Home|Products|Doc|Community/RZB|Company
With a balance sheet total of € 156.9 billion as of 31
December 2008, Raiffeisen Zentralbank Österreich AG (RZB) is the third-largest
bank in Austria
and one of the country's leading commercial and investment banks. It was
founded in 1927 and is headquartered in
RZB considers itself in particular a partner in Central and Eastern
Europe (CEE). Via listed subsidiary Raiffeisen International
Bank-Holding AG, RZB
operates one of the leading banking networks in CEE. 17 markets of the region
are covered by subsidiary banks, finance leasing companies and a number of
other financial service providers. Over 14.7 million customers are attended to
through more than 3,200 business outlets.
The RZB Group is also represented with branches and offices in a number of
international financial centres and in the growth markets of Asia.
High Performance Value at Risk (VaR) aggregation
based on VectorNova's VectorSTAR
Background
Primarily, RZB had
the challenge to aggregate dynamically 15 billion net present values (NPVs).
The first tests on
midrange servers and with traditional standard relational database management
systems (RDBMS) resulted in approx. 8 hours run time.
It has been soon recognised
that spending more on faster hardware will not solve the problem of the long
run times sufficiently but will increase hardware invests and system software
maintenance costs dramatically.
Therefore, RZB
looked for new technologies running on standard hardware and detected
VectorSTAR by VectorNova.
Execution / Timeline
a) Evaluation Phase (May 2008 – July 2008)
Two different approaches for the Value at Risk (VaR) aggregation based
on a test set of 5 billion net present values representing 100.000 deals with
5.000 scenarios were evaluated.
In order to compare the performance of the different approaches the run
time for the VaR aggregation a random portfolio of 1.000 deals out of the
complete set of 100.000 deals was compared.
VectorSTAR
approach – run time of less than 2 minutes
A prototype implementation was provided by VectorNova. The test was run under
Linux on an off-the-shelf standard hardware with two CPUs with 2 cores each and
in total 64 GB RAM.
RDBMS
approach – run time of 2 hours 46 minutes
A prototype implementation was provided by RDBMS vendor. The test was
run on hardware comparable to the VectorSTAR test.
Since the VectorSTAR approach was drastically faster than a RDBMS
approach a VaR aggregation implementation based on VectorSTAR was decided.
b) Implementation
Phase (July 2008 – August 2008)
The VaR aggregation implementation based on VectorSTAR
was done during 2 weeks.
Since off-the-shelf standard hardware can be
used for VectorSTAR the total cost for a) the hardware in two computer centres
and b) VectorStAR licence is rather low (ca. 50 K€).
During fall 2008 VectorSTAR has been encapsulated
as a web-service which can be linked to any other application in the bank for
statistical analysis or vector calculations.
Successes / Benefits
In RZB's ‘Value at
Risk (VaR) aggregation’ project VectorSTAR enabled the bank to:
·
implement
VaR aggregations in a 2 week implementation time frame to run VaR aggregations for
15 billion net present values (NPVs) in less than 10 minutes – this is 100 times faster
compared to traditional implementations
·
to use
off-the-shelf standard hardware components and thereby reduce the sum of software
and hardware investments and maintenance costs (for a comparable high performance VaR
aggregation) by 100
·
to provide a web-service
that calculates VaR aggregations for an arbitrary portfolio of a couple of
hundred of RZB’s deals and
·
to
setup a high performance and extremely flexible platform for all kind of vector and statistical
calculations in the bank
Outlook
RZB is planning to
integrate the statistical manipulation language R into the VectorSTAR
web-service.
