Google Search Appliance and Outsourcing


Infoworld reports that Google has released a bigger, beefier version of their search appliance.   The new search engine extends "the search capabilities to 3 million documents and 150 queries per minute."  We probably don't need 150 queries per minute, but we definitely have that many documents or more.  The new appliance also uses a clustering approach to HA.  This sort of support needs to be added to our web infrastructure platform that McNamee is working on

We use an internally developed search engine right now.  Its actually a great piece of work, but we have a tough time keeping up due to all the other work that needs to get done.  I'm often asked if I'm a supporter of outsourcing.  The answer in general is "no."  I think IT is too important to an organization to turn it over lock, stock, and barrel to an outsider.  I'm am, however, in favor of "oursourcing" any internal development that can be replaced by off the shelf purchasing so that those resources can be turned to other important activities.

I learned this lesson the hard way in a prior life when we'd built a large system based on CORBA before application servers were widely used or understood (at least by me).  We spent a lot of time working on distribution code, load balancing, error recovery, threading, etc.    After a while we just retooled the system, bought Weblogic and solved most of our headaches.  In the end, there was no way a handful of guys in my shop could come close to keeping up with the development staff Weblogic had placed on the problem and the price of the software was well worth having my guys working on product development rather than infrastructure tweaking.