IBM's PowerLinux is the combination of a Linux-based operating system running on IBM Power Architecture technology. This combination produces a customizable system capable of handling large quantities of data more efficiently without the need for additional hardware. To determine real world performance, we composed a test scenario with two comparable systems running large database operations. This test also shows that Linux can be an abstraction layer on different types of hardware that are better at running certain workloads. We believe that we're entering an era where customers will choose Linux first, then choose the hardware platform most capable for the workload such as ARM, x86, Power or others.
For our test, we have compared two systems that are in the same price category. As the architecture of power and intel are fundamentally different, we've chosen to compare two systems: a HP Blade with Intel Xeon architecture, and an IBM PowerPC.
We ran two separate tests: a full read-write test with high congestion to simulate a very busy database, and a second read test without any congestion so the database server could use all resources available to perform as fast as possible. These two tests simulate the worst and best conditions; your typical database usage will lie between these values. We have performed the tests on RAM-based disks, so we could fully test the performance of the architecture and processing power instead of testing the performance of the storage layer.
In the first test we've performed a test based on the TPC-B test: a test designed to simulate the most difficult operations for a database: perform SELECT, UPDATE and INSERT on the same records to simulate the congestion a database will face when performing day to day operations. As the test will often try to update and lock the same records, this is a representative test, as most of your operations will happen on the current orders your company is working on.