Comments on: So, how many machines/CPUs do we need? http://www.udidahan.com/2007/02/12/so-how-many-machinescpus-do-we-need/ Enterprise Development Expert & SOA Specialist Sun, 14 Mar 2010 06:27:00 -0500 http://wordpress.org/?v=2.8.4 hourly 1 By: » [Podcast] Space-Based Architectures for the Web http://www.udidahan.com/2007/02/12/so-how-many-machinescpus-do-we-need/comment-page-1/#comment-1295 » [Podcast] Space-Based Architectures for the Web Wed, 11 Apr 2007 05:28:05 +0000 http://udidahan.weblogs.us/archives/377#comment-1295 [...] So, how many machines/CPUs do we need? [...] [...] So, how many machines/CPUs do we need? [...]

]]> By: Guy Nirpaz http://www.udidahan.com/2007/02/12/so-how-many-machinescpus-do-we-need/comment-page-1/#comment-466 Guy Nirpaz Wed, 14 Feb 2007 04:48:20 +0000 http://udidahan.weblogs.us/archives/377#comment-466 Udi, Nice writeup! My two cents: Most IT systems should create value by providing business logic that operates on business data. Since most current architectures are client/server or tier based, if you measure the value generating code vs. mandatory tier communication/serialization/data mapping you'll find an avg. of 70% for non-value code and 30% (hopefully) for code that actually creates business value. Why is that? The problem is in the middleware! My suggestion is to change the perspective and look at the application as a whole. In a scale-up scenarios look at each processing unit (partition of the application) as a complete application that can do all business value and not partition by technical value, you'll get to a complete new paradigm. The first step is to partition by data elements and not by technical elements. This is the essence of SBA. You can easily do this with <a href="www.gigaspaces.com" rel="nofollow">GigaSpaces</a>, but others like Ebay and Google do this by themselves. Udi,

Nice writeup!
My two cents:
Most IT systems should create value by providing business logic that operates on business data. Since most current architectures are client/server or tier based, if you measure the value generating code vs. mandatory tier communication/serialization/data mapping you’ll find an avg. of 70% for non-value code and 30% (hopefully) for code that actually creates business value.

Why is that? The problem is in the middleware! My suggestion is to change the perspective and look at the application as a whole. In a scale-up scenarios look at each processing unit (partition of the application) as a complete application that can do all business value and not partition by technical value, you’ll get to a complete new paradigm. The first step is to partition by data elements and not by technical elements.

This is the essence of SBA. You can easily do this with GigaSpaces, but others like Ebay and Google do this by themselves.

]]> By: Regu http://www.udidahan.com/2007/02/12/so-how-many-machinescpus-do-we-need/comment-page-1/#comment-454 Regu Tue, 13 Feb 2007 09:46:11 +0000 http://udidahan.weblogs.us/archives/377#comment-454 Udi, In a web scenario, how do you correlate multiple events in a SBA that have been created out of a single request? Also, how would you ensure that the response to the user is synchronous and meaningful? Of course we may use a correlation ID across the events. In this case will the request thread wait until all events for the request are processed? Udi,
In a web scenario, how do you correlate multiple events in a SBA that have been created out of a single request? Also, how would you ensure that the response to the user is synchronous and meaningful?
Of course we may use a correlation ID across the events. In this case will the request thread wait until all events for the request are processed?

]]>