Jon kicks off the show by asking Udi if one can run a high-availability and high-scalability site (like stackoverflow.com) on two servers with ASP.NET? Udi asks “how high is high” and talks about the importance of defining the scope of the architectural problem.
Udi comments on the how site behavior (high reads or writes, static or dynamic content, etc) dictates the way availability and scalability concerns can be handled.
Udi talks about how websites can be broken into fragments and how one can take advantage of content distribution networks and caches.
Udi stresses that high-availability inherently makes scaling more difficult and how scalability is a multidimensional cost function.
Kevin asks about common scalability mistakes and solution patterns. Udi replies with talk of SOA, web services, REST, one-way messaging, pub/sub and offloading things to the background in an asynchronous way.
Kevin follows up with additional questions: Why don’t more developers turn to messaging patterns to solve their scalability issues and can one introduce scalability changes incrementally?
Udi discusses the scalability patterns employed by RIA Services for Silverlight and gives his opinion on Azure as a scalability system.
Scott K asks about geographical distribution and Udi gets to the heart of the matter – high-availability means writing data to multiple places which means you are slowing your system down.
K Scott asks Udi about the AppFabric service bus provided by Microsoft and how it compares to already available distributed cache solutions like .NET Service Bus, MassTransit, and NServiceBus.
This entry was posted
on Sunday, January 10th, 2010 at 1:00 amand is filed under interview.
You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.