Tuesday, August 28, 2012

The new IBM zEnterprise EC12 - A technical introduction

General Introduction

Ok, ok, I’ll anticipate a bit. I’m sure some of you will remark when they’ve read through this : is that all ? Does that justify a new System z announcement ? IBM zAware, Flash Express, some kind of discount on the MLC pricing and … oh, that’s it ? Well, I guess it is. Or not ? It's definitely not what people at IBM will tell you when you make this remark. They really invested a lot of money and people into this system and it really has a lot more to offer than you would think at first sight.
Well, you know what, let me walk you through it and at the end you can let me know which side you are choosing.

The new system that’s being announced today is the zEnterprise EC12 or zEC12 (machine type 2827) : ‘The IBM zEnterprise EC12 - proven hybrid computing designed to manage multiple workloads, with the simplicity of a single system (ZG12-0262)’. Although dimensions haven’t really changed, the zEnterprise has a completely new look.

Click on image for larger version

What about the name ? If you follow the logic of the zEnterprise z196, this should be the zEnterprise z2A1, but it isn’t. It’s the zEnterprise EC12. zEnterprise is clear, this is the new brand name since it was introduced in 2010. Apparently EC pops up again for the Enterprise Class Models. And we seem to have picked up the count again where we left off after the z10, skipping 11 and now resuming at 12. Hence : zEnterprise EC12 or zEC12. Thank God IBM keeps its naming conventions transparent.

I’ll first give you the general technical details. Then we’ll have a look under the covers with some special attention to connectivity and physical planning. Then we move on to the new features : Flash Express and IBM zAware. Other topics covered are Operating Systems, zBX, the HMC, software pricing, Statements of Direction, documentation and some key dates.

Technical specifications

Let’s start by giving you an overview of the most important technical specifications with all new elements indicated in blue.

Click on image for larger version

Models and sub-capacity settings

Similar to the z196 the zEC12 has four regular models and one large sized model : H20, H43, H66, H89 and HA1 (the ‘A’ standing for ‘10’, so ‘A1’ stands for ‘101’). For the regular models, every book has 27 PUs with the HA1 each book has 30 PUs.
There are 2 spares designated per system. We see 4 SAPs per book. So if we take e.g. the H20 with 27 PUs, we subtract the 4 SAPs and the 2 spare processors, that leaves us with, hey, 21 PUs. Well, each system has 1 PU which is reserved for future use. And, no, don’t ask, I really don’t know what it will be used for. As usual processors can be defined as Central Processors (CPs), ICFs, IFLs, zAAPs zIIPs and optionally as additional SAPs.
A full processor (the 701 or a specialty engine) has a capacity of 1.514 mips as opposed to 1.202 on the z196. We have again three sub-capacity levels (4-, 5, -6) now for up to 20 CPs.

The zEC12 supports two different internal I/O infrastructures : the InfiniBand I/O infrastructure first introduced on the z10 supporting the 6 GBps InfiniBand I/O interconnect and the PCI Express Generation 2 (PCIe Gen2) I/O infrastructure introduced with z196 and z114 supporting the 8 GBps PCIe I/O interconnect. Only the last one is available on new systems.


In general we have the same RAIM Memory design as introduced on the z196. The system minimum is 32 GB and goes up to 768 GB per book and up to 3 TB for the entire system. This means the same maximum as on the z196. You can go up to 1 TB per LPAR. The fixed HSA is now set at 32 GB.

Where are the improvements ?
First of all the Out of Order core execution has become more performant than on the z196. There’s the introduction of 2 GB page frames, increasing efficiency for DB2 buffer pools, Java heap and other large structures. Decimal format conversions enable broader exploitation of Decimal Floating Point facility by COBOL programs.

And there’s the improved cache, too. A small comparison :

  • L1 private 64k i, 128k d
  • L2 private 1.5 MB
  • L3 shared 24 MB / chip
  • L4 shared 192 MB / book
  • L1 private 64k i, 96k d
  • L2 private 1 MB i + 1 MB d
  • L3 shared 48 MB / chip
  • L4 shared 384 MB / book

And next to that, there’s the additional Flash Express, which I will discuss later in this post.

Processor speed

Processor speed has now gone from 5.2GHz on the z196 to 5.5GHz on the zEC12

Click on image for larger version


In short you can upgrade from any z10 EC or any z196 to a zEC12.

When a z196 with a zBX is upgraded to a zEC12, the zBX Model 002 is converted to a Model 003. Mind you, this takes a considerable amount of additional planning. So be sure to walk this through with BP or IBM Subject Matter Experts.

