I’m not saying that Microsoft should have used LAMP or anything else, but that their architectural choices resulted in a very expensive system and that a different architecture on the same platform could likely have achieved better performance at a lower cost.

And I do prefer Lambos

RE: “I’ll concede that it’s not an apples to apples comparison”

Won’t it be more practical to compare on some other metric which make more sense. Myspace works on a different business model when compared to PlentyOfFish.com. facebook and myspace have more in common and hence the large number of servers.

also, what else was microsoft suppose to use for myspace? LAMP?

apart from that I can’t think of any use for these figures unless u are saying that Microsoft’s platform are too expensive and useless for large scale development?

People who can afford microsoft or oracle will always go for it in my experience no matter what you tell them. People who love LAMP will use it forever. It like preferring a Lambo over Ferrari or Porsche.

Sorry for the tangent but I was expecting something more from the article (at least a M$ bashing would have been nice!).

Imran

I’ll concede that it’s not an apples to apples comparison.

> I can’t really think that putting up so many servers costs so much.

It’s not just the licenses. It’s the power/cooling, administration, total cost of ownership thing. The fact that Microsoft is the one doing MySpace effectively makes the licensing costs zero – quite different than for other companies. The whole deal still has to be quite expensive.

Thank you for your response, yes it answered my question.

Anything less just doesn’t seem realistic to me. Who would put out 500 DB Servers if each and every one of them costs 8 a full SQL2008 per-proc license.. Assuming you get the license at the ultra low price of 1k you’re still paying 4 million only for SQL server licenses.

greetings Daniel

Right – forgot about Plenty Of Fish.

Thanks for your kind words.
I’ve been keeping an eye on CCR and similar MS developments. I wouldn’t recommend using it in the path of a given web request.

While it can be interesting for certain in-process multi-threaded high-performance scenarios, that would be best used as an implementation detail under higher-level frameworks.

I have one client that may have been well served by using some of its features under NServiceBus for their low-spec devices.

Hope that answers your questions.

http://highscalability.com/plentyoffish-architecture

Has 2 Web Servers and 3 Database Servers.

Myspace on the other hand has 1.5 billion page views a day.

http://highscalability.com/myspace-architecture

Has 4500 web servers, 1200 cache servers, 500 database servers

Facebook however is meant to have something like 14,000 web servers, and 2500 database servers. (Personally I think facebook performs like crap. So much AJAX, failing all the time.)

By your experience would you even consider using CCR? If so do any particular scenarios come to mind?

Always love reading your blog and your articles, you are one of the most insightful guy that I know .Net or otherwise.

The client was using those patterns as well as doing the “thank you page, and email back” model.

Was this the same client (60 million users) and architecture as described in the post below?

http://www.udidahan.com/2008/07/30/scaling-long-running-web-services/

Simon