Thursday, July 22, 2010

IBM zEnterprise System : a new dimension in computing

General Introduction

Do you have a minute ? Or two ? Or perhaps a bit more ? This is not your regular every once in a while new mainframe announcement. Previous announcements were all in all pretty straightforward : we have a new Enterprise Class system and this time it’s called the zEnterprise 196. Surprise, no System z11 and here it is :

Click on image for larger version in new window

This time, however, it’s much more. Of course, there’s a great similarity with the z10 EC but this time the mainframe is just a part of the entire announcement, the IBM zEnterprise System. This complex consists of three fundamental elements as you can see below : the IBM zEnterprise 196, the IBM zEnterprise BladeCenter Extension (zBX) and the IBM zEnterprise Unified Resource Manager.

Click on image for larger version in new window

With this announcement, IBM delivers on its Dynamic Infrastructure vision that was designed during the past years. If you look e.g. at this presentation, you’ll see how IBM wants to create “a single management and policy framework across Web serving, transaction, and database” with a central role for the mainframe. Starting point for this is that we live in a heterogeneous world where a one size fits all approach for all workloads does no longer stand a chance. Each platform has its specific strengths. What IBM wants to do now is bring those platforms together into one system, the IBM zEnterprise System. The goal is to use the mainframe as a central system to manage the heterogeneous platforms which are connected to it. In the beginning we will be mainly talking about managing, setting up, monitoring and automating these systems. But in the end we should be talking about a common policy based Workload Management. This would be directed from the mainframe but executed on several platforms. Today we take the first steps along this route.

I will try and give you some insight in each component and the way in which they are interconnected.

IBM zEnterprise 196 (Model 2817)

First of all this remains of course the announcement (ZG10-0349) of an Enterprise Class mainframe, the successor to the System z10 EC. Although it keeps belonging to the System z family, it has a new name : zEnterprise 196, because it really represents a new generation. What does the 196 stand for : 1 for the first generation of the new mainframe and 96 because it now has 96 engines.

Of course I can only focus on some of the highlights of this system.


Technical specifications

Click on image for larger version in new window

There’s a strong resemblance to the System z10 EC. The IBM zEnterprise 196 has four regular models again and one large sized : M15, M32, M49, M66 and M80. We have 2 spare engines per machine and the rules for implementing specialty engines also remain the same.

For the first 15 CPs there’s sub-capacity again. The full engines (701-715) can be defined as 4xx (approx. 20%), 5xx (approx. 49%) or 6xx (approx. 64%). The specialty engines on such a system are of course always running on full capacity.

You have a minimum of 32GB of memory ranging to 768GB per book and up to 3TB for the entire z196. There have been some memory enhancements with the introduction of RAIM or Redundant Array of Independent Memory. You immediately see the resemblance with RAID (Redundant Array of Independent Disks). This should prevent further memory problems or outages.

Click on image for larger version in new window

The zEnterprise 196 again has a faster processor. It goes from 4.4GHz to 5.2GHz. It’s not as spectacular as the previous step forward but it’s enough in order to remain the fastest processor. There have also been some enhancements for the cache levels.

Click on image for larger version in new window

The Server Time Protocol (STP) remains optional, but do beware that connecting to External Timers is no longer possible with the z196.

Regarding connectivity nothing much has changed. Some cards are no longer supported but can still be carried forward like e.g. the FICON Express4 and the OSA-Express2 cards.

No spectacular changes for Capacity on Demand (CoD) either. A couple of things : the administrative On/Off Capacity On Demand test has been reintroduced. Because there was a lot of confusion about the installation of CoD records, 4 records will now be automatically ‘installed’ in the plant. For a lot of customers this will mean they will no longer have to run the process to ‘install’ their ‘staged’ records.

Upgrade paths are from the z9 EC or the z10 EC. Power consumption per book remains the same for a zEnterprise 196 compared to the z10 EC (1800 Watts).


What’s inside and physical planning

With the z10 BC the I/O drawers were introduced instead of the I/O cages. It’s a bit odd, but the z196 has kind of a mix of both. Depending on the necessary I/O slots you will have a mix of one or more cages (28 slots) and drawers (8 slots). As a customer you don’t have a choice. If you don’t have more than 32 cards, you will only have drawers. Otherwise, you will only have cages or a mix of both.

