Thursday, February 23, 2012

Simple Workflow Service - Amazon Adding One Enterprise Brick At Time



Yesterday, Amazon announced a new orchestration service called Simple Workflow Service. I would encourage you to read the announcement on Werner's blog where he explains the need, rationale, and architecture. The people I spoke to had mixed reactions. One set of people described this as a great idea and were excited that the developers can now focus on writing domain-specific code as opposed to writing plumbing code to orchestrate their actual code. The other set of people felt that this service creates a new cloud lock-in making it difficult for the developers to switch from one cloud to another as well as being able to interoperate at the orchestration level.

I believe this is a brilliant idea for a variety of reasons. Orchestration has always been painful. Ask the developers who have been involved in managing task execution across a cluster that required them to code for load balancing, handling exceptions, restarting hung processes, tracking progress etc. This is not a core competency the most developers have but they do end up writing such code due to lack of better alternative. The frameworks such as WS-BPEL were not designed to run in cloud-like environments and there has been no single standard REST orchestration framework out there that people could use.

From a vendor's perspective, I admire Amazon's ability to keep innovating via such services that differentiate them as a leading cloud vendor. As computing becomes more and more commodity, competing based on price alone isn't a good idea. If you're a cloud vendor you need to go above and beyond the traditional IaaS attributes even though you excel in all of them. I also see PaaS eventually bleeding into IaaS as IaaS continues to become a commodity. As far as PaaS goes, federated or otherwise, we're barely scratching the surface.

I don't see this service as a cloud lock-in but it certainly makes EC2 more attractive and sticky. I would be concerned if Amazon were to force the developers to use their SWS for orchestration. This is their version of how they think orchestration should be done and the developers can opt in if they want. And kudos to them to think beyond their cloud. The folks who worry about cloud lock-ins also talk about Amazon not following the standards. I believe that we should not create standards for the sake of creating standards. I am a believer in first showing that something works and later, if there's enough interest, figure out a way to standardize it. All these talks about standard-first even before you write that first line of code doesn't make any sense.

It's yet to be seen how this service turns out, but this is a huge step forward for getting more enterprise software customers onboard. Orchestration is one of the most chronic problems of enterprise software and with the challenges of a hybrid landscape to be able to orchestrate across on-premise and cloud-based solutions, this service is certainly a step in the right direction. Right Scale has been using a Ruby workflow Ruote for their workflow needs and now they orchestrate these workflows using SWS  to achieve fault tolerance and concurrency. As you can see, Amazon has opened up a gold mine for start-ups. The back-end execution has always been challenging. Now, there is an opportunity to write your own enterprise grade workflow engine or scheduler that runs in the cloud.

No comments: