Information Technology Reference
In-Depth Information
Server operating systems. Search engines, web media, e-commerce
sites, and email systems are hosted on computers in data centers; each of
these computers runs an operating system, often an industrial strength
version of one of the desktop systems described above. Usually, only
a single application, such as a web server, runs per machine, but the
operating system needs to coordinate thousands of simultaneous incoming
network connections onto that application. Servers operate in a hostile
environment, sometimes receiving many thousands of packets per second
attempting to subvert or block the service. At the same time, there is a
premium on responsiveness. Amazon and Google both report that adding
even a sub-second delay to each web request can dramatically reduce their
revenue per user.
Server clusters. For fault tolerance and scale, many web sites are im-
plemented on distributed clusters of computers housed in one or more
geographically distributed data centers. If one computer fails due to a
hardware fault, software crash, or power failure, another computer can
take over its role. If demand for the web site exceeds what a single com-
puter can accommodate, web requests can be partitioned among multiple
machines. As with normal operating systems, server cluster applications
run on top of an abstract cluster interface to isolate the application from
hardware changes and to isolate faults in one application from affecting
other applications in the same data center. Likewise, resources can be
shared between various applications on the same web site (such as Google
search, Google earth, and gmail), and resources can be shared between
multiple web sites hosted on the same cluster hardware (such as with
Amazon's Elastic Compute Cloud).
1.3.6
Future operating systems
Where do operating systems go from here over the next decade? Operating
systems have become dramatically better at resisting malicious attacks over the
past decade, but they still have quite a ways to go. Provided security and
reliability challenges can be met, there are huge potential benefits from having
computers tightly control and coordinate physical infrastructure such as the
power grid, the telephone network, and a hospital's medical devices and medical
record systems. Thousands of lives are lost annually through trac accidents
that could potentially be prevented through computer control of automobiles. If
we are to rely on computers for these critical systems, we need greater assurance
that operating systems are up to the task.
Beyond mission critical systems, whenever the underlying hardware changes,
there is work to do for the operating system designer. The future of operating
systems is also the future of hardware:
Very large scale data centers, coordinating hundreds of thousands or even