Under the covers

Here’s a picture of how an air cooled model might look like on the inside

Click on image for larger version

The cooling has been changed for this new system. As far as I understand it for the moment, you have ‘fill and drain’ kits for both air and water cooled models. The radiator is filled with some kind of cooling fluid which is used for cooling the cores in the book. This fluid is cooled with water in the water cooled models and with air in the air cooled models. It’s an entirely new system which is completely incompatible with the z196.

Drawers and connectivity

Must I remind you of the fact that ESCON is no longer supported ?
Let’s have a look at the other card types.


Well, nothing spectacular concerning FICON. The FICON-Express8 card can be carried forward and the FICON-Express8S card for the PCIe drawer was already available. There is some improvement with zHPF as the 64k byte limit is eliminated allowing the card to fully exploit its available bandwidth.


There’s now also a new OSA-Express4S 1000Base-T card for the PCIe I/O drawer.

The card has 2 ports accounting for 1 CHPID.

There’s nothing much to tell about the other OSA-Express cards. It’s pretty straightforward as already saw on the illustration above.


The new Crypto Express4S card has one adapter per feature and you initially have to order two features. There are three configuration options for this PCIe adapter, but only one configuration option can be chosen at any given time. Switching between configuration modes will erase all 'card secrets'. The three options are : Accelerator, Secure IBM CCA coprocessor (default) and IBM Enterprise PKCS #11 (EP11) coprocessor, which is new.

Just a final remark on connectivity in general : there are limitations on carrying forward I/O cages and drawers. A maximum of 45 Non-PCIe Feature Cards can be carried forward following strict rules. So, do your planning carefully if you still have a lot of these cards.

Physical Planning

Measurements, weight, clearance

As far as I can see, as good as no changes regarding the footprint. There will apparently be identical floor cutouts for the zEC12 as for the z196 with no significant increase in weight. The depth of the system with covers will increase by 64 mm or 2.52 inches.

Top exit power cabling and non-raised floor option

The overhead Power option is shipped separately and installed on-site to allow for door clearance. In some cases this can avoid ordering the machine with height reduction saving the IBM CE some extra work.

The only thing for which there is no overhead support is for the water supply on water cooled systems. So it also speaks for itself that for the new non-raised floor models, only air cooled systems will be available.

Let’s move on to the really new features now.

Really New 1 : Flash Express

As a picture tells more than a thousand words, I guess this illustration shows you immediately what Flash Express is about :

Flash Express is used to deliver a new tier of memory, which is indeed slower than real memory but which is a lot faster than external disk. So, Flash Express introduces Solid State Drive (SSD) technology to the zSeries family. An operating system, such as z/OS, will be able to access blocks of flash storage as storage locations within a logical partition.

The main application of internal Flash in the zEC12 GA-1 is paging store for z/OS. z/OS paging subsystem will work with a mix of internal Flash and External Disk. This leads to improved paging performance realized through a simplified configuration. That way Flash Express can improve availability and reduce latency during transition periods and spikes. It helps accelerate start of day processing. It enables faster snapshots of diagnostics (e.g. SVC dump, standalone dump).

With pageable large (1MB) pages it can improve performance of DB2 and Java making it also ideal for applications with random read access and high read/write ratios.

Flash Express is implemented via NAND Flash SSDs (Solid State Drives) mounted in PCIe Flash Express feature cards.

They plug into PCIe I/O drawers in pairs. They are cabled together to form a Raid 10 Mirror for redundancy thus providing 1.6 TB of storage.

A maximum of 4 pairs is allowed per system.

Internal Flash is accessed using the new System z architected EADM (Extended Asynchronous Data Mover) Facility, an extension of the ADM architecture used in the past with expanded storage. Definition in IOCDS is not required.

You assign flash memory to partitions like main memory by minimum/maximum memory amount, not by feature. Each partition’s flash memory is isolated like main memory. You can dynamically increase the partition maximum amount of flash or dynamically configure flash memory into and out of the partition.

For the moment Flash Express is only used by z/OS 1.13 (4 GB LPAR or larger) plus PTFs. In the future it may also be exploited by Linux on System z … For availability, check the key dates below.

Really New 2 : IBM zAware

IBM zAware grew from the awareness (no pun intended) that systems are changing very fast and are gradually getting more and more complex. It becomes a real challenge to detect problems soon enough before they cascade and lead to system failures at some or other point. The challenge lies in the ability to identify the causes of system anomalies and other problems proactively and repair them as quickly as possible, within SLA conditions before the system lets you down. That’s why IBM developed IBM zAware (IBM System z Advanced Workload Analysis Reporter).

Here’s an illustration of how IBM zAware monitors your sysplex(es).

