Communications of the ACM
Taming the Clouds at OOPSLA
Workshops of OOPSLA 2009 started on Sunday and continued through to Monday. I participated in the two cloud computing workshops, the design workshop and the operational implications workshop and I'm just blown away by the amount of interest and research that is being initiated in this space . Academics, Startups , Enterprises, everybody is keenly working to get an edge in the Cloud space and many different and exciting ideas are emerging.
Some of the interesting talks that caught my attentions are the following
- Stuart Charlton (Stu) from Elastra. His talk on Cloud Independent Modeling of applications, mentioned three languages, ECML (Configuration), EDML (Deployment) and EMML (Management). The interesting fact is that these three languages are in fact ontologies. According to Stu, ontologies (a semantic technology that industry looks at with skepticism) offers relationships (at least the inheritance) for free and most of all a reasoning capability. It is indeed remarkable to see these semantic technologies getting wider adoption in the industry. Elastra will be soon opening up these languages and do follow their Website if you are interested. Beware though, these are first-cut drafts and may not be extensively documented.
- Nati Shalom from Gigaspaces . Gigaspaces is a provider of cloud abstraction services. Gigaspaces has the capability to distribute applications across different clouds and is capable of automatically provide load balancing. The beauty of their architecture is that since they have a logical representation of a machine resource, they can seamlessly integrate machines across different clouds and systems.
Apart from these highlights, the workshops had breakout discussions that were very productive and insightful. Following are some of the interesting discussions we had.
- What is cloud computing ? Initiated by Dave Nielsen, this was a discussion that was heated and also really insightful. It became obvious the term Cloud computing has a different meaning based on what kind of consumer you are. The audience was a mix of Industry and Academia and the presence of technologists from different tiers of both sides led to very interesting discussions. Unfortunately we did not zero in on a definition but agreed that terms on demand, abstraction and delivery method are important. In my opinion the NIST definition of cloud computing seems to encapsulate almost all of the important features and good enough as a working definition although it tends to overwhelm a newcomer.
- Modeling and designing for the cloud - A somewhat uncharted area for now but definitely needs (research) attention. The experience in the area of pattern based development for clouds is poor and just as in any other case, best practices and patterns emerge with experience. Cloud computing is still maturing and we are yet to gather enough experience to identify patterns. However the lessons learned during SOA may be applicable in the clouds to some extent.
- Licensing, SLA and other business level considerations about the cloud are still evolving but as important as the technical issues. For example elastic nature of the computing Clouds make some of the existing licensing models impossible to apply. There are some Cloud related licencing models that are emerging (such as the IBM's PVU based licencing model) but these are yet to be scrutinized and evaluated. Just like the modeling and designing aspects, business aspects of the cloud are subjected to evolve with experience.
I did receive good feedback on both my presentations, the one on middleware based best practices and also the ideas in using Web Service Level Agreements (WSLA) on top of the Cloud. I'm convinced that Altocumulus, the Cloud Middlware Platform we are to demonstrate on Wednesday is indeed a step towards the right direction. The conference is to start tomorrow (Tuesday) but I'm already convinced that it is going to be very exciting.
No entries found