|
In November 2003, B4E, together with Cap Gemini TMNG, completed a customer paid benchmark of
B-oo-levard® at the Sun Microsystems benchmark centre in Germany.
The Customer Requirements:
(1) Handle an installation with 12 million subscribers mirroring a typical and real business
environment (e.g., tariff plans, discounts) on a Sun Fire 6800 with 96 GB of main memory
(2) Prove full clientele capability for more than 550 independent
service /content providers (ISPs)
(3) Prove 100% transaction safeness
(4) Update tariffs / products without stop and restart of the application
(5) Perform reference data updates (EAI-transactions) against the running system without
influencing the ongoing rating or billing processes
(6) Demonstrate performance creating service provider bills
(7) Demonstrate performance creating event record files for ISPs
(8) Demonstrate performance creating various reports
|
The Benchmark Results:
(1) With the given hardware 12 million subscribers could be handled processing
completely about 6 million event data records per hour (please refer to
B-oo-levard®OneStepRating to understand our different approach
to rating and billing)
(2) Clientele processing was proven for all ISPs (more than 550)
(3) Application and database were manually “stopped and restarted” – no data was lost
(4) New tariffs were created, affirmed, and correctly processed against special
Event Data Records without shutting or slowing down the system
(5) Subscriber data, pricing schemas and more were changed using EAI transactions
without any negative implication to the rating performance of B-oo-levard®
(6) Bills for all 564 ISPs were created within 4 minutes while B-oo-levard®
was continuing with rating in parallel; about 60 % of all EDRs were considered for
these bills
(7) A throughput of 10 million EDRs per hour for the files to be sent to ISPs was
achieved while the rating process was running continuously in parallel
(8) High spender report, daily revenue report, and state authorities inquiry were
created within 5 to 8 minutes while the rating process was running continuously in parallel
|