USENIX Technical Program - Abstract - USENIX Annual
Conference, General Session - June 2000
Gecko: Tracking A Very Large Billing System
Andrew Hume, AT&T Labs - Research; Scott Daniels, Electronic Data
Systems Corporation; Angus MacLellan, AT&T Labs - Research
Abstract
There is a growing need for very large databases which are not practical
to implement with conventional relational database technology. These
databases are characterised by huge size and frequent large updates;
they do not require traditional database transactions, instead the
atomicity of bulk updates can be guaranteed outside of the database.
Given the I/O and CPU resources available on modern computer systems, it
is possible to build these huge databases using simple flat files and
simply scanning all the data when doing queries. This paper describes
Gecko, a system for tracking the state of every call in a very large
billing system, which uses sorted flat files to implement a database of
about 60G records occupying 2.6TB. This paper describes Gecko's
architecture, both data and process, and how we handle interfacing with
the existing legacy MVS systems. We focus on the performance issues,
particularly with regard to job management, I/O management and data
distribution, and on the tools we built. We finish with the important
lessons we learned along the way, some tools we developed that would be
useful in dealing with legacy systems, a benchmark comparing some
alternative system architectures, and an assessment of the scalability
of the system.
|