Tuesday, May 27, 2008

Saving money with the zIIP

The zIIP is the specialty engine that can also be used by ISVs to redirect workload to it. And allthough I claim to keep a close watch on the mainframe world, a couple of things slipped my attention. There are already more softwares using it than I thought. I guess I'm not alone so here's how I got to them.

Last week Trevor Eddolls started me thinking with his blog post on 'Saving money on mainframes'. Mr. Eddolls wonders whether specialty engines (especially zIIP and zAAP) can save you money :
Basically, any workloads run on these specialty engines do not form part of an organization’s contracted mainframe processing capacity. So their use results in a reduction in that organization’s Total Cost of Ownership (TCO). As a consequence, not only do they get reduced software costs, they also get additional processing capacity – and that can be used to eliminate or delay the next upgrade.
Saying that every workload which is redirected to zIIP or zAAP results in a reduction of your software cost is a bit too straight-forward. You still have to keep in mind that your monthly software cost (MLC) is based on a 4-hour rolling average. If you have your peaks at night running a heavy IMS batch workload, you won't see any direct savings on that side by adding a specialty engine. Still, I think the second part of his statement is correct. The real savings come from redirecting workload to the specialty engines and delaying a next upgrade. Delaying this upgrade may also mean delaying higher software costs. If not MLC software, then perhaps for softwares which are still based on the full GP engine capacity of your machine.

Mark Fonteccio comments on that with 'Mainframe specialty processors: Do they really save money?', making the same reflection as I did : "By taking workloads to those processors, you can free up room on the central processors. That’s what might matter the most".
But both have the impression that specialty engines might be the talk of the town, but people aren't convinced yet. I think we're reaching (or have reached) a turning point : companies are actually starting to buy them. Why, because there are more and more softwares using it. More than I was aware of and I'm sure I'm not alone.

So, I was glad to read the comments of Edward Jaffe (Phoenix Software) and Tom Harper (Neon Enterpirse Software) on this subject on the IBM-Main discussion list. Not only do they point out that according to IBM specialty engine sales are up 85% year-over-year, but they also indicate that both their companies have their softwares redirect workloads to the zIIP. So eventually I came up with more than just DB2 and IPSec redirecting workload to the zIIP. Here's a little list :
  • Of course we have the DB2 and IPSec workload from IBM. And that's not the end : DB2 9 now includes support for native SQL stored procedures and most z/OS DFSMS SDM (System Data Mover) processing associated with zGM/XRC was also added.
  • Phoenix software has (E)JES. A little quote : "(E)JES V4R5.0 makes a significant portion of execution resources eligible for redirection to zIIP processors, including all end-user interactive host client environments except CICS, all batch scripts running in the background or foreground under TSO/E, when calling the (E)JES API, and when running (E)JES as the Operating System Interface within the workstation component server."
  • Neon Enterprise Software had an announcement on May 15, 2008 for their IMS Utilities suite : "The NEON Eclipse Reorganization Utilities include iUnload, iLoad, iBuild, iCheck, iSurvey, iExtract, iReorg and iCopy. Version 5.1 of these products includes the initial support for zIIP processors. With NEON Eclipse Reorganization Utilities, it is possible that a customer could experience capacity gains of more than 70 percent for some IMS database maintenance processing. NEON is planning a release in the second half of 2008 that will provide 97 percent zIIP processing support for all the NEON Eclipse Reorganization Utilities, including iCopy".
  • CA was one of the first to announce support for the zIIP with a range of products including e.g. CA-Vtape Virtual Tape System and CA Tape Encryption. CA also put CA-IDMS and CA-DATACOM on the roadmap, but nothing really official has been announced so far.
  • Syncsort for z/OS is another one that has been using the zIIP since January, 2008 : "SyncSort exploits the new MIDAW and System z9 Integrated Information Processor (zIIP) facilities available on z9 processors. SyncSort's use of the MIDAW facility reduces CPU time and elapsed time. The zIIP facility allows many sorts to have a portion of their processing directed to the zIIP, thereby lowering the traditional CPU time cost associated with sorting. The zIIP exploitation also liberates conventional CPU cycles for use by other applications that do not exploit the zIIP facility".
  • Who's missing so far ? BMC ? I saw some announcements with statements on zIIP support, but I don't know whether it's really having any products yet. If so, please do let me know!
So, the list is definitely longer than you might've thought and I'm sure it's still going to grow significantly in the next couple of months.

Thursday, May 22, 2008

Show and Tell

When I started this blog and had to chose a name I wanted to add Belgium to indicate my location. But what about the main title. Weren't mainframes supposed to be dynosaurs, extinct, definitely not cool. Even IBM turned to zSeries, System z and what next ? So I still decided to stick with the old-fashioned term 'mainframe' because I liked it and people around me knew what I was talking about. And now I see practically everywhere the term mainframe gets to be used again. Even in this nice, little 'Show and Tell' movie. I know it mentions EMC, but still, I like it, especially when the father explains what mainframes are doing. Talking about a culture clash. Just watch it :

And while I was browsing through some mainframe movies at Youtube I came across another, I'd say, Show and not even Tell. The Mainframe and The Green Data Center explained in less than a minute.

Monday, May 19, 2008

Live Virtual Classes

Next week, there's an interesting Live Virtual Class on 'High Availability for Linux on IBM System z Servers'. Its contents :
Linux virtual servers are increasingly being used to support critical applications in IBM System z and zSeries server environments. As the mission becomes more important; so does the need to ensure its supporting infrastructure is highly available. But how? z/OS system programmers know the best practices for eliminating single points of failure for their systems; but may be unsure how to translate those techniques to the Linux world. Admins coming from a distributed Linux background may wonder how a virtualized environment on zSeries changes the game. And both groups may ponder how to best marry the Linux and z/OS worlds to maximize availability.

Wonder no more. This session will cover a set of high availability architectures for Linux virtual servers (LVS); in the context of serving data to WebSphere applications (though WebSphere itself won't be the main focus). We'll discuss:

1. Single points of failure in an LVS environment; probabilities of each; and the relative cost to eliminate them -- with examples of how to do so
2. The power of virtualization to minimize degradation of service in the wake of failures; and to reduce the need for large clusters of redundant servers
3. How software running on Linux virtual servers can work cooperatively with a Parallel Sysplex and z/OS data sharing groups
4. Sample architecture specifics; including architectural decisions and tradeoffs; configuration options; and product technologies used
Date : Wednesday, May 28, 2008
Time : 11:00 AM EDT (New York), 10:00 AM CDT, 4:00 PM London, 5:00 PM Brussels
Registration over here.

There's another one the day after on z/VSE Tools.
The z/VSE web site offers more than 20 tools ('as is'; at no additional charge) for download. The tools are designed make certain z/VSE tasks easier and more interesting. Because of the sheer number of tools; you may have lost track about what tools are provided and what they are for. There may be a tool available that you are not aware of; for a task you need to perform. This session will provide an overview of each the tools that are currently available on the VSE homepage. For every tool; a short description and usage scenarios will be discussed.
Date : Thursday, May 29, 2008
Time : 11:00 A.M. EDT New York, 10:00 AM CDT, 3:00 PM London, 4:00 PM Brussels
Registration over here. (the link's not working for me at the moment, hope it works for you).

If you've missed any Live Virtual Class, you can still find them for playback at the Z/VM Live Virtual Class page and at the z/VSE Live Virtual Class page.

Tuesday, May 13, 2008

Red Alert : Sysplex-wide impact due to loop in GRS on z/OS 1.9 and higher

As yesterday was a day off here in Belgium (Whit Monday), I just picked up this Red Alert today. The contents of this Red Alert :
In z/OS 1.9 GRS recovery for GQSCAN retries with incorrect registers. This can result in an ABEND0A0 followed by repetitive ABEND0C6 or ABEND0C4 in ISGQNX.
As a result of the program check loop in ISGQNX's recovery no GRS work is able to run on the failing system. This may cause sysplex-wide impact (...)'.
Check out all the details and what recommended actions to take on the Red Alerts Page.

And if you haven't subscribed to them yet, please do so over here.

Thursday, May 8, 2008

10 reasons for IT failure

This caught my attention when I received the ZDNet Newsletter. It pointed to the blog of Michael Krigsman with the rather uncommon name : 'IT Project Failures'. He points to the original author of this list, Mike Kavis, who writes about this in his blog on 'Enterprise Architecture'. Here's a summary, but just take same time to read the entire topic. It's really interesting, perhaps somewhat confronting, as you'll recognize a lot. At least I do.
1. Poor Communication
2. Underestimating or ignoring impact of change
3. Lack of Leadership
4. Lack of strong executive sponsorship
5. Poor project management
6. Poor Planning
7. Trying to do it cheap
8. Lack of technical knowledge
9. Lack of sound business case
10. Poor vendor management
How about telling us how we do have to do it ?
Well, he doesn't leave us entirely clueless :
1. Build strong business case
2. Secure executive sponsor and top level buy in
3. Create a Road Map
4. Communicate the Road Map
5. Empower Others to Act on the Road Map
6. Start small, deliver early and often (agile)
7. Expand, leverage reuse
8. Govern
As I always say, just check it out !

Tuesday, May 6, 2008

z10 EC : OSA-Express3 and InfiniBand

Here's an addition to the original z10 EC announcement (ZG08-0289). This new z10 EC announcement (ZG08-0354) gives the availability dates for the OSA-Express3 cards and the InfiniBand coupling links.
The OSA-Express3 10 Gigabit Ethernet LR, with 2 ports per feature, was announced on February 26, 2008 and becomes available on the z10 only on May 30, 2008. The number of ports is doubled and it has improved throughput and reduced latency. As a consequence, next to some other System z features, announced here (ZG08-0355), the OSA-Express2 10 GbE LR will no longer be available on the z10 after June 30, 2008.
Not announced on February 26, 2008, but also available on May 30, 2008 is a new OSA-Express3 Gigabit Ethernet SX and LX card with four ports - as opposed to two ports on its predecessor.
The InfiniBand Coupling Links also become available on May 30, 2008. As announced earlier, they are also available on z9 EC and z9 BC S07 dedicated coupling facilities. It is called PSIFB : Parallel Sysplex using InfiniBand. These new coupling links bridge a distance of 150 meters with a throughput of 6GB/s in each direction for the z10. This connectivity is realised with 50 micron cables with new MPO connectors. It is defined with a new CHPID : CIB (Coupling using InfiniBand). Using 2 physical ports you can define 16 CHPIDs : this allows using the same physical Coupling Links for multiple sysplexes. It might be a good replacement for ISC-3 links used within the data center.
I think Balanced Power Plan Ahead is also something new. This is really not my cup of tea so I'll just quote the announcement : "Phase currents are minimized when they are balanced among the three input phases. Balanced Power Plan Ahead (#3001) is designed to allow you to order the full complement of bulk power regulators (BPRs) on any configuration, to help ensure that the configuration will be in a balanced power environment".