Grappling With the Gorilla

Oracle unquestionably is the 800-pound gorilla of database companies. For
more than a quarter of a century it has specialized in relational database
management software (RDBMS), building its revenue up into the $10 billion
range. According to IDC, Oracle had 39.4% market share in 2002, compared
with 33.6% for IBM/Informix and 11% for Microsoft.

But as any silverback (an older, usually dominant male gorilla with gray or
whitish hair on the back) knows, there will always be younger rival ready to
challenge for leadership. In this case it’s Microsoft SQL Server. Once
considered a lightweight database for small business, SQL Server has bulked
up and is starting to make forays into Oracle’s enterprise market territory.

This makes for exciting times. Neither Bill Gates nor Larry Ellison are
known to back down from a fight. Will Microsoft be successful in this arena,
or will Oracle slap it around like an upstart pup?

The Entourage

It’s pretty easy to find an Oracle customer. Just grab a copy of the Fortune
500 and you could probably find at least one Oracle database in most of
them. But last year was not a good one for the company. While Oracle’s
revenues turned around toward the end of its fiscal year 2003 (May 31), it
was down 2% compared to FY02, which was 12% below its 2001 revenue.

Part of this is due to an overall decline in the market. According to
Gartner, Inc., the database market overall declined in 2002, with new
license revenues dropping from $7.12 billion in 2001 to $6.63 billion last
year, a 7% loss.

“Along with many of the other IT markets, the RDBMS markets are feeling the
pressures of reduced IT spending,” said Colleen Graham, industry analyst for
Gartner’s Software Industry Research group. “However, despite this slowdown
of spending on new licenses, we find many organizations very actively
utilizing and furthering development of their major RDBMS-based
applications.”

But Oracle has also been losing market share. IBM took over the #1 spot from
Oracle in 2002 going from a 33.9% to 36.2% share, while Oracle dropped from
39.7% to 33.9%. Half of IBM’s database revenue, however, is from the
mainframe market, which it dominates. When looking at the Unix/Linux/Windows
market, Oracle is the largest vendor. Even so, according to Gartner, Oracle
lost both market share and revenue in this smaller platform market.

So, who has been attracting these fans? Microsoft. It increased its share of
the Unix/Linux/Windows market from 18.1% in 2001 to 22.8% in 2002. It also
boosted its revenue in this area by 17%, despite market shrinkage. And this
growth isn’t limited to small-scale database projects.

The State of North Carolina, for example, uses SQL Server 2000 Enterprise
Edition and Windows 2000 Datacenter Server for its North Carolina Floodplain
Mapping Program (NCFMP). Following 1999’s Hurricane Floyd’s $3.5 billion
damage, the state decided it needed to computerize its paper flood insurance
rate maps since eighty percent of the damage occurred in areas not
identified as flood prone.

The state hired Watershed Concepts (Greensboro, N.C.) as the prime
contractor for establishing the mapping system, which uses the ArcGIS
geographic information system software from Redlands, Calif.-based
Environmental Systems Research Institute, Inc. (ESRI). After eliminating
IBM’s DB2 and Informix databases as possible contenders, Watershed Concepts
came up with two possible routes: running a single 16-way Unisys Corporation
(Blue Bell, Pa.) ES7000 Orion 200 server with Windows and SQL Server or two
10-way Sun Microsystems, Inc. (Santa Clara, Calif.) E4500 servers with
Solaris and Oracle 8i.

“The goal of our due-diligence phase was to minimize any risk in meeting our
scalability and availability needs,” says Watershed Concepts’ president
Scott Edelman. “The Microsoft and Unisys solution did just that, with 40
percent lower lifecycle costs than Solaris and Oracle.”

For failover, NCFMP split the ES7000 into two eight-way partitions, each
with 8GB RAM. External storage is on a 10TB SAN from EMC Corp. (Hopkinton,
Mass.). To meet future needs, they can upgrade both the servers and SAN.

“We feel very good about our upgrade path and that we’re not building
ourselves into a corner,” says Tim Johnson, acting director of the North
Carolina Center for Geographic Information and Analysis. “When more
processing power is needed, we can easily double the capacity of our
existing ES7000 by adding another 16 processors to the chassis, moving to
two 16-way partitions to retain failover capabilities.”

The Weigh-In

For years the consensus has been that Oracle and SQL Server didn’t even
belong in the same weight class. But this perception is gradually changing
as Microsoft continues to develop its database including creating a 64-bit
version and boosting the number of processors supported up to 32. So, what
do the numbers now show about how the two match up in terms of performance
and cost?

First, taking a look at speed, the top two slots on the Transaction
Processing Performance Council’s (www.tpc.org) list of the non-clustered
database servers in early September 2003 were both held by the
Hewlett-Packard Integrity Superdome with 64 Itanium processors. The top
position was when it was running the Oracle 10G database and HPUX operating
system. The number two was the Superdome running Windows and SQL Server. The
difference between the two was less than five percent. The TPC report,
however, also showed that the price/performance for the Windows machine was
about 20% lower than the Oracle one.

Oracle, however, touts that its software is less expensive to use than
Microsoft’s when used in conjunction with packaged applications from
PeopleSoft, SAP and Siebel. Consulting firm INPUT (Reston, VA) surveyed 30
firms on their use of SQL Server and Oracle and found that the annual cost
per user for Oracle was $499.07 compared to $936.67 for Microsoft.

The Bout

So, we have a fight. Both are now capable of managing large databases. Which
of them should win your business?

Well, the numbers can only tell you so much. As the investment firms say in
their disclaimers “past performance is no guarantee of future success.”
Neither are white papers or benchmarks necessarily an indicator of how well
something will meet your own needs. Odds are, you are not going to purchase
a Superdome. If you do, then those TPC numbers cited above may figure into
your decision making. But they are useless if you are running a 4-way Dell
server.

Same thing with the cost figures. The North Carolina Flood Plain Mapping
Program and found that Windows was less expensive for its applications. The
entities cited in the INPUT study, on the other hand reported that Oracle
was half the cost per user of using Windows. It should also be noted,
however, that those companies using Oracle had, on average, five times as
many end users as those with Windows. Oracle’s lower cost, therefore, may be
due to economies of scale.

So, you can’t really rely on how well the databases performed at other times
or in other configurations. That information can give guidance, but the one
contest that truly matters is how well they perform on your own equipment
and applications. Only then can you tell whether the underdog will upset the
champ.