Tuesday, July 23, 2013

zEnterprise BC12 - A technical Introduction

General Introduction

Really folks, if you want to know when a new System z box will be announced, don’t bother scrutinizing the social media, bribing IBM people or consulting Madame Soleil. Just ask me when I’m planning my holidays. Take the first Tuesday and you’ll be spot on.

So today IBM announces the IBM zEnterprise BC12 or the zBC12. This is the announcement : 'IBM zEnterprise BC12 (zBC12) extends modern mainframe capabilities for enterprises of all sizes to help cut costs, improve service, and fuel innovation for business growth (ZG13-0195)'.

No new naming, it's just plain and simple as predicted : zBC12. I just wonder what IBM will come up with for the next generation. What was that song again from Stevie Wonder. Oh yes, Superstition.

I'll give you the usual survey of the new system starting with some technical specifications. There are also a couple of entirely new elements like zEDC, RoCE and the IFP. The Software pricing is always an interesting part and there are quite a lot of statements of direction too. I'll conclude with some key dates and a couple of references to already available documentation.

As usual, along with the announcement of the new Business Class model there's also the GA2 for the Enterprise Class model 'IBM zEnterprise EC12 enhancements enable data, cloud, and security-ready capabilities (ZG13-0192)'.

Technical specifications

The zBC12 gets machine type 2828. Just like its BC predecessors it's also a one frame, air-cooled machine. Just like the z114 it has top-exit I/O cabling and top exit power cabling. Allthough I hear that not every one is happy with the extra chimneys that are added, making it difficult to fit it into its usual place or in the hot aisle/cold aisle strategy. And just to warn you : you cannot order just top exit power cabling. This immediately includes the top exit I/O cabling ànd the chimneys too.

Again, we have two models : the H06 and the H13. The CPC drawers have 9 PU cores. The H06 has one drawer (9 PUs) and the H13 has two drawers (18 PUs). As you can guess the H06 has 6 configurable engines, the H13 has, yes, 13 of them. This means that an extra engine has been added to the base model, because previous generations had 5 configurable traditional engines. So what about the other PUs ? Well, each model has 2 SAPs and an IFP (Integrated Firmware Processor). I'll come back to that one later. Just like with the z114 the H13 also has 2 dedicated spares.

Both models can have up to 6 traditional CPs ranging from A01 to Z06 giving us 156 capacity settings instead of the usual 130 settings.

 Click on image for larger version in new window

You know what PCI stands for in the graph. But have a good look at the lowest setting. For the last years this has always been around 25 mips, oh, sorry, 25 PCIs. The lowest model is still called A01 but it now has 50 mips. This is apparantly due to the increased clock speed by which it's not possible to go under this 50 mips limit. But there's some good news here too. The zBC12 A01 still has zELC pricing and even though it now has 6 MSUs, the software price for the A01 remains the same as for the z114 A01 with its 3 MSUs.

Other engines can be defined as specialty engines. Here's an overview of the possibilities.

Click on image for larger version in new window

The memory ranges from 8 GB up to 240 GB for the H06 and from 16GB up to 496GB of usable memory for the H13. This doubles the maximum capacity for both models. Up to 112GB we have 8GB increments from then onwards it's 32GB. I have the impression that we have larger ranges of memory that allow for concurrent upgrade without the need for Plan Ahead Memory. I like that, because when we take a look at the last two generations, you can almost be sure that the zBC12 will be End of Marketing by half 2016 for MES upgrades and by half 2017 for LIC upgrades. Be sure to get hold of these ranges before ordering. As an example I give you the ranges for the H06. You might feel a lot more comfortable by ordering e.g. 56GB instead of 48GB of memory.

Upgrades are possible from any z10 BC or z114. You can also (disruptively) upgrade from an H06 to an H13. Only the H13 allows an upgrade to the larger zEC12. The processor now reaches 4.2GHz.

Under the covers

Here you see an 'under the covers' illustration of a zBC12 Model H13.

As you can see this is pretty much the same as with the z114. Just a quick remark : you can still carry forward a traditional I/O drawer. Empty slots however cannot be filled after the upgrade to the zBC12. All new cards go into the PCIe I/O drawer that was introduced with the z114. Here's a reminder of the differences between the traditional I/O drawer and the PCIe I/O drawer.

And this is the new PCIe I/O drawer.


Here's an overview of the current connectivity features.

Click on image for larger version in new window

If you take a close look at this overview, you'll see a couple of new things : The OSA-Express5S cards, the 10GBe RoCE Express card and the zEDC Express card. And one reminder : No more ESCON. So, have a look at the Optica FICON Converters for that.

So, first, what about the OSA-Express5S cards. Apparantly nothing much new. It seems to be merely a matter of naming. All OSA cards are now Express5S but the performance remains the same as with the Express4S cards.

New : 10GBe RoCE - RDMA over Converged Enhanced Ethernet

First, let's get the pronuciation right. What's your favorite Rocky film ? That's it, you nailed it. It's the Rocky card.

I must say I had to turn to Wikipedia for a definition of RDMA (Remote Direct Memory Access) : "RDMA supports zero-copy networking by enabling the network adapter to transfer data directly to or from application memory, eliminating the need to copy data between application memory and the data buffers in the operating system. So it is a transparent mapping via the network. Such transfers require no work to be done by CPUs, caches, or context switches, and transfers continue in parallel with other system operations. When an application performs an RDMA Read or Write request, the application data is delivered directly to the network, reducing latency and enabling fast message transfer".

An example ?  WebSphere accessing a database using TCP/IP requires CPU resource consumption. Therefore TCP/IP communications are executed in software which consumes processor cycles. Lots of applications today use TCP/IP to communicate server to server. With this new feature Shared Memory Communications via RDMA (SMC-R) can be implemented on a RoCE card that can communicate memory to memory, avoiding the CPU resources of TCP/IP, thus reducing network latency and improving wall clock time. What you get is a 'Hipersockets-like' performance across physical processors on different CECs.

Other exploiters are :
  • Any TCP/IP Application, without change.
  • WAS type 4 connections to remote DB2, IMS, CICS instances
  • Cognos to DB2 connectivity
  • CICS to CICS connectivity via IPIC
  • Optimized Sysplex Distributor intra-sysplex load balancing
The applications themselves do not have to be changed.

The technical details :
  • Only one of the two 10 GbE SFP+ ports enabled per feature
  • Each feature must be dedicated to one LPAR. (Reconfiguration is supported)
  • Minimum configuration of two features per LPAR is recommended
  • A switched connection requires an enterprise-class 10 GbE switch
  • Point-to-point connection is supported
  • Not defined as a CHPID and does not consume a CHPID number
  • Exclusive to zEC12 GA2 and zBC12
  • z/OS V2.1 Exploitation
You can find more information over here where you can also find a  2-page brochure on the topic.

    New : zEDC - IBM zEnterprise Data Compression

    zEDC is all about compression. Why, will you ask, do we need a compression card. System z has already hardware compression on CPU level which is working just fine. Apparent reasons for the zEDC card are the offload of CPU cycles to this card and the fact that you will start to compress more than you did in the past. More compression also means less disk usage. Using this card (accelerator) also means you will not use extra compression software (read, yet again, CPU cycles) either.

    It's also a different kind of compression. The System z hardware compression is an IBM proprietary compression while the zEDC compression is not. The code scans text and looks for the  re-use of a phrase and puts in a back reference.  For large files, zEDC can be very efficient.  As said, compression is done on the card, not using CPU resources.

    Planned exploiters are SMF, BSAM, QSAM and Java 8.

    There is also a (free) tool available which can help you estimate the possible utilization of one or more zEDC features : the Batch Network Analyzer (BNA).

    Some practical details : there is one coprocessor per PCIe I/O feature card. It supports concurrent requests from up to 15 LPARs when given inputs of 1 MB or larger. Up to 8 features are supported by zEC12 and zBC12. A minimum of two features is recommended. You'll also need z/OS 2.1.
    The zEDC card must be defined in the IOCP. I won't go into detail here and I'm sure the IOCP User Guide will cover all you you need to know about this.

    And what if you have an older D/R machine or you need the data on a system which still has z/OS 1.12 or 1.13. Well z/OS 1.12 and 1.13 support decompression of the compressed data. Of course there you will need CPU cycles to make this happen. So, if you want to use zEDC, perhaps it's best when all the systems sharing the data are zEC12 or zBC12 with z/OS 2.1.

    As with the RoCE card this feature will not become available on previous System z generations.

    You can find more information over here where you can also find a 2-page brochure on the topic.

    New : IFP or Integrated Firmware Processor

    Here's also were the Integrated firmware processor (IFP) I mentioned at the beginning kicks in. The IFP is allocated from the pool of available PUs. Actually its use is completely transparent for the customer. You can't choose to install an IFP nor do you have to pay for it. It comes with the box and it's used when you have a 10GbE RoCE or a zEDC card installed. It has two Resource Groups (RGs) which have firmware for the 10GbE RoCE and zEDC Express features. It's initialized at POR if these features are present. I'm not exactly sure what this means when one of these cards is added later. Do you need a POR then too ?

    Other hardware features

    Just to be sure I want to mention that IBM zAware and Flash Express which were first introduced on the zEC12 are now of course also available on the zBC12.

    Further the HMC supports FC0091 (introduced with the zEC12) and FC0092. Older HMCs (FC0090 and earlier) can not be upgraded to control zEC12 or zBC12.

    What about zBX and zManager. The zBX Model 003 which was announced with the zEC12 becomes now also available for the zBC12. There are two major enhancements :

    Software pricing

    It just occurred to me that with the announcement of the z114 I had some good news and some not so good news for certain customers. Well, it's a bit similar this time except that the not so good news is restricted to customers who are now already two generations behind.

    Let's start with the good news. Similar to what happened with the zEC12 we don't see a new pricing mechanism but an 'update' to the AEWLC pricing. It means that for MLC pricing there will be a reduction of 5% (below 46 MSUs) or 4% (above 45 MSUs). On the zEC12 some (older) softwares were excluded from this. I guess this will be the same for the zBC12 but in general all important softwares are included. The announcement for this can be found here : 'Technology Update Pricing for AEWLC offers price-performance advantages for IBM zEnterprise BC12 servers (ZP13-0385)'. The list of eligible softwares for the reduction is also included in this announcement. This reduction does not apply to IPLA (OTC) softwares.

    As I already mentioned the A01 has 3 extra MSUs but remains on zELC pricing. This means the price remains the same as on previous A01 models.

    What's the not so good news ? Like last time with the z114, IBM will now apparently impose another increase on the EWLC priced software. This time it should only be a 3% increase. This impacts the z10 BC and older machines.

    Operating Systems

    • z/OS V2.1 with PTFs (Exploitation plus zEDC Express and 10GbE RoCE Support) 
    • z/OS V1.13 and V1.12 with PTFs (Exploitation) 
    • z/OS V1.11 with PTFS (Toleration, Lifecycle extension required)
    • z/OS V1.10 (Toleration, Lifecycle Extension required)
    As you will have already noticed, z/OS V2.1 will become essential for any new
    functions that are announced now or will be announced in the future. z/OS 2.1 is also officially announced today. You can read the entire announcement over here (ZP13-0371).

    • SUSE SLES 10 and SLES 11 
    • Red Hat RHEL 5 and RHEL 6
    • z/VM V6.3 for exploitation support
    • z/VM V5.4 and V6.2 – toleration support
    • z/VM V6.1 is End of Support since April 2013
    z/VM 6.3 ? Yes, it is also announced today. The announcement is over here (ZP13-0376).
    • z/VSE 4.3
    • z/VSE 5.1

    zEC12 GA2

    Let me just summarize the most important functionalities that also become available on the zEC12.
    • Integrated firmware processor (IFP)
    • LPAR absolute physical capacity setting
    • IBM zAware extensions
    • FICON – 24k devices per channel
    • OSA-Express5S 10 GbE, GbE, and 1000BASE-T
    • 10GbE RoCE Express
    • zEDC Express
    • CFCC Level 19 – Thin Interrupt Support to enable production use of shared logical processors in some cases
    • CFCC Level 19 – MQ Shared Queue overflow Flash Express Exploitation
    • STP – Improved SE time accuracy

    Statements of direction

    There are a lot of statements of direction. So I'll only include some of them here.
    • z/VM support of 10GbE RoCE Express:
      IBM plans to offer future z/VM support for guests to exploit the 10GbE RoCE Express (FC0411) feature on the IBM zEnterprise EC12 and IBM zEnterprise BC12 systems. 
    • Removal of support for connections to an STP Mixed CTN
      The IBM zEnterprise EC12 and the zBC12 are the last System z servers to support connections to an STP Mixed CTN. After that the Sysplex Timer (9037) can no longer be used.
    • IBM System z Integrated Information Processor (zIIP) and IBM System z Application
      Assist Processor (zAAP) simplification

      The IBM zEnterprise EC12 and the zBC12 are planned to be the last System z servers to offer support for zAAP specialty engine processors. IBM intends to continue support for running zAAP workloads on zIIP processors ("zAAP on zIIP"). This is intended to help simplify capacity planning and performance management, while still supporting all the currently eligible workloads.
    • CFCC Level 19 Exploitation of Flash Express
      IBM intends to provide exploitation of the Flash Express feature on zEC12 an zBC12 servers with Coupling Facility Control Code (CFCC) Level 19 for certain Coupling Facility structures in the first half of 2014. This new function is designed to allow list structure data to be migrated to Flash Express memory as needed when the consumers of data do not keep pace with its creators for some reason, and migrate it back to real momory to be processed. When using WebSphere MQ for z/OS Version 7 this new capability is expected to provide significant buffering against enterprise messaging workload spikes and provide support for storing very large amounts of data in shared queue structures, potentially allowing several hours of data to be stored without causing interruptions in processing.
    • zEDC Express exploitation by z/OS for DFSMSdss and DFSMShsm
      IBM intends to provide exploitation of the zEDC Express feature for DFSMSdss and DFSMShsm by the end of third quarter 2014.
    • z/VM guest exploitation of zEDC Express feature
      In a tuture z/VM deliverable IBM plans to offer z/VM support for guest exploitation of the IBM zEnterprise Data Compression (zEDC) Express feature (#0420) on the IBM zEnterprise EC12 and IBM zEnterprise BC12 systems
    You can find all the SODs over here.


      Apparantly, it's still a bit early to come with full documentation, but this is what is already available for the moment.

      Redbooks (drafts) :
      Web :

      Some key dates for the zBC12 and zEC12 GA2

      July 23, 2013
      • Day of announcement
      • First day for GA orders
      • Resource Link support available
      September 20, 2013
      • New zBC12 models H06 and H13
      • Upgrades from z114 or z10 BC
      • Field-installed features and conversions that are delivered solely through a modification to the machine's Licensed Internal Code (LIC)
      September 30, 2013
      • z/OS V2.1 zEnterprise Data Compression (zEDC) acceleration for SMF log data
      4th Quarter 2013
      • Target availability of the Batch Network Analyzer (BNA) tool.
      October 31, 2013
      • MES features with already existing PCIe I/O drawers like zEDC
        Express, 10GbE RoCE Express, OSA-Express5S, Crypto Express4S, Flash
        Express and FICON Express8S
      December, 2013
      • z/OS V2.1 zEDC acceleration for extended format BSAM/QSAM sequential data
      December 31, 2013
      • All other MES features
      As usual, I could only scratch the surface and touch upon the highlights of this new system.
      There's much more to discover about it. So I'd say : stay tuned !

      No comments: