Company Collaboration Efforts Date
NASA R&D; Google provide funding for a new joint R&D center September 2005
America Online AOL adopts Google’s search engine; Google advertises in AOL network December 2005
Bearing Point Enterprise search technology services with unstructured data February 2006
Source: Various news sources
The focus of information technology at Google for both software and hardware is speed and cost.
These two metrics are valued more than any other criteria such as reliability of machines or highperformance
enterprise computing hardware. Ultimately, the result must transform a response
time of user query using Google’s search engine to be completed within a one second time-frame.
Started in Larry Page’s dormitory room, the information technology at Google has transformed
into a full-blown large cluster PC network that functions similar to a computing grid.iv Even
though information technology infrastructure has changed dramatically over the years, the model
of IT use at Google has stayed the same. This model follows the original principles adopted by
the co-founders of building a prototype system that uses commodity hardware and intelligent
software. The shift of computer industry with PCs becoming commodity electronic hardware
over the years has worked in favor of Google’s IT strategy in getting the best cost performance
ratio (Patterson & Hennessy, 2004). Thus, instead of purchasing the latest microprocessors,
Google IT performs calculations to look for the best value of processing power per dollar and
purchasing many PCs that are only a few months old in the market, but at a much lower
discounted price. This is suitable for Google because the framework of their search engine is
built around parallelizing many user query requests across multiple machines and if more
processing is required, the system can simply increase more machines to serve even greater user
requests. The overall price per performance is more important than individual peak
performances, and this enables Google to achieve superior speed at a fraction of the cost rather
than using a few, but expensive high-end server systems. The end equation for Google’s IT in
selecting machines is calculated by the cost per query, and is derived by the sum of capital
expenses and operating costs divided by performance. For accuracy, the calculation takes into
inherent effects due to hardware depreciation and maintenance repairs. At the data centers, the
primary cost factor is capital expenditure credited to hardware, followed by personnel and
hosting costs (Barozzo, et al., 2003).
No comments:
Post a Comment