Monday, December 14, 2015

IBM DB2 12 for z/OS Early Support

Intro
I was thinking that there was little use in going through those older announcements I missed the last couple of months, but when I look at the information that's easily available, I have the impression that I'm reading announcements a bit differently than other people. So that's why I think some of you might nevertheless enjoy or get something out of my information.

Announcement
So, October 6, 2015 IBM announced the early support program for DB2 12 : 'IBM DB2 12 for z/OS Early Support Program (ZP15-0504)'. A select group of customers will be stepping into this program that becomes available March 4, 2016. If you're interested in participating, contact your IBM representative.

CPU savings and performance improvements
IBM makes a lot of work on CPU savings while increasing the performance. A reverse trend of what we used to know in the past that seems to steadily continue. Some figures : "Compared to DB2 11, DB2 12 clients can achieve up to 5% CPU savings for various traditional OLTP, heavy INSERT, and SAP workloads. Select query workloads may see even more benefits, with up to 25% CPU savings. Read-intensive OLTP workloads may see up to 10% CPU savings with DB2 12. Decompressing data is much faster using DB2 12 with significant performance improvements, delivering up to 15% CPU savings when running the queries against compressed tables".
Since IBM did already a great job on improving the use of CPU resources in DB2 V10 and DB2 11, it's amazing they still succeed in improving it again. Would earlier DB2 coding really have been that lousy ? I guess not, it just wasn't a focus at the time, but the competition is fierce nowadays.
Some more stuff is also offloaded to the zIIP adding to a better price/performance rate.

Use of (much) larger memory in z13
What else ? I think it's only logical IBM is heavily using the new features of the z13. I particularly think of the memory that - as the announcement says it - has become "more affordable", when they just mean "much cheaper". Having a bunch of memory at your disposal (forget GB we're talking TB) eases a lot of things. One example : "enhancements to compression aids DB2 utility processing by reducing elapsed time and CPU consumption with the potential to improve data and application availability" or in-memory performance improvements like
  • Contiguous and larger buffer pools
  • Use of in-memory pipes for improved insert performance
  • Increased sort and hash in-memory to improve sort and join performance
Prerequisites
That's why I think you will get the best benefit out of DB2 12 with a z13. Allthough the prerequisites are not that stringent of course
  • z/OS V2.1 Base Services, (5650-ZOS), or later
  • z196, or higher, processors running z/OS V2.1, or later
Planning
One other thing : back in 2012 I made the following little table when DB2 11 was pre-announced
  • V6 : GA 1998, EOS 2005
  • V7 : GA 2001, EOS 2008
  • V8 : GA 2004, EOS 2012 (support lasted an extra year)
  • V9 : GA 2007, EOS 2014 (back to normal)
  • V10 : GA 2010, estimated EOS 2017
This trend seems to continue. V10 is EOS in 2017. Announcement of DB2 11 was in 2013, so I guess the announcement of DB2 12 will be somewhere by the end of 2016. So, if you're still on DB2 10, you have some two years left to get to DB2 11. Don't wait until DB2 12 is out, because apparently no skip level migration will be available. 

No comments: