Despite the traditional news dead zone as many people take a well-earned
summer holiday, this month has seen a couple of interesting announcements from
IBM which will impact many of us over the next year or two.
I've been spending time with a few large customers recently, helping them
upgrade from DB2 for z/OS V7 to V8. This upgrade is usually one of the more
painful ones, as the move to a 64-bit platform can result in some
performance/CPU regression for certain workloads, and there are some other
pre-requisites for the upgrade, which can be problematic. However, Version 7
has been out of support for over two years now (end of support was June 2008)
and very few large organizations can justify running mission critical applications
on software no longer supported by the vendor.
The bad news for those customers who have only just got to Version 8 is that
IBM has just announced the end of support for that release as well. The key
date is April 30th 2012, which gives V8 users just over 20 months to
justify, plan and implement their next migration. This process can take 12-18
months for many large DB2 sites, so if you're a Version 8 customer and you
haven't started planning yet, perhaps this announcement will give you the
motivation to do so.
As I outlined in one of my previous columns, at Version 8 you have a choice
as to whether to follow the normal upgrade process and move to Version 9, or
take advantage of the skip-release option and go directly to Version 10. My
previous advice still applies: take the normal upgrade route to V9 if you've
been on V8 for a while and have your migration to V9 in plan. If you have only
just made it to V8 (or worse still you're still in the process of getting
there) you might want to consider skipping directly to V10 at some point during
the next two years, so you can get ahead of the upgrade curve again rather than
constantly having that "End of Support" date hanging over you like
the Sword of Damocles.
If you go down the V9 upgrade route and need some help in justifying the
upgrade, I've included a link below to a White Paper I produced for IBM a
couple of years ago which outlines the underlying business value of the new
features and functions within V9 which you might find useful when talking to
management. I'll be producing a similar paper for Version 10 over the next few
months, and will link to it here when it's done.
pureScale on System x
Last month in this column, I wrote about IBM's DB2 pureScale technology and
how I consider it to be the single most important new feature in the DB2 for
Linux, Unix and Windows product for the past 10 years.
As originally announced, DB2 pureScale was supported only on IBM's AIX-based
POWER6 and POWER7 midrange server platforms. These are powerful machines
capable of supporting large workloads using advanced technology such as
hardware parallelism, PowerVM virtualization and Dynamic Logical Partitioning.
The combination of DB2, pureScale and the POWER6/7 platform provides a
compelling solution to organizations needing high performance, extreme
availability and scalability for their database without investing in mainframe
technology. However, it is also relatively expensive compared to the
Intel-based servers used by many small/medium businesses to run their IT
On 5th August, IBM announced that the DB2 pureScale feature would
also be available on selected models of their Intel-based System x server line
(specifically, the x3650 M3, x3690 X5 and the x3850 X5) running SUSE Linux.
This announcement significantly increases the potential customer base for
pureScale by decreasing the minimum cost of entry and allowing organizations
with a strict Linux OS strategy to consider a pureScale solution. IBM can now
offer a highly available, highly scalable, application transparent database
clustering solution on System z, System p and System x, covering the complete
spectrum of server profiles.
Although the cost for this solution is likely to be lower than that for a
POWER7 based environment, it still won't be down to commodity prices. The
announcement specifically limits support to the three IBM System x servers I
mentioned above, and you'll need to budget for enough capacity to host the
coupling facility, either in the form of separate dedicated System x servers or
as logical partitions within the main server nodes. (The coupling facility will
host the same PowerHA clustering architecture I discussed last month, but in a
software-emulation mode running under Linux). You're also going to have to spec
Infiniband network links for the various servers, which are significantly more
expensive than good old-fashioned Ethernet cards. However, when you consider
the scalability and availability advantages of pureScale this will still be a
remarkably cost-effective platform.
Triton has been fortunate enough to be involved in the early beta program
for DB2 pureScale on System x for the last few months, and I've been absolutely
astonished at the price/performance that is possible with this technology. It
will be interesting to see where else IBM takes this in the coming months, and
what other platform/operating systems will be supported in the future.
See All Articles by Columnist