Monday, April 6, 2009

Talk on SOA & cloud computing at SOA Symposium

I had the pleasure of participating in the 2009 SOA Symposium - Government and Industry Best Practices on Thursday and Friday this past week. This was an event organized the US Department of Defence's Business Transformation Agency as part of its efforts to deploy SOA in the DoD.

Dennis Wisnosky, who was the "patron" of the event, is the Chief Architect and CTO of Business Mission Area of the Office of the Deputy Chief Management Officer, U.S. Department of Defense. He has the uneviable task of trying to get an incredibly large organization (of 3.5m+ people) adopt some type of consistent SOA architecture! That's when you realize how easy we technology providers have it .. its a LOT harder to get real, large organizations to adopt/adapt than to come out with yet another piece of cool technology :-).

I gave a talk titled "Scalable SOA in Your Own Open Source Cloud." Feedback welcome!


I also met Thomas Erl again (since the January meeting in Germany) and also finally met JP Morganthal and Mark Little f2f! Very much a pleasure indeed .. I just wish I wasn't so damned jetlagged and out of it most of the time. Arriving into DC without my bag didn't help either.

4 comments:

Eran Chinthaka said...

A nicely presented set of information on Cloud without any marketing bluff.

Vendor lock-in is a problem in Cloud, true. But at the same time for the economies of scale to kick-in it has to come from vendors with each individual company trying to optimise their platforms and APIs.
"Open" can not come from sets of individuals who have few computers to play around (as it is with most of "Open Cloud Forums"). Also the three main players of the cloud industry, i.e Google (SaaS), Amazon (IaaS) and Microsoft (PaaS) are on three places in the cloud spectrum. And interoperating these three itself is a hard problem . So most of the companies or individuals has no other option than to follow them. For example, when Microsoft comes with all the Azure service integrated with Windows next version, it will become the de-facto for "most" people.

The other issue on Cloud as I see is the myth about clouds on its ability to run "any" application. First the application should be scalable to run on clouds. Its very hard to get an existing application as it is and run on clouds and boast about the scalability. But this is exactly what is/was happening with most of the companies trying to move to Clouds.
Second is that not every parallel programming model is suitable in Cloud. Yes, map-reduce style models consisting of individual standalone worker nodes might work well. But MPI style models requiring. better communication infrastructure will suffer heavily on most of the platforms. Some people consider HPC and Clouds as "twins separated at birth". So considering cloud as a silver bullet will be a disaster, IMHO.

Enrico Lisk said...

Almost three years down the line, whats your take-on on the adaptation of cloud in a Sri Lankan context. from a vendor & client perspective.

Enrico Lisk said...

Almost three years down the line, whats your take-on on the adaptation of cloud in a Sri Lankan context. from a vendor & client perspective.

Enrico Lisk said...

Almost three years down the line, whats your take-on on the adaptation of cloud in a Sri Lankan context. from a vendor & client perspective.