The Office of the Comptroller of the Currency has created an information architecture to help it build a data warehouse, one data mart at a time, over the next several years. The warehouse has to be built as a multimart environment, the office's senior data analyst Vance Kane said. "We have different lines of business—financial institutions, human resources. To have one database warehouse is fantasyland from my perspective," he said.
The Office of the Comptroller of the Currency has created an information architecture
to help it build a data warehouse, one data mart at a time, over the next several years.
The warehouse has to be built as a multimart environment, the offices senior data
analyst Vance Kane said. We have different lines of businessfinancial
institutions, human resources. To have one database warehouse is fantasyland from my
perspective, he said.
Yet, OCC did not want independent islands of information either, added
information services director Charles Wright.
Unless the offices plan combined the disparate data, it would not make any
improvement over the legacy database systems OCC now has. So the office is planning a data
warehouse that links data marts customized to meet the needs of particular user groups.
OCC will build the various data marts over several years. We cant wait and
roll out the data warehouse all at one time. It has to be done incrementally, Wright
To achieve its goal, OCC assembled a data architecture plan that requires that the
office standardize information before entering it in a data mart. For example, OCC chose
one day to be the basic time measure, Kane said. Though one day may be too small a measure
for data that is updated quarterly, OCC needed a measure that would work for everyone, he
If queried on the same question, each data mart must give the same answer based on the
time period given, Kane said. In the past, two databases would give different answers
because they were using different units of measurement or were basing the answer on
But with the caveat, based on time period given, OCC has the leeway to keep
to the standard even though one data mart may not be as up-to-date as another, Kane said.
To get the data into the same format, OCC set up what it calls a data warehouse staging
area. It is here that the office converts and standardizes data for inclusion in a data
mart, Kane said.
The warehouse environment is the factory for the data mart, he said.
So far, OCC has built one data mart for its U.S. and Foreign Branches Division and has
begun work on a second for its Country Exposure and Transfer Risk Division.
Before the data mart, the U.S. and Foreign Branches Division had to go to four
different systems get the same data, and field examiners lacked access altogether,
division director Jose Tuya said.
We are a small, self-contained unit but have all these sources of data; we made
an ideal test case, Tuya said.
OCC has settled on Microsoft SQL Server and IBM DB2 software for its data marts. Users
will access data housed on Compaq 1850R, ProLiant 6500 and ProLiant 7000 servers running
Microsoft Windows NT.
The first data mart is using SQL Server, but for data marts targeted at a larger
audience we would use DB2, Kane said.
The selection of front-end software is a good example of OCCs effort to meet
users needs. For the first data mart, OCC used BusinessObjects decision support
software from Business Objects Inc. of San Jose, Calif.
But OCC economists may want software that will provide more sophisticated statistical
analyses, Kane said. One possibility is the Statistical Analysis System from SAS Institute
Inc. of Cary, N.C., he said.
The data mart documentation and source data definitions are stored in a warehouse
repository, separate from the front-end software, Kane said. Keeping the information used
by all data marts separate from the front-end software will reduce retyping and give the
office the freedom to use other software for future data marts, he said.
The office also wants the flexibility to change part or all the equipment and software.
It could be conceivable that we may need a larger platform than SQL Server or
DB2. From talking to other organizations, [we know] data warehouses tend to grow very
fast, Kane said.
One likely change will be to make the data marts accessible via an intranet. Using an
intranet looks and feels similar to browsing the Internet, and most people like the
Internet, Kane said.
For now, OCC will wait for Web server software products on the market to mature a bit.
Web software, in general, is behind client-server software in the capabilities it offers,
It is possible that the move to an intranet will delay the debut of a mart for
OCCs Country Exposure and Transfer Risk Division.
The first mart, for the U.S. and Foreign Branches Division, has its users concentrated
in the Northeast, and the second marts users are spread throughout the country, said
Frank Carbone, a senior adviser in the International Banking and Finance Division.
Setting up a data warehouse, sounds easy, but as you work your way into it, there
are stumbling blocks, Kane said.