Click on image for larger version

Let’s have a look at the several components on this picture. IBM zAware is running in its own LPAR (on a CP or an IFL), kind of acting as e.g. a coupling facility does. Therefore it is also defined from the HMC. So it needs its own memory as well and this is estimated to start at 4GB and an additonal 200MB per monitored system. It also requires its own DASD and connectivity via HiperSockets or OSA ports. It can monitor any System z Server running z/OS 1.13 + PTFs. However, it can only run on a zEC12.

IBM zAware goes through massive amounts of OPERLOG messages (including all z/OS console message, ISV and application generated messages) and turns this into useful information. It creates a 90 days baseline. Starting from this baseline it performs machine learning, pattern recognition, and statistical analysis on the new streaming messages. It looks for unexpected patterns detecting unusual behavior of z/OS images in near real time.
The results are presented in a single browser based view showing the health of your Parallel Sysplex via a relative weighting and color coding as you can see in the example below.

The colors give you an indication on the timeline of possible problems with their weight and you can drill down on any problem (color) in order to get more details about it.
IBM zAware can also provide information in XML format that can feed other processes or tools.

There are some overlaps with other products like e.g. z/OS Health Checker, z/OS PFA and z/OS RTD. Still it goes one step further because it’s a self learning pattern based message analysis tool able to diagnose complex z/OS problems across your sysplex reporting problems that might bring down your system(s).

IBM zAware is a fee-based feature based upon the number of z/OS images it is monitoring. The good news : IBM zAware licenses for DR machines are included in the base purchase, so no extra cost there.

Operating Systems

z/OS Support Plan :
  • z/OS V1R10 with PTFs (Toleration, Lifecycle Extension Required)
  • z/OS V1R11 with PTFs (Toleration, Lifecycle Extension required)
  • z/OS V1R12 with PTFs (Exploitation of some functions like e.g. Crypto Express4S)
  • z/OS V1R13 with PTFs (Exploitation of most functions plus Flash Express and IBM zAware Support)
  • z/OS V2.1 with Full Exploitation in base
z/VM Support Plan :
  • z/VM 5.4, 6.1, and 6.2 will support zEC12 in compatibility mode with PTFs
  • Exploitation support to be provided with z/VM 6.3 (to GA after zEC12)
z/VSE Support Plan:
  • z/VSE plans to provide compatibility support at zEC12 GA1 for:
  • z/VSE 4.3 with PTFs
  • z/VSE 5.1 with PTFs
Linux for System z Support Plan :
  • IBM intends to support zEC12 with the following distributions:
  • SUSE SLES 10 and SLES 11
  • Red Hat RHEL 5 and RHEL 6

zBX and hybrid computing


When you take a look at the zBX Model 003, apparantly there was no money left to do a restyling for this box too. As a matter of fact, when you take a closer look at new functionalities or features, there’s nothing much that really accounts for introducing a new model at all. There’s no big deal happening under the covers either. To put it bluntly : no new types of blades (not even for the DataPower XI50z), just some extra connectivity for redundancy and throughput reasons, and that’s about it. Does that justify the extra amount of planning and work it takes for the forced upgrade from model 002 to model 003 when upgrading to a zEC12. I’m just not sure about it.

So what ís new here ?
For the Unified Resource Manager dynamic discovery and configuration of storage is added. There’s a fulfillment of the statement of direction to implement new Unified Resource Manager APIs to enable discovery, monitoring and management of ensemble resources using external tools. As an example, IBM Tivoli is using these APIs.

For connectivity the 10 Gb cables between BladeCenter 10Gb switch and 10 Gb TOR (Top of Rack) switch are doubled enabling a potential of 20 Gb Ethernet bandwidth via link aggregation.

By the way, if you should have a IBM Smart Analytics Optimizer (ISAOPT), it is no longer supported on the Model 003. You will have to migrate to the IBM DB2 Analytics Accelerator (IDAA) which is the separate Netezza box. If this was all you were running on the zBX, you probable won’t even need a Model 003.

I’ll finish this part with a spark of good news : at least there is no cost for upgrading from Model 002 to Model 003 and transferring zManager from z196 to zEC12.

And of course the zEnterprise keeps its enormous wealth of possibilities that were already introduced since it's first announcement back in 2010.

Operating system support on zBX

Support for Linux environments on System x blades in zBX
  • 64-bit version support only
  • Red Hat RHEL 5.5, 5.6, 5.7, 6.0 6.1
  • SUSE Linux Enterprise Server (SLES) 10 (SP4) and SLES 11 SP1
Support for Windows environments on System x blades in zBX
  • Microsoft Windows Server 2008 R2 and Microsoft Windows Server 2008 (SP2)
  • For either Windows the Datacenter Edition is recommended
Support of AIX environments on POWER7 blades in zBX
  • AIX: AIX 5.3 Technology Level 12 or higher,
  • AIX 6.1 Technology Level 5 or higher, AIX 7.1


It’s simple : only HMCs with feature code FC0091 are allowed on the zEC12. When carrying them forward previously supplied FC 0091 consoles are upgraded to the newly required 16 GB with Driver Level 12 adding an additional 8GB to them. FC0091 shipped prior to zEC12 have only 8GB.

