Page 154 - 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

without them, and they will find that if their application en-
ters an infinite loop, they have little choice but to shut it down
and commission another server to try again. If a server com-
ponent fails and the server grinds to a halt, so does the work-
load. But unless you have a service-level agreement, Amazon
is not responsible. The answer to the problem, its technical

support tells users, is for the customer to learn how to build
failover into the application, then rent a second server in EC2
to stand by as the failover destination if something happens to
the first one. All of this is fair for a supplier that is providing
only infrastructure, unless you’ve purchased an SLA that says
otherwise.
Mastering this distribution of responsibility will be one

challenge. Another will be meeting the CFO and CEO’s ex-
pectation that the cloud is going to drive down the cost of their
computing infrastructure. The CFO understands that the
cloud supplier provides the capital outlay for equipment, so
he’s looking for reduced capital expenses while keeping op-
erating expenses in hand.
One way to do that is to develop the capacity management
skill of “cloudbursting,” a name for resorting to cloud resources
as a way of offloading spikes of activity that would otherwise

tend to require more servers. Cloudbursting offloads a burst
of activity to an infrastructure provider, where the cost of pay-
ing for a few hours of server time several times a year is far
lower than that of purchasing and configuring more servers.
Cloudbursting is a catchy name, but it carries its own risks.
What if the cloud server fails to start as a spike of activity hits?
Will that jeopardize your company’s service-level agreements

134