About Web/Grid Services convergence

I have probably already blogged about the “Web/Grid Services convergence” marketing song that the Grid community has been singing since the introduction of WSRF. Jim Alaterascomment in my last entry made me want to write about it again :-)

We wouldn’t even be talking about ‘convergence’ if the Grid community hadn’t followed a path away from the one the industry was very busy preparing around Web Services technologies. OGSI represented an attempt to extend Web Services with behaviors “necessary” to build Grid solutions. The suite of technologies around OGSI were named ‘Grid Services’. Web/Grid Services were being developed in parallel and as a result we had two paths running at different directions. The community complained (well, ok… I had something to do with it too :-) and so we ended up with WSRF replacing OGSI. Although a good step towards the right direction (I am not going to go into resource- vs service-orientation in this post), WSRF was NOT supported by all players in the WS space. In fact, this is still the case and I suspect that we haven’t seen the last of this (the next GGF in Athens is going to be very interesting… I haven’t decided whether I will go or not :-). So, where’s the convergence?

To me, the suggestion that there is a convergence doesn’t make sense. It’s like saying that Bioinformatics or Banking Services converge with Web Services. There is a separation of concerns; however, I accept, the terminology doesn’t help in making it clear. Web Services (WS) is the term used for the suite of technologies built around XML to enable messaging (SOAP), metadata descriptions (WSDL, policy, etc.), quality of service (e.g. security, reliable messaging, transactions, etc.), and more. Grid/Banking/Bioinformatics/etc Services represent application-domain-specific artifacts which can be implemented using any underlying technology (WS being one of them). Grid Services should be seen as domain-specific services employed to realise the vision of service-oriented, high-performance, distributed computing.

While at Newcastle, I was always advocating in favour of a conservative approach to WS usage as the underlying implementation technologies for Grid Services. The idea was to always use the most widely accepted specifications for the underlying pluming and never-ever come up with new ones (unless having everyone on-board in advance), and to focus on product-quality, interoperable, specifications for building the higher-level stacks. Instead of arguing for years about pluming within GGF, we could have worked on defining high-level, interoperable, domain-specific services using standard and widely accepted Web Services technologies as the underlying pluming (this is exactly what a group of us suggested back in 2003). And, please, don’t tell me about the need to model state! Amazon, Google, and other providers exposing their business functionality through SOAP-based messaging do quite well without WSRF.

And you know what? I suspect that we haven’t seen the last of this, which will, once more, show that the Newcastle folks were right all along :-)

(Disclaimer: Please note that I am not expressing any new opinions above which I haven’t already publicly expressed before I joined




‘s position may be completely different from my personal one.).

Comments are closed.