Recently I shared my impressions of the latest version of IBM DB2 Analytics Accelerator. Now I'm here to let you know that there's a new Redbook that details the features found in Version 4.1.
"Reliability and Performance with IBM DB2 Analytics Accelerator 4.1" is currently a draft publication. The formal publishing date is set for the end of July. I'll list just a few highlights:
* Chapter 5 examines new query accelerator features, including static SQL support as well as support for “NOT IN <subquery>” and many other new sub-query predicates. Numerous customers I talk to believe that their current workloads aren't large enough to run cost-effectively on DB2 accelerators. However, with the recent query enhancements, it may be time to reconsider this. Use the monitoring tools and you may discover that SQL statements that previously couldn't run on the accelerator now can.
* Chapter 6 provides a look into the primary factors that affect throughput of loading data into the accelerators -- partitions and table number and sizes; processors (number and speed), and; Netezza host. The chapter goes through different scenarios to help you understand your database design and the best way to load the data.
* Chapter 7 covers online data archiving with the High-Performance Storage Saver. When archiving occurs, the table space partition is put into a persistent read mode with a status of PRO. The process to archive and restore is handled by two separate stored procedures. However, I'm not sure how to remove the archived partition when you no longer want to save it on both DB2 and the accelerator. Hopefully the process of purging old data from the accelerator will be clarified in the final version of the Redbook.
* Chapter 9 is devoted to monitoring enhancements. Having spent most of my career tuning SQL, I find this information extremely interesting. There are several options for monitoring the accelerator: DB2 Accelerator Studio, DB2 commands, query history tables and DB2 monitoring tools from IBM or third-party vendors.
The DB2 command –DIS ACCEL(name) DETAIL provides information about active and queued queries, wait times and disk storage utilization for the named accelerator as well as totals for all subsystems connected to the named accelerator.
As for real-time monitors, IBM continues to add information to the standard interfaces it and third-party vendors use to monitor DB2 activity. The first set of information allowed customers to see what is happening on the Netezza box. The next set of information provided more insight into which statements are running on the accelerator and how much was saved from general DB2 processing. In addition to monitoring utilization levels, customers need to know that the statements they create are capable of running on the accelerator. Making this information available real-time via the monitoring tools adds considerable value to this product.
Here's the complete table of contents:
Chapter 1. The analytics lifecycle
Chapter 2. Database design considerations
Chapter 3. Overview of IBM DB2 Analytics Accelerator V4.1
Chapter 4. Installation and maintenance procedures
Chapter 5. Query enhancements
Chapter 6. Load and incremental update
Chapter 7. Online data archiving
Chapter 8. High availability, disaster recovery, and workload balancing
Chapter 9. Monitoring enhancements
Chapter 10. The Accelerator Loader
Appendix A. Preparing input data for Accelerator Loader
I feel this publication is well worth your time. If you read it, please share your thoughts in comments.