Database Performance
Situation: 1.2 TB database was crashing twice a
week. Nightly schedule could only be run every
third day because it took 46 hours to run. No
documented understanding of database contents.
Action Taken: Database engine was tuned, data model was
redesigned, major processes were redesigned and re-written and
database was re-laid out with performance and stability in
mind. Periodic cleansing of the database was instituted
and database documentation was generated.
Lasting Effect: A nightly cycle which
runs in 8 hours. A database that is
half the size of the original while containing
three times as much information. No
engine crashes in a six month period and
appropriate documentation for every table in
the system. In addition, new utilities were developed to aid in the identification of
newly added tables.
A Failed Project
Situation: After 18 months of development the project team was on the verge of delivering a
400 GB database system with a 76 hour nightly cycle, a 105
day monthly cycle and a
yearly cycle projected to run across an undetermined number of
weeks. The system did not perform some of the basic tasks for
which it was originally designed and there was no consistent
software build process resulting in fixed bugs showing up in later
releases of code.
Action Taken: In three months, both database and engine were tuned.
All non-performing processes were re-written and a configuration
control process was put in place to ensure consistent software
builds.
Lasting Effect: A nightly cycle that
runs in 2 hours. A 7 hour monthly cycle and
a 15 minute year-end cycle. System could
also be consistently rebuilt while meeting all
planned business requirements.
|
