ServicesResourcesConferencesOur TeamWeblogsAboutContact
   

Subscriptions

Post Categories

News

    MicrosoftBizTalkBlogs

Affiliations

Archives

Rich on Volta's tier splitting: Halleluja!

Usually, I do not post just "+1"... but this time - and I have been fighting with myself not to rant too much about Volta - I am 100% with Rich.
I just think there is too much "Oh, that is easy: just one attribute and a switch and I'm done"-mantra in there...

Well, keeps me busy :)


posted on Thursday, December 13, 2007 1:28 PM

# re: Rich on Volta's tier splitting: Halleluja! @ Saturday, December 15, 2007 4:34 PM

Hi Christian,

well, haven't taken a closer look at Volta yet, but the direction MSFT is taking is clearly the RIGHT way to go.

Same situation with all the new products. Be it Workflow Foundation, LINQ, Volta: Abstract at a higher level (in this case a fairly high level), let the guys in the wild concentrate on the business logic ONLY. Then some domain specialists build a runtime for this which includes the necessary smarts (optimizations) to make the stuff performant. And this includes fancy stuff like runtime optimization strategies, parallelising stuff etc.

The beauty of LINQ is not that it is less code than the datareader stuff. And currently you give up a small bit of your performance (well, Rico said they got LINQ performing at 96 percent of the manual data reader implementation).
The beauty is that the guys at MSFT who build the "runtime" can throw all the strategies they want at it - parellising queries, caching, virtualisation yada yada. Same with Volta. The interface gives you just enough freedom to specify the business logic (which is what you are getting paid for anyway).


Funny that the object-oriented/Service Oriented stuff has taken so long (well, still a long way to go) to reach the same level of abstraction which SQL always had: Specify what you want, not how to do it. Let the smart guys decide HOW to do it. Most of the time, they'll do it better than you. SQL query processing and Garbage Collection are two beautiful examples of how a high level of abstraction actually achieving performance benefits in most of the cases.
Sure, if you code it manually in Assembler, build your own memory management and everything, you may outperform the "easier" stuff. History will tell you that the human brain is simply too small for that, and that we always screw things up. Well "never try to outsmart the system".


Sorry, but you are wrong this time. Just wait and see what the future brings. Volta might in its current shape not survive, but the direction will be: do just the business logic, without thinking about where it runs, how it runs, which calls it makes, how it caches.

That has always been the MSFT direction.

Now I can understand you are a smart guy who prides himself (and presumably enjoys learning how stuff works big time) in knowing the internals and is able to do a textbook approach on building complex systems. 99 percent of your colleagues are not.

Cheers.

Brian


Powered by Community Server, by Telligent Systems