterprises running on hardware or software systems, some of which are end-of-lifed. They continue to run as the cost of finding an alternative or migrating to the new version or system is much higher than running it with higher maintenance for its useful life before retiring them. It may not be easy to re-create environments which require setting up an old out-of-support version of Oracle for a customer who some part of the ERP is connected to an application which runs on this environment. The older the systems and the more customized the applications, the greater would be the barrier to move them to the cloud.
  • Lack of control for businesses over IT. Now this is something which seems pretty rampant but not discussed so often : it is not uncommon to find businesses running small (or even mid to large) IT departments outside the “corporate IT” for certain applications and systems that they closely work on. In some F500 companies, if aggregated across teams, this is even upto 50% of the size of the corporate systems. In other companies, while businesses use the common services of “corporate IT” they dictate the requirements : “Well, this is a mission-critical application related to blah blah ... and we need a dedicated environment for this application. We cannot have it share computing power with any other application and need a high-end server with ....” With cloud, to some extent, businesses will lose such control and may not be so amenable to have their applications moved to a cloud.

Overall the adoption to cloud will happen in a phased manner where enterprises will test the waters or rather test the clouds, with some early applications. That process is bound to take a while, but moving test and QA environments would be a better bet for lot of reasons. That is the subject for another post!

No comments: