IBM has just released a very interesting new paper, looking at the performance improvements delivered by DB2 11 for z/OS within an SAP application environment. The paper can be downloaded from IBM using this link: IBM zEnterprise System: DB2 11 for z/OS with SAP® Performance Report
Here's the report abstract:
The performance of DB2 for z/OS is critical to the SAP solution on System z. As a result, IBM invests heavily in it. With DB2 11 for z/OS, large investments were made in both development and test. Over forty new features were put in DB2 11 specifically for SAP. Changes were made in DB2 11 to improve performance and to enhance continuous availability, scalability, and security for business critical information. These investments have paid off with DB2 11 for z/OS in the SAP environment.
DB2 11 for z/OS delivers “out-of-the-box” DB2 CPU savings of up to 10% for complex online transaction processing (OLTP) and up to 10% for update-intensive batch workloads, when compared to DB2 10 running the same workloads. Even greater savings can be seen with the SAP Banking Services (SBS) Day Posting workload. Complex queries may enjoy up to 40% CPU savings. Utility performance improvements have also been delivered.
DB2 11 for z/OS delivers scalability improvements. Support for larger relative byte address (RBA) and logical record sequence number (LRSN) is available. A larger dynamic statement cache is supported. Reductions in DB2 latch class contention in buffer management have been delivered.
SAP AG completed certification of DB2 11 for z/OS at the same time that IBM made DB2 11 for z/OS generally available. This is the earliest SAP certification of any DB2 for z/OS release. See SAP Note 1850403 (Release of DB2 11 for SAP Components).
The IBM SAP on System z Performance Team, located in Poughkeepsie, NY, worked very closely with the DB2 for z/OS Development and Performance teams for over two and a half years throughout the development of DB2 11 for z/OS. We evaluated the performance of early DB2 11 drivers, prototypes, and tuning recommendations. We collected and analyzed performance data. We identified and helped debug performance issues, as well as functional problems. We drove performance improvements into DB2 11 for z/OS. And, finally, we executed a set of measurements comparing DB2 11 with DB2 10.
We focused on running DB2 11 for z/OS with SAP on an IBM Enterprise Class machine, a zEC12 . Specifically, we ran the SAP Banking Services (SBS) Day Posting workload, which is a good representation of a customer OLTP workload. We used various configurations for these tests to show how DB2 11 performed in different environments and how it compared to DB2 10. We focused on areas that are important to our customers running SAP on System z. We ran in both DB2 single system and data sharing environments. We ran with different database sizes and different number of partitions in the database. We ran with all general purpose processors (CPs) and a mix of CPs and zIIPs.
This paper is a culmination of all this work. It is a joint effort between the IBM SAP on System z Performance Team and the IBM DB2 for z/OS Performance Team. In it, we will discuss the results of key measurements that were done and our experiences with DB2 11. The measurement tests that were done were stress tests, not certified benchmarks.