(Originally posted 2007-04-19.)

Parwez Hamid: System z9 BC GA2 and System z9 EC GA3 Update

Available from 11 May, Driver Level 67 provides the BC GA2 / EC GA3 enhancements. The current one is 63. Parwez set the expectation the upgrade can be done non-disruptively if you plan carefully.

Common to BC and EC machines:

• LPAR Group Capacity Limit. Group of LPARs on one footprint need not be in the same parallel sysplex. Requires z/OS Release 8.
• Coupling Facility code (CFCC) Level 15. Exclusive to z9. Designed to provide a basis for future CF enhancements (which will require z/OS R.9). z/OS R.6 with OA17055 for function and OA17070 for RMF support. Right now provides more allowable tasks (112 vs 48). (I mentioned this as part of Joan’s presentation yesterday.)
• System-Initiated CHPID Reconfiguration reduces operator burden when taking CHPIDs offline. z/OS R.6 with PTFs required.
• Multi-path IPL.
• CHPID type FCP (QDIO) and OSA performance enhancements. (I mentioned these as part of Connie’s presentation yesterday.)
• QDIO Diagnostic Synchronisation (Connie again).
• On/Off Capacity on Demand flexibility improved so may well provide more responsive activation of temporary capacity.
• HMC Logon LDAP Authentication enhances security for the Hardware Management Console (HMC). There are other HMC enhancements.
• The System Activity Display (SAD) now has the system’s power consumption. The example of a 10-IFL machine is 4.7kW. Parwez said the biggest z9 configuration only takes 18kW – when I asked if his example was realistic. Personally I think these numbers are VERY good. And here’s an example:

BC only:

• R07 Capacity Backup (CBU) Enhancement allows conversion to the capacity limit of a S07. Note: Some R07 limitations (compared to S07) remain. S07 and R07 are the hardware models, of course.
• New FICON Express4-2C SX card – has fewer channels than EC FICON Express4, to meet “BC customer” requirements.
• Crypto Express2-1P single processor card for “BC customer” requirements, compared to EC card which has two processors.

Statements of Direction:

• Removal of support for Dynamic ICF Expansion – which allows CF LPAR to expand from dedicated ICFs to use shared General-Purpose Engines (GCPs).
• z9 is the last system which will support older than “n-2” technology – whether as the CF or another coupled system.
• z/OS XML processing on zAAP and zIIP specialty engines.
• IPSec encryption and decryption on zIIP is expected be available in August 2007 – using Enclave SRBs. zIIPs support CP Assist Crypto Function (CPACF) as all characterisable engines do. Support is in z/OS R.8 with PTFs.

Me: DB2 I/O with DS8000

2 people in my session – one a hardware storage vendor and the other a business partner who does DB2 education. So we had a nice wide-ranging chat using the foils.

Isabel Arnold: Attracting Young Developers to the Mainframe using Eclipse-Based Tooling

Websphere Developer for z is Eclipse-based, so has the usual Eclipse behaviours. There is a division of labour between the Eclipse-based workstation components and host components. I noticed EGL as one of the supported languages. I’m psyched enough to want to play with it! It would be fair, though, to note that things like refactoring for mainframe languages are not YET as advanced as those for java.

Hendrik Wörner: Websphere Portal V6 for z/OS

A good presentation on why z/OS is a good place to run Websphere Portal. Those advantages are mainly the same as for Websphere Application Server on z/OS.

Isabel Arnold: Mainframe Asset Discovery, Modernization and Integration

A very good SOA-oriented overview presentation, with more detail on componentisation. Websphere Studio Asset Analyser and Application Transformation Workbench work together to help with transformation.

(Originally posted 2007-04-18.)

Me: Memory Matters in 2007

I was pleased that about 60 people attended this presentation.

Connie Beuselinck: System z9 OSA-Express2 Update and Performance Benchmarks

Network Traffic Analyzer was announced today for QDIO problem determination. It’s exclusive to z9 and OSA-Express2. QDIO Diagnostic Synchronisation (also announced today) allows z/OS to signal to OSA-Express2 to synchronise software (z/OS) and hardware (OSA-Express2) traces.

Also announced today was OSA Dynamic LAN Idle, which is designed to reduce latency and improve performance. This feature, which requires z9 and OSA-Express2, affecting OSA inbound blocking and packing.

Connie had some nice but very busy graphs that show z9 performance vs z990 in this area.

Joan Kelley: Pushing The Limits of Parallel Sysplexes: Bigger, Smaller and Further Apart

CF Level 15 was announced today – increases tasks to 112. Recommend re-running CF Sizer as some structures will increase in size.

CF Level 14 had a reduction in “CF to CF” communication service times for duplexing. She reminded us that CF Level 14 also provided a more realistic view of CF Utilisation.

A customer has a 70GB CF structure for MQ!!!! There have been some improvements to allocation that get the allocation time for structures that big down to 3 minutes.

z/OS Release 8 has a new option – Message-Based Protocol – which reduces read I/O to the Couple Data Set, using XCF signals. You need a new format Couple Data Set and to turn it on with a SETXCF command.

Joan has a good discussion – with measurements – of CF Dynamic Dispatch.

APAR OW51813 adjusted the timing thresholds for CF Request Sync to Async conversion. This adjustment has only happened once.

Bob Rogers: Dealing with more Data, A Sysprog View

A good review of the many things we’ve done to allow disk volumes to get larger, the most recent being:

• Multiple Subchannel Sets in z/OS 1.6
• HyperPAV
• MIDAW

(Originally posted 2007-04-17.)

I just started talking to the original developer of the IFASMFDP SMF Dump Program about adding more record selection sophistication to it.

I can promise absolutely nothing but are there any requirements any of you have for the SMF Dump Program?

(Originally posted 2007-04-17.)

Paul Rogers: z/OS Releases 8, 9 and Beyond

This is very much an overview presentation. Other sessions go into more detail.

Paul provided a salutary reminder that z/OS Release 6 goes out of service at the end of September. We’re now supporting only the current release and the 2 previous ones.

APAR OA20314 (in support of zNALC) allows you to IPL with LICENSE=ZNALC rather than hardcoding the LPAR’s name according to the old NALC licencing convention (ZNALxxxx). zNALC is a big step forward in licencing, among other things allowing any modern System z server to participate.

Multiple subchannel sets (z9-109 only) is interesting in that it allows PAV aliases to be defined in the second subchannel set (SS0), freeing up addresses in SS0 (given the 64K limit). HyperPAV also (pardon the pun) addresses this. So the two functions together help greatly for large installations. (They work well together.)

Paul also presented on IBM’s programme of simplification for z/OS System Management. One could argue this takes all the fun out of it. 🙂 But seriously I think it’s an important strand going forward.

The Health Checker appears to be a big success story: When it was owned by the ITSO (the Red Books people) there were more than 3000 downloads. Now in z/OS Release 8 it’s fully integrated into the base operating system.

The writing is clearly on the wall for HFS. zFS is obviously the way to go. Note: This can be a migration worthy of serious planning. Major helpers became available in z/OS Release 7. I’d not noticed, by the way, that RMF had support for zFS.

Paul mentioned briefly an enhancement to WLM-Managed JES2 Initiators in z/OS Release 8. I’m going to have to do more research before commenting further on this one.

Bob Rogers: z/OS R.8 and R.9 System Programmer Goody Bag

About 10% of his audience are already on z/OS Release 8. (He didn’t need to ask who’s on Release 9.)

On Release 8 some things struck me:

• Page Frame Table entries (PFTEs) are not included in SVC Dump. This helps with Release 8’s support of more than 128GB in an LPAR as the PFTEs can be huge. SVC Dump records real and virtual addresses to support this. The net effect is that SVC Dump is quicker.
• PDSE Buffer Manager and Index Manager use 64-bit addressed buffers. There’s also an option to retain buffers after close – which helps with frequently-closed members (such as those read at logon).
• XML parsing is part of the operating system. It is non-validating, with support for C, C++ and Assembler, and parses in segments (to support big XML files). It supports 31- and 64-bit, SRB and TCB modes, and cross-memory. The plan is to produce other kinds of XML parsers in due course. DB2 Version 9 exploits this.
• JES2 is introducing a function to balance initiators across the JESplex (without involving WLM). There are also improvements for duplicate job names.

And on Release 9 (which has only been previewed at this point):

• z990, z890 and z9 BC and EC support the “ASID and LX Reuse Facility”. z/OS Release 6 supported Linkage Index (LX) reuse. Now R.9 supports ASID reuse. This could prolong the period between IPLs. Bob made the comment this helps certain MQ customers.
• SMF can now use the System Logger – potentially for performance. You use the new IFASMFDL program to unload from the logstream. IFASMFDL has OUTDD filtering to reduce the need for multi-pass processing.
• TRSMAIN (Terse) is now supported as AMATERSE. (It’s also available for R.7 and R.8 with OA19194.) The DDs used are different – SYSUT1 and SYSUT2. There are some functional requirements.

Kay Adams: What’s New In zSubCapacity Pricing

Group Capacity Limits supported by z/OS Release 8 – on z9 EC and BC only. You can have multiple groups in the one machine. You can set individual LPAR limits as well as a group limit for the group it’s in.

Kay also talked about zNALC, which became available a month ago. NALC and z/OS.e are withdrawn from marketing at the end of September 2007. I questioned whether Linux applications could be considered for certification, as well as UNIX and Windows. The good news is they can. This IS good news because lots of new applications are being built on the LAMP stack (where L stands for Linux). I’m thinking of eg blogging and wiki software.

It’s not my area but z/VSE 4.1 introduces Subcapacity Licensing for VSE.

Pat Artis – MIDAWs

An excellent explanation that, though difficult material for most, gave a good explanation of what MIDAWs are and why they help when they do.

Me – Much Ado About CPU

I was very pleased to have around 50 people in my audience. And I got some good questions – mostly afterwards.

(Originally posted 2007-04-17.)

It’s been really nice meeting with so many acquaintances. The highlight of Day 1 was undoubtedly Pat Artis presenting on HyperPAVs. As I have a couple of foils on HyperPAVs from the DB2 perspective on Thursday it’s going to be nice not to have to explain them in too much detail – as Pat did such a great job.(The Thursday session is on I/O and DB2 Versions 8 and 9, a much wider ranging version of Jeff Berger’s 2006 MIDAWs presentation. Again these are Jeff’s foils rather than mine.)

(Originally posted 2007-04-16.)

The agenda this time is chock full of great sessions. I’ve “marked my card” for every single session.

Bob Hoey – marketing guy though he is – is giving a keynote speech that fills one with a lot of hope for the mainframe. (Bob, if you know nothing else about him, participated in the infamous “IBM Mainframe Marketing” videos that appeared on You-Tube a few months ago.)

So I’m really looking forward to this week, knowing I’ll meet a bunch of great customers, developers, colleagues in the field, business partners and vendors. In fact I already did. 🙂 And you’ll be relieved to note I read all of my 3 presentations in the past few days, including 2 of them on the plane over.

Will be Twittering through the week and maybe will blog more on the conference. Anyone reading this also here?

And it’s warm and sunny here in the München Hilton, right by the Englischer Garten. 🙂 Wish you were here.

(Originally posted 2007-03-23.)

This year’s IBM System z Technical Conference is 16-20 April in Munich.

Now I’ve got my three presentations together I can relax and have just perused the agenda: There are some very good sessions and some excellent speakers.

My sessions are:

• Much Ado About CPU

  I’ve worked pretty hard to move the z9 stuff into the main flow and also added stuff on z/OS R.8 and a tiny bit on R.9. I think it’s altogether a much better presentation.

• Memory Matters in 2007

  Completely revamped to add in more stuff about subsystems and also the RSM rewrite in z/OS R.8.

• I/O Performance on the DS8000 with DB2 for z/OS Version 8 and 9

  This was written by Jeff Berger of the DB2 lab and I’m pleased he’s asked me to present it for him. There’s less about MIDAWs, which is a good thing as it makes room for other stuff such as DB2 Version 9 and HyperPAVs.

I look forward to seeing you there.

(Originally posted 2007-03-22.)

Here’s a nice picture from the Minding the Planet blog:

(Originally posted 2007-03-20.)

My good friend Linda August (of the IBM Washington Systems Center and previously of the global Performance Management team I was a member of) has written a very good paper (and an accompanying presentation she gave at SHARE in February).

The theme is highly topical: The management of uniprocessor LPARs. Topical because of the increased uniprocessor speed of the z990 and particularly the z9 processor families, which is leading customers to want to define LPARs with perhaps only a single logical engine.

It’s a thumpingly good read, covering a number of topics in the general areas of

• Changes in ready dispatcher to help SRM run work more efficiently in general
• WLM policy and MVS settings to help manage uniprocessor environment
• Managing CPU-intensive work
• z/OS functions which override WLM assigned dispatching priorities

The above list taken direct from her agenda foil.

The paper goes all the way back to the MVS/ESA “Reduced Pre-emption” line item. But it contains information about changes over all the years since. There are things in the paper and presentation that I didn’t know about – but I wouldn’t claim to be “omniscient in detail”.

You can get the paper from here. If questions or comments arise I’ll either handle them myself or point Linda to them. So do feel free to comment on the paper here.

(Originally posted 2007-03-20.)

I guess I must be part of the “ungrateful peasant” generation of computer geeks…

I never really liked FORTRAN. And now John Backus who led the team that developed it is dead.

For me the elegance of other languages biased my view – but that’s to miss the point: FORTRAN preceded them. And with all such pioneers elegance of form is often lacking.

I did have some practical acquaintance with FORTRAN: In 1984, when doing my final year Physics project I used it to do the “heavy lifting” (and an Acorn Atom to do the graphical lightweight display stuff). UCL at that time had a licence for the NAG set of mathematical libraries. As my project involved numerical integrations to solve Schrödinger’s Equation (in only 1 dimension) I turned to the NAG libraries. To use them I had to use FORTRAN. My practical experience of programming at that time had, to be fair, consisted of a mixture of Z80 Assembler, BASIC, and a home-grown programming language at “gap year” employer. So I think I had a reasonable view.

And that view, whether reasonable or not, was that FORTRAN was frustratingly restrictive, even by the standards of Assembly languages. But you could get much more done with FORTRAN than with Assembler more quickly. I expect one could’ve called the NAG routines from Assembler – but that would’ve been even more work.

So, I think FORTRAN gets a lot of stick. But it’s reasonable to suppose that all the other more elegant languages depended on the pioneer step that was FORTRAN.

Now, does anyone remember using FORTRAN (or even Assembler) on the 3090 Vector Facility? I think maybe 1 or 2 IBM customers in the UK did – but they weren’t in my “social circle” of customers I called on. The legacy today of the Vector Facility is a very large number of reserved opcodes, though I suspect we’re starting to use those now for other things. Does anyone know?