I could explain further but it would be a long blog post. We don’t need a complex solution. We only need a simple distributed state sync protocol that uses asynchronous events and knows when it is disconnected or out of sync. Just like a browser UI.

Not sure in what stage of Events API Amazon AWS is working right now or if they really started working on those features requests. Definitely good to have these features also with other cloud operators and not only AWS.

From the point of view of [platform] developers, this was good — we wouldn’t have to go redo everything. However, from the point of view of web service developers, it was scary — too many things with which to comply before you even got off the ground (and few tools to assist).

Much of this, I think, came about because “web services” were “plain old services” (CORBA, whatever) with a syntax — so we all knew what would be necessary, and just built it in from the beginning (along with all the other things that had been missing — see ‘second system effect’).

Although not terribly familiar with the Cloud world yet, I would hope that we learn from previous experience and allow the standards to have a relatively low cost of entry. Just as was attempted in WS-Notification & friends (I don’t know how well it worked), it would be good to have “simple” events and then standard extensions that might allow for appropriate scaling of the performance as well as the feature set as time goes on (& need develops). While I don’t mean to imply that there are lots of “small clouds”, it would be nice to define such a standard in an approachable means — a relatively small, uncluttered, core set of features whose more feature-complete-yet-complex extensions can be implemented as time allows.

Put another way, while the issues in the blog-post are undeniably required “in the fullness of time,” is there some way to define a layered standard so that one can start without providing everything?

Just a thought…

In general, I find that service management processes (as in the ITIL process framework and others) which includes service operation and others are somewhat of an after thought in the cloud. It seems the focus is on service transition or build (me a VM or something).

Whether such standards such as the WS-splat dealing with wisemen and wisdom can ever be reused is another more politically charged consideration.

Ultimately, business processes could span multiple clouds (be it public or private in-house) and non-cloud legacy things. Tying all the service management processes together could be a nightmare without such technology enablement.

Life was somewhat less automated before even SNMP and yet we still had events and management of them that happened, in that case through the evil proprietary bigco systems.

Ultimately, the success of the cloud could very well depend on its service manageability in addition to security, across the stack (IaaS, Paas and SaaS); without which it will never be leveraged in an Enterprise Architecture.