For more 10g new features
information, you may want Mike Ault's best-selling book
Oracle Database 10g new features. It's 30% off at this
link and you get immediate access to the working 10g scripts.
Ellison on Oracle 10g Enterprise Grid Computing
by
Donald K. Burleson
2015 Update:
Please see these 2015 updates to Oracle hardware
architectures and
the
costs and benefits of server deconsolidation.
Introduced with much fanfare, Oracle CEO Larry Ellison last week
revealed his vision for the future of database management and
Oracle's plans to intercept the changes. Ellison began his keynote
with his view of the history of how computer technology has changed:
"Back in 1964 nobody had heard about the Beatles ... Elvis was the
king of rock-and-roll ... It was a different world back then". As I
recall, in 1964 the Beatles were taking the world by storm, but we
can forgive Larry for this little mistake. His point was that
technology has changed all aspects of life.
When
listening to Ellison, we get the impression that he spends a great
deal of his time speculating on the future of technology. Oracle
always strives to be proactive and predict the next big movement,
often with mixed results. For example, Ellison was far ahead of the
curve when he predicted the advent of Web appliances back in 1996.
His attempt to sell Web appliances was far ahead of it's time, and
while his prediction that the future of PCs was quite correct, the
market wasn't ready. Today, we can clearly see that the future of
PCs is with all application software being located in cyberspace.
We
also have to remember that the market does not always move in
exactly the same direction as Oracle has predicted. In 1997, Oracle8
was introduced with the promise that object-orientation was the wave
of the future. While history will show that Ellison was correct and
Java and C++ have become a de-facto standard, Oracle was
surprised to see that object-oriented databases were not embraced by
the market.
Given
his mixed history of brilliant insights and a vision that is
sometimes too far ahead of the times, Ellison should best be viewed
as a visionary with a firm stake in the future of computing. And
having slipped from his position as the fourth richest person in the
world, he has a vested interest in ensuring that his new Oracle10g
products are embraced by the market.
Ellison began his presentation with his version of history and the
current market. He mocked Microsoft's TPCC database benchmark on a
64-CPU Windows server, claiming that Microsoft is myopic by using a
large mainframe-like server. When discussing a visit by Microsoft to
IBM labs, Ellison said: "They should have turned left into IBM
research, but instead, they turned right into the IBM museum."
Ellison's disdain for large, monolithic servers was justified by
noting that mainframe-like systems have a single point of failure
and are constrained b 128 CPUs. "Big computers have limited
capacity," said Ellison, predicting that 128 processors would not be
sufficient for the largest processing systems.
"Computer systems are outgrowing server capacity," he continued,
predicting that the future lies in the scale-out idea of Enterprise
Grid computing. Ellison acknowledged that his "Enterprise Grid"
concept is quite different from the traditional view of grid
computing and he went into great detail to distinguish his vision of
grid computing from Scientific Grid computing (in which many
computers are used to solve massively parallel problems).
Enterprise Grid computing gives the illusion of being a single
computer, said Ellison, and he predicted that companies will soon
embrace the low-cost 64-bit Intel server blades. Ellison also noted
that the Intel processors are the fastest and cheapest available:
"If you want the fastest processors, then they will be forced to pay
less." said Ellison, "Intel processors are the fastest in the
world."
In
Ellison's vision of Enterprise Grid computing, low-cost Intel server
blades can be added and subtracted from the grid as-needed,
providing "unlimited performance and capacity." He also promised
that the grid monitoring software will provide automatic load
balancing, just as is found in a traditional SMP server.
Ellison also noted that Oracle 10g is becoming more automated, and
less manual work will be involved in installing, maintaining, and
tuning Oracle10g. He noted a new self-managing storage system, a
self-tuning instance, and self-tuning SQL as new components that can
reduce the manual workload. Ellison suggested that, with Oracle10g,
companies could save millions of dollars by cutting their DBA staff,
but he did not dwell on that new feature (a wise decision, since
many of the attendees at his keynote were Oracle database
administrators).
For
now, many Oracle professionals are left to ponder the costs of grid
computing versus the large, monolithic systems offered by the server
consolidation vendors. One such vendor was quite upset by Oracle's
claim that mainframe-like configurations have a single
point-of-failure, noting that many of the large 16- and 32-CPU
servers have built-in fault tolerance and will never suffer a
catastrophic failure. The critics also note that Oracle's own
failover products (DataGuard and Oracle Streams) can be made to
provide 100% fault tolerance.
The
market is split between the idea of server consolidation (moving
many instances onto a single SMP server) and using server blades in
a loosely-coupled configuration. However, it is clear that the
Oracle10g Grid movement is economic, and that hardware savings are
the driving force. The server consolidation experts say that a
back-to-the-mainframe approach provides the best resource sharing,
and without sophisticated software to allocate and de-allocate
blades. The grid proponents maintain that server blade technology
will provide the most scalable and flexible solution.
For
now, it appears clear that few companies are going to replace their
hardware with Intel 64-bit servers until their aging systems are no
longer capable of serving their loads. So it may be a few years
before we will know whether Larry Ellison's vision will be widely
adopted in the marketplace.
-----
