The
Physics
Database Services section provides
Oracle-based database services
for the Physics community at
CERN. Currently we host ≈150
Oracle 10g RAC nodes and ≈850 TB
of raw disk space to offer
database services to the LHC
experiments in the context of
the World LHC Computing Grid.
The physics community requires
high-end database services in
particular for high
availability, reliability and
performance. The current
implementation at CERN deploys
RAC on ASM, Streams and Data
Guard, following the best
practices of the Oracle Maximum
Availability Architecture.
The expected data growth is roughly ≈30 TB
per year per experiment. The experiments need to have all data
available at any time not only during the experiment lifetime
(10-15 years), but also for some time afterwards, as the data
analysis will continue. To meet this need we have to provide an
efficient way of accessing and storing the few Petabytes of
mostly read-only data. The answer to our challenge is the
compression available in ORACLE 11G Release 2 on the Database
machine.
We have tested the Advanced Compression
Option with Oracle’s Exadata I. The half-rack was located in
Reading, UK and accessed remotely from Geneva during two weeks.
It consisted of four nodes and seven storage cells with 12 disks
each. The tests focused mainly on OLTP and
Exadata Hybrid Columnar
Compression (EHCC)
of large tables for various representative
production and test applications used by the physics community,
like PVSS, GRID monitoring and test data, file transfer (PANDA)
and logging application for the ATLAS experiment. Tests on
export datapump compression were also performed. The test
results are rather impressive: we achieved 2-6X compression
factors with OLTP compression and 10-70X compression factors
with EHCC archive high.
The EHCC can achieve up to ≈3X better compression than tar bzip2
compression of the same data exported uncompressed.
Oracle Compression offers a win-win solution,
especially for OLTP compression as it shrinks the used storage
volume while improving performance.
We were invited to
present this to the Oracle OpenWorld 2009 conference and
received excellent feedback. The following talks were given:
Svetozar Kapusta
CERN
openlab