Exploring IBM Z Sort Accelerator Support in Syncsort MFX
For organizations running IBM Z mainframe systems, several converging trends in recent years have made it more challenging than ever to achieve the needed organizational performance at the best possible price point.
The volume and velocity of data have increased dramatically as mobile devices, IoT sensors, and smart machines have proliferated. An increasingly diverse array of applications running on those devices must, in many cases, access information that resides on the mainframe. Expectations are higher than ever, as end-users expect price quotes, account balance inquiries, and other queries that rely on mainframe data to occur almost instantaneously.
That has prompted IT leaders in mainframe departments to seek out ways to speed processing, especially with regard to many of the mundane processing tasks such as sorting, copying, merging, compression, and report generation. These typically consume a disproportionate share of costly CPU processing time and related resources, often up to 25 percent of processor usage and up to 30 percent of I/O channel resources.
Fortunately, there are tools available that make it possible to run these processes far more efficiently, both by offloading them to dedicated hardware components and by implementing software utilities such as Syncsort MFX. The combination of these two options can be especially effective, as we’ll discuss a bit later in this article.
Watch our Webinar
To learn more about Syncsort MFX enhancements, watch our webinar which will highlight Syncsort MFX enhancements in more detail.
IBM Innovation Addressed the Challenge with Specialty Engines
In 2004, IBM responded to these challenges by offering specialized hardware components, essentially spare CPUs dedicated to performing specific functions, which IBM refers to as “specialty engines.” These are typically activated to handle specific workload types, provided that the proper additional licensing is in effect.
Offloading tasks from the primary CPU to a specialty engine frees up the primary CPU to do more work and achieve faster performance, including on core enterprise applications such as OLTP, batch, and ERP. There are additional costs associated with the use of specialty engines, but these are generally one-time costs, whereas processing time on the primary CPU is subject to IBM’s usage-based licensing fees.
IBM’s first specialty engine was called the z Application Assist Processor (zAAP) and was limited to running certain Java code and processing XML. Later, IBM introduced the z Integrated Information Processor (zIIP), which was assigned to handle Db2 workloads, as well as the Java and XML workloads that were formerly handled by the zAAP.
Precisely’s Innovative Addition to the Mix
For 50 years, Precisely’s Syncsort MFX has been the most frequently installed third-party software product on IBM and plug-compatible mainframes. Deployed in over 85 countries, Precisely’s Syncsort MFX solution is a high-performance sort, copy, and join utility designed to exploit the advanced capabilities of the z/OS operating system and IBM Z mainframes.
Initially, Precisely’s Syncsort product was focused on optimizing sort processes. Over the years, it evolved to take advantage of virtually every significant hardware and operating system advancement released by IBM, delivering improvements of up to 25 percent in sort-related CPU processing time.
Naturally, the developers of Syncsort MFX looked at IBM’s zIIP as yet another opportunity to improve system performance and free up CPU time, saving Precisely customers money. In 2014, Precisely introduced a new capability in Syncsort MFX, which enables sorting/copying workloads to be easily offloaded to zIIP – especially with data-transformation features such as INCLUDE/OMIT, INREC/OUTREC, and OUTFIL.
Customers using Syncsort MFX and the zIIP offload capability on copy work have reported reductions of 25 percent on elapsed times and reductions of up to 90 percent on CPU times. With SMS compression via Syncsort ZPSaver, CPU time reductions of as much as 80 to 90 percent may be achieved.
New Sort Innovations
As new technology is delivered by IBM to the Z Platform, Precisely continues to embrace and extend those innovations. Two recent IBM enhancements have provided an opportunity for Precisley to again deliver more value to customers. IBM’s Integrated Accelerator for Z Sort is a new coprocessor designed for the z15. This new accelerator can help reduce CPU usage and improve elapsed time for eligible workloads by speeding up sorting, shorten batch windows, and improving select database functions. Precisely worked closely with the IBM z/OS team in Poughkeepsie, NY, and their hardware architects for the sort accelerator. We have developed new algorithms in Syncsort MFX to take advantage of the coprocessor and will release this capability later in 2021. With the new algorithms, our customers can see dramatic improvements to batch sort job performance. In our lab, we are seeing CPU and elapsed time improvements of up to 35% depending on the key length, record length, file size and other factors.
The second innovation is in security. The IBM Z platform offers highly secure ways to help deal with compliance and regulatory requirements. The IBM Z Pervasive Encryption solution is a method to enable extensive encryption of data in-flight and at-rest to meet these protection standards. The challenge to implementing data encryption is that it consumes processor cycles, so it doesn’t come without penalty. Although mainframe customers love the pervasive encryption approach, no one likes the additional resource consumption. For Syncsort MFX users, If the input or output data set is encrypted, BSAM must be used instead of our high-performant low-level IO access methods, and there is an extra cost from that perspective. Once again working closely with IBM, we have identified a way to continue to use our low-level IO for encrypted data sets and improve encryption performance.
We have seen great performance improvement from our benchmark testing. For Syncsort MFX, we have seen up to 45% CPU and 40% elapsed savings, and for Syncsort MFX with zIIP offloading enabled, we have seen up to 80% CPU and 40% elapsed time savings when processing encrypted, basic and large format datasets.
To learn more about Syncsort MFX enhancements, watch our webinar New Mainframe Sort Innovations Build on IBM Z Platform Enhancements which will highlight Syncsort MFX enhancements in more detail.