Physically the system has the same measurements as the z10 EC. Still, two optional features may influence the construction of the z196 and have an impact on your physical planning. The first feature is the optional ‘overhead cabling’ or 'Top exit' option. Cables will no longer go under the raised floor but are directed to the top of the machine. For this purpose cable towers are attached to both sides of the z196. This makes it 12 inches (about 30 cm) wider and 5.5 (about 14 cm) higher. Mind you, this does not include the power cables, as you can see on this picture.

Click on image for larger version in new window

Also the statement of direction about optional water cooling has been fulfilled. You choose this option at your initial order. There’s no MES to change this afterwards.


New software pricing : AWLC

IBM has developed a new pricing mechanism especially (and exclusively) for the z196 : Advanced Workload License Charge (AWLC) as the successor for VWLC. This new pricing mechanism will give “reduced price points, versus VWLC, for all MSU levels with more than three MSUs”. AWLC will be used for standalone as well as for sysplex pricing. For sysplex pricing it will only be applied when all systems within the sysplex are z196 systems. There will be a transition program for people who are gradually replacing their older systems by z196 machines.


IBM zEnterprise BladeCenter Extension (zBX)

General features

The zEnterprise BX (model 2458-002) (ZG10-0263) is a 42U Enterprise rack in which two 9U Blade Centers can be mounted. They fill up less than half the rack, but the reason is that IBM wants the zBX to comply to the same severe (RAS) conditions as the mainframe itself. A zBX can consist of 1 to 4 racks. This means a maximum of 8 Blade Centers with a total of 112 (8x14) blades. With this first release, IBM will only allow Power 7 Blades, with AIX as operating system. The servers running on these blades will always run in a virtualized environment. The communication with the various components is realized via the so-called Top of Rack (TOR) switches. These blades can also have their own I/O (disks) and network connectivity. The distance between the zEnterprise 196 and the zBX is limited to a maximum of 26 meter.


As you could see on the first general illustration, IBM makes a functional distinction between just blades and optimizers. Blades are running applications in a traditional manner as we have always known them. Optimizers are specific Workload accelerators that can give you a significant performance enhancement and/or a cost reduction per transaction (just like specialty engines on the mainframe do).The goal is always the same : optimizing workloads across different platforms. The optimizers are always pre-packaged.


IBM Smart Analytics Optimizer

For the moment there’s only one example of these optimizers which has been announced : the IBM Smart Analytics Optimizer for DB2 for z/OS (ZP10-0339) (ISAOpt). You know this one by now, it has been talked about a lot but now it’s getting announced as part of this new concept. I will not go into great detail on this ISAOpt, as you can already find lots of information about it elsewhere. But it’s quite clear that this is a very good example of this Dynamic Infrastructure set of mind and I think the illustration below clarifies a lot.

Click on image for larger version in new window

This is actually a data warehouse solution. The DB2 workload is really spread across different platforms. Some part of the transaction runs on the traditional processors, some part is executed on the zIIP and another part is running on the Smart Analytics Optimizer on the zBX. The blades on the zBX contain portions of the data warehouse in memory. This allows ultra fast access to those parts of the database and takes care of a real performance boost which would never have been possible with a standalone mainframe. The figures I’ve seen so far are really impressive starting at ten times (yes, not ten percent but ten times) better performance. And, as with the zIIP, offloading workloads to the ISAOpt may result in an additional cost reduction.


IBM zEnterprise Unified Resource Manager

Functionality


The description of the zEnterprise Unified Resource Manager will make it clearer how the communication between the zEnterprise 196 and the zEnterprise BX is set up and how the management of different components is happening. You might call the zEnterprise Unified Resource Manager the glue sticking all the components together. It’s installed on the HMC of the mainframe and has the following functionalities :
  • Operational management
  • Virtual Server management
  • Network management
  • Hypervisor management
  • Platform performance management
  • Energy management
Management means Installation, Monitoring, Optimalization, Diagnosis . . .

As you will see, with this first generation the Unified Resource Manager is mainly focusing on the zEnterprise BX with its Power blades.

Below you see a couple of illustrations showing the connectivity between the various components.

Click on image for larger version in new window


Click on image for larger version in new window


Connectivity

As you can see the Unified Resource Manager is installed on the HMC of the mainframe. Via new functionalities on this HMC all components can be managed. This is done through a 1Gb connection between the mainframe and the various virtual servers. Thus, every ‘Hypervisor’ (z/VM, PowerVM) can communicate with the zEnterprise Unified Resource Manager. For data transfer a 10Gb connection is used. Next to that, we still have the usual connectivity for the customer network.