For those still using dial up connections : HMC application LIC for zEC12 and zBX Model 3 will NOT support dial modem use. Modems on installed HMC FC0091 hardware will no longer work and future HMCs will no longer have modem hardware.

Another point of attention : the HMC LAN switch (FC0070) which could be supplied with former systems is no longer provided. They can be carried forward or the customer must acquire the switches separately to implement the required HMC LAN connectivity.

Related announcement : 'Hardware Withdrawal: Hardware Management Console (HMC) (ZG12-0244)'

Software Pricing

The software pricing is pretty straightforward this time. It may come as a surprise, but the pricing metric remains the same for MLC : AWLC. Still, there’s a benefit of on average 5%. How is it realized ? Quite simple, there’s a reduction on some softwares depending on the MSUs.

Let me give you an example. You have a zEC12 with a machine capacity of 1.500 MSUs and your SCRT reports 1.300 MSUs for z/OS and in a smaller partition 300 MSUs for Cobol. Well, it’s not the reported MSUs that count. Since your machine has 1.500 MSUs, you’ll get a reduction compared to the z196 of 5% on those softwares. So, no need to make elaborate calculations starting with 2% on the first 45 MSUs, then 4% from 46 to 315. No, plain and simple 5% because the machine has 1500 MSUs. You see, even software pricing can be easy from time to time.

More information on software pricing can be found in a separate announcement : 'Technology Update Pricing for AWLC offers price-performance advantages for IBM zEnterprise EC12 servers (ZP12-0418)'. Here you also find the list of software to which the reduction is applicable. Transition offerings are available for mixed sysplex environments.

Statements of direction

The zEC12 will be the last high server to support connections to an STP Mixed CTN. This includes the Sysplex Timer(R) (9037): After zEC12, servers that require time synchronization, such as to support a base or Parallel Sysplex(R), will require Server Time Protocol (STP), and all servers in that network must be configured in STP-only mode.    

Removal of support for Ethernet half-duplex operation and 10 Mbps link data rate: The OSAExpress4S 1000BASE-T Ethernet feature is planned to be the last copper Ethernet feature to support halfduplex operation and a 10 Mbps link data rate. The zEC12 servers are planned to be the last IBM System z servers to support half-duplex operation and a 10 Mbps link data rate for copper Ethernet environments.   

Removal of ISC-3 support on System z: The zEC12 is planned to be the last high-end System z server to offer support of the InterSystem Channel-3 (ISC-3) for Parallel Sysplex environments at extended distances. ISC-3 will not be supported on future high-end System z servers as carry forward on an upgrade.  
The same applies for OSA-Express3 support and FICON Express4 support.

IBM System z Integrated Information Processor (zIIP) and IBM System z Application Assist Processor (zAAP) simplification: IBM zEnterprise EC12 is planned to be the last high-end System z server 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. In addition, IBM plans to provide a PTF for APAR OA38829 on z/OS V1.12 and V1.13 in September 2012 to remove the restriction that prevents zAAP-eligible workloads from running on zIIP processors when a zAAP is installed on the server.


IBM Announcement page.
Here you'll also find more product details, the data sheet and an elaborate FAQ, which also gives you an idea of some price features

The Technical manuals will be available in the Library section of Resourcelink.

Two redbook drafts are available : the IBM zEnterprise EC12 Technical Introduction and the IBM zEnterprise EC12 Technical Guide.

I'll come back with a larger overview as soon as all documentation is fully available.

Key Dates

August 28, 2012 :
  • Announcement Day
  • First Day Orders for GA Systems
September 4, 2012
  • zEC12 Models H20, H43, H66, H89, and HA1
  • z10 EC and z196 upgrades
  • z196 with zBX model 002 upgrades to zNext with zBX model 003
  • IBM zAware
  • Flash Express - orderable August 28, 2012, and deliverable September 19, 2012. Planned availability of z/OS exploitation of functionality is December 14, 2012
December 7, 2012
  • Model conversions for IBM zEnterprise EC12
December 14, 2012
December 31, 2012
  • MES features for Models H20, H43, H66, H89, and HA1
  • zBX model 003 move from one zNext to another zNext
  • z/OS support for 2 GB large pages and dynamic reconfiguration for Flash Express
And to wrap things all nicely up, here's a video that walks you through all components of the new system

No comments: