How Google Works
Baseline Magazine has an excellent summary on how Google manages their information technology, suggesting that businesses in general can learn from the Google approach. Google, in case you're not aware, is very secretive of the details of their IT investment and strategy. It is, after all, their bread and butter -- the key differentiator that keeps them one step ahead of Yahoo and Microsoft, in the ability to deliver relevant searches, instantly. Baseline has put together a pretty good story however, based on what little has been publicly released by the company -- and what has made its way out via analysts, former employees and Google's own presentations to academia. Google is apparently more open to academia -- probably because of their roots -- and probably because they recruit a vast number of new employees directly from schools.
We all know of Google's prowess in the search arena -- but what of the technology that Google uses to run their own business? Well, there are a few best of breed applications that Google has purchased -- most notably, they use Oracle's Financials. However, quite the majority of Google's technology is built in-house. Everything from their infrastructure, to the software they use. In fact, the same vast network that Google employs for their search and other ancillary businesses, is leveraged for running their own business. As Douglas Merrill of Google puts it, "We're about not ever accepting that the way something has been done in the past is necessarily the best way to do it today."
Some of Google's technology of note that separates them from pack:
What a concept. Does it really have to be that revolutionary? How many systems are you aware of that requires a precise process to leverage it? How many go strange when an enduser does something unexpected -- if it responds at all?
Updated: July 30, 2006
We all know of Google's prowess in the search arena -- but what of the technology that Google uses to run their own business? Well, there are a few best of breed applications that Google has purchased -- most notably, they use Oracle's Financials. However, quite the majority of Google's technology is built in-house. Everything from their infrastructure, to the software they use. In fact, the same vast network that Google employs for their search and other ancillary businesses, is leveraged for running their own business. As Douglas Merrill of Google puts it, "We're about not ever accepting that the way something has been done in the past is necessarily the best way to do it today."
Some of Google's technology of note that separates them from pack:
- Google's Infrastructure -- built from cheap off the shelf PC components, stitched together with velcro and software to operate as one gigantic supercomputer that spans the globe via the internet. Google apparently has the ability to rapidly deploy prefabricated data centres anywhere they own access to fibre. Google's infrastructure is built with the explicit assumption that the components -- entire servers -- will fail. Software makes sure that endusers don't notice.
- PageRank -- the secret sauce of finding the right information on the internet, that was developed by Larry Page when he was still in school. Stanford apparently holds the patent for PageRank, and will have the ability to license it in 2011. Google has made modifications to PageRank since its initial development -- which is probably a good thing, as Page published his ideas, including the formula, in a paper while at Stanford. The original secret has been out for sometime now.
- BigFiles and the Google File System (GFS) -- BigFiles was originally developed by Google's founders as a virtual file system to stitch together the storage from all the individual servers to make one giant pool of storage. It has been replaced by GFS, developed by Google Labs. GFS basically makes sure that there is redundancy in Google's system, and manages traffic on the network.
- Big Table -- Google's own database management system, which stores structured data used by Google's applications.
- MapReduce -- developed to take the load off developers coding for a distributed architecture. MapReduce takes execution code, breaks it up into chunks and parses it out to processors on the Google network. This allows developers to not worry how their code will execute on the network. It just does. Quickly.
What a concept. Does it really have to be that revolutionary? How many systems are you aware of that requires a precise process to leverage it? How many go strange when an enduser does something unexpected -- if it responds at all?
Updated: July 30, 2006
Andrew Whitchcock has an overview of Google's BigTable, from a presentation by Jeff Dean at the University of Washington. Check it out!
Comments
Post a Comment