Page 66 - How Cloud Computing Is Transforming Business and Why You Cant Afford to Be Left Behind

MANA GEMENT STRATEGIES F O R THE CL OUD R EV OL UTION

the software governing the cluster. Software can supply fault
tolerance, or the ability to cope with the failure of any single
component, by routing work around that failure and redoing
it in a separate component within the cloud.
Fault tolerance has always been expensive to solve in hard-
ware. Google uses the cheapest highly reliable parts and lives

with failures as they occur. Corporate data centers pay a high
price for the reliability they buy in hardware, but it’s worth it
to them not to have critical business systems go down. Solving
the problem in software makes it a simple task to add a hard-
ware server without a lot of advance preparation. In the un-
likely event that the server fails early in its normal life span,
the cluster can cope.

Yahoo! recently offered a more specific example of fault
tolerance in software that it has implemented in its cloud-style
database system, Hadoop, running on clusters of up to 4,000
servers at a time and 25,000 servers in all. Hadoop sorts
through data gained from each active site on the Web. With-
out Hadoop, sorting the results of a complete Web crawl took
so long that new sites appeared on the Web before the task
could be completed, and the results were badly out of date.
With Hadoop, the indexing of the Web can be done in 73

hours, so a Yahoo! Web search is only three days behind the
actual state of the Web.
Google, Amazon.com, and now Yahoo!, as it embarks on
its own use of cloud data centers internally (a private cloud),
have all adopted a design principle from the Internet itself.
The Department of Defense, through its Defense Advanced

46