This article is an excerpt from DB2 11: The Database for Big Data and Analytics (MC Press, 2013).
In many situations, the speed at which analysts can understand and react to changes in a dynamic business environment is limited by the IT infrastructure’s ability to process complex queries against large volumes of data. Ideally, business analytics would be unconstrained by infrastructure limitations. Organizations could leverage near-instantaneous analytics and run reports when they need them, even in real time. This is where DB2 query accelerators come into the picture.
A DB2 query accelerator is a transparent-to-the-user tool designed to boost database speed and performance. Accelerators can improve response time dramatically and reduce CPU utilization by offloading eligible queries to specifically designed hardware. They can help deliver faster, more reactive business insight by executing analytics when they are required and at high speed.
The IBM DB2 Analytics Accelerator can substantially improve performance and reduce the cost of analytics in DB2 for z/OS environments. Available as an add-on appliance built on IBM Netezza technology, the accelerator is designed for easy, rapid deployment. Users simply instruct DB2 to consider the query accelerator the new access path for eligible queries on eligible objects. The appliance requires very little configuration, and accelerator performance information is integrated with the usual DB2 traces.
For candidate queries, the results can be astonishing. Queries often run significantly faster than they ever have before. Response times for well-tuned queries running on current-generation traditional hardware can shrink from hours to seconds. Queries that previously ran too slowly to be useful can often be completed in minutes. As response times approach what you might expect from an OLTP environment, realtime analytics can become an everyday reality.
Potentially, a good portion of the CPU consumed by queries running in DB2 can be eliminated by offloading the query processing to the accelerator. Nevertheless, offloaded queries returning a large result set could consume a noticeable amount of CPU time in DB2. This scenario is improved with a performance enhancement introduced in DB2 11 for z/OS.
Figure 1 shows the CPU utilization and evolution of the millions of service units (MSU) 4-hour rolling average during the execution of a data warehousing workload. This scenario is the combination of short, medium, and long intensive queries executed entirely in DB2.
Figure 1: Workload before acceleration—Total LPAR CPU utilization
Figure 2 shows the same scenario after adding an accelerator appliance to the environment. The most complex queries were offloaded to the accelerator. As a result of the faster accelerator performance, the whole scenario was accomplished in a fraction of the original elapsed time. Most of the z/OS CPU was removed from the LPAR as queries were executed in the accelerator, reducing the impact in the MSU 4-hour rolling average, which is the usual software license cost metric.
For clarity, the chart scale is unchanged. This example shows how eligible workload can be executed faster and less expensively in the accelerator.
Figure 2: Workload after acceleration—Total LPAR CPU utilization
The secret of this speed resides in the highly specialized hardware and software that is tuned for serving analytics queries. The DB2 Analytics Accelerator appliance exploits massive parallel processing on dedicated CPUs, disks, and memory in a highly and linearly scalable architecture.
Figure 3 depicts the relationship between an application, DB2 for z/OS, and the accelerator. The accelerator appliance connects to z/OS and DB2 using a private network. The DB2 objects to be accelerated are defined and loaded in the accelerator using DB2 stored procedures and a graphical user interface. Once an accelerator is installed and enabled, the DB2 optimizer considers the appliance as a new access path and will offload SQL processing transparently when it judges that doing so would be more efficient.
Figure 3: IBM DB2 Analytics Accelerator and DB2 for z/OS
The business value of a query accelerator resides in its close and transparent integration with DB2 and System z. You get hyper-speed analytics in the highly reliable, secure, and stable mainframe environment that you already know and love. DB2 Analytics Accelerator not only can help reduce total cost of ownership but also can change how analytics are executed on DB2 and System z, the platform of choice for enterprise warehousing infrastructures.
LATEST COMMENTS
MC Press Online