[From Dr. Dobb's Journal]
Anton Okmianski explains how his company settled on Berkeley DB as its database of choice for meeting the high-performance requirements associated with provisioning broadband devices.
"To meet the high-performance requirements associated with provisioning broadband devices, we at Cisco Systems (http://www.cisco.com/) designed our Broadband Provisioning Registrar (BPR) to have a distributed architecture. At the heart of the system was a central database to store device-configuration data. However, we were wary of using standard relational databases, which are bloated with features we didn't need, and performance and management overhead we didn't want. In the end, we settled on Berkeley DB, a lightweight embedded database from Sleepycat Software.
"When considering Berkeley DB, our first challenge was to design a database layer that mapped our relatively complex data schema into the Berkeley DB's simple key/data pairs. The next challenge was to deliver performance of at least 150 provisioning transactions per second on relatively modest hardware -- 2x750-MHz Sun Fire 280R with 10,000 RPM disks in a RAID5 setup. In this article, I'll explain how we tackled these challenges."
This article continues at
Back to Database Journal Home