But how do we implement this ? For the mainframe, communications are realized through the I/O cards. For the 1Gb Intra Node Management Network (INMN) two different connections on two different 1000BaseT cards must be defined in the IOCDS. To distinguish them from the rest, they get a new CHPID type : OSM. For the 10Gb Intra Ensemble Data Network (IEDN) again two different connections on two different 10Gb cards are defined, also with a new CHPID type : OSX. These connections run to the Power Blade Centers. The components communicating with the zEnterprise Unified Resource Manager are the purple blocks on the illustration above.

A couple of examples to make this a bit more clear.
  • The Intra Node Management Network is used by the Unified Resource Manager for installing the hypervisor PowerVM on an ‘empty’ blade. So this hypervisor is entirely installed, started/stopped and managed by the zEnterprise Unified Resource Manager. Once PowerVM is operational, virtual servers can be installed on it. Again this is completely initiated from and executed by the Unified Resource Manager.
  • The Intra Ensemble Data Network is a private network which is e.g. used by the Smart Analytics Optimizer to transfer data from the Blades to DB2 on z/OS.

Ensembles en Nodes


You saw these new terms a couple of times by now : nodes and ensembles. A node is a zEnterprise 196 with one or zero zEnterprise Bladecenter Extensions. When you combine one or more (maximum 8) nodes of which at least one contains a zEnterprise BX you create an ensemble. These nodes and ensembles are created by the zEnterprise Unified Resource Manager.


HMC

So, the zEnterprise Unified Resource Manager runs on the mainframe HMC. No wonder you’ll need a new HMC with feature code 91. And if you’re using the zEnterprise Unified Resource Manager you must always order two consoles. One of them will be used as primary console, the other is the alternate console. The first console on which you define an ensemble automatically becomes the primary console. The alternate console is used to back up all the management data the Unified Resource Manager is using. This console is automatically activated when there’s something wrong with the primary console.

Keep in mind that, just like the Smart Analytics Optimizer, the zEnterprise Unified Resource Manager only becomes available on November 19, 2010.


Statements of Direction

  • The zEnterprise 196 will be the last System z generation to support ISC-3 links. So, time to have a closer look at InfiniBand.
  • The zEnterprise 196 will also be the last generation to support ESCON. It might be possible that you will still be able to carry them forward to the next generation but I wouldn’t count on that. Also here : time to replace remaining ESCON channels or to have a look at the Optica Converters.
  • The zEnterprise 196 is planned to be the last high-end server to support dial-up modems for Remote Support Facility (RSF). If you haven’t done so, you might want to take a look at setting up a broadband connection.
  • In the first half of 2011 the zEnterprise Unified Resource Manager should support System x blades in the zBX with Linux on System x.

Key dates

September 10, 2010
  • z196 Models M15, M32, M49, M66, and M80
  • Water Cooling for z196
  • Manage suite for z196
  • z9™ EC upgrades
  • z10 EC upgrades
November 19, 2010
  • Manage suite enhancement functions for z196
  • Automate suite (#0020) for z196
  • IBM Smart Analytics Optimizer for DB2 for z/OS for z196 in blade quantities of 7, 14, or 28 POWER7 blade for z196 in a zBX
December 17, 2010
  • IBM Smart Analytics Optimizer enablement on z10 EC and z10 BC
  • IBM Smart Analytics Optimizer for DB2 for z/OS for z196 in blade quantities of 42 or 56
December 31, 2010
  • MES features for Models M15, M32, M49, M66, and M80

Documentation

Some IBM pages
Redbooks (some are still drafts)
Other sources
  • Here you find a virtual tour of the zEnterprise 196. You can e.g. see how an I/O card or a processor book is removed
  • I posted a little movie in my previous post. If you take a look over here, you'll find several others on Youtube.

Final remark : This is really a LOT of new stuff and I know I can only superficially touch upon all these new functions and features. So I hope you get at least some first impression of where IBM is heading with this new zEnterprise System. I will definitely write more on a lot of these aspects.

Next to that there were also announcements on z/OS 1.12, z/OSMF 1.12, z/VM 6.1, z/VSE 4.3, DS8700, PL/I for z/OS V4.1 etc. I'll also try and write some more about those.

But don’t worry, I’m off on holiday now, so you can take your time for this one . . . and . . . have a nice holiday too !

No comments: