By Bill Moran, Rich Ptak,
The advertisement (at right) appeared on the front page of the May 23rd
issue of the Wall Street Journal (WSJ). We confess that we envy the budget that
Oracle has for placing these ads in the WSJ. As you can see, the ad attacks
IBM’s Power 7+ systems.
Obviously, Oracle believes the cost of a front page ad in
the WSJ is money well spent, especially so if they convince the paper’s readers
that Oracle solutions are both more powerful and cheaper than IBM’s solutions.
This appears to be a bit of a stretch for Oracle as in our experience
their software users do not associate the company with low prices, especially
in connection with an Oracle database.
That said; let’s move on to examine the ad details. First,
note that the performance figures are given without any attached metrics.
Certainly, Oracle’s 27, 843 number (highlighted in red) is much larger compared
to IBM’s 10,902 (also in red). But what is being measured? The actual metric is
given in the fine print at the bottom of the ad. It is SPECjEnterprise2010
EJOPS. An executive reading the paper should reasonably ask: what does an EJOPS
number have to do with my IT operation? The answer: “Not Much”.
The benchmark result quoted by Oracle is an industry
standard benchmark created and maintained by SPEC[1]. However, the benchmark
focuses on Java operations. We are at a loss to identify any real world
installation operation that relates very closely to the benchmark. Thus, the claim that is made in the headline
of the ad, namely that Oracle gets 2.6x better performance, is limited to this particular
benchmark environment. And, it cannot be generalized to any other environment.
It is also worth pointing out that the body of the ad makes
it appear that the benchmark applies to IBM’s Power 7+ and Oracles SPARC T5.
However, reading the fine print (at the bottom) you find that the benchmark
only compares a single, specific IBM system, the IBM 780 to one specific Oracle
system, the T5-8. Therefore, the results do not apply across the board to all
Power7+ systems or even all T5 systems.
Finally, we look at the cost quoted in the ad, namely
$805,000 for IBM versus $299, 000 for Oracle. The SPEC benchmark[2] upon which the ad is
ostensibly based does not include any cost. Therefore, the dollar figure which
appears to be associated with the benchmark is, in reality, totally unrelated
to it[3]. It is just plugged in by
Oracle.
So, let’s just stop for a moment and analyze the number.
This appears to be the price just for the server[4] without any software, such
as the operating system, database, etc. or necessary hardware (like storage,
network, etc.) It is frequently the case for Oracle systems that the software
represents the major (and largest) part of the cost. Therefore, without knowing
the prices of these additional items, the cost comparison is not informative.
What Oracle is doing is roughly equivalent to comparing automobiles by only
looking at the cost of the engines. No one would accept such a comparison as
valid.
We devoted this time to parsing Oracle’s ad because of where
it appeared. In general, WSJ readers are unlikely to be familiar with the
technical details of a SPEC benchmark and how to interpret them. We thought it
worthwhile to provide some additional insight into what is and isn’t being
said.
Now, Oracle may actually be able to put together a credible
case of its competitive advantage against IBM. But, this ad definitely does not
make that case. However, if Oracle was only trying to attract the attention of
executives reading the WSJ, they will likely see this ad as a success. We like
to think this analysis helps those executives to put these claims in proper
perspective.
[1]
None of our comments are meant as criticisms of SPEC. They do an excellent job
of maintaining a diverse set of benchmarks. Details about their benchmarks are
on the website: http://www.spec.org/jEnterprise2010/results/.
[2]
All of the details of the benchmark are on the SPEC website: http://www.spec.org/jEnterprise2010/results/. It is also worth noting that Oracle did not quote
the best IBM result.
[3]
Some industry standard benchmarks, such as TPC do require that the vendor
provide a price for the benchmark system so that price/performance can be
calculated.
[4]
IBM informed us that the pricing for the configuration in the ad does not appear
on any IBM website.
No comments:
Post a Comment