Wednesday, October 22, 2008

z10 BC (part 2) and z10 EC GA2

In my previous post I discussed the new functionalities for the System z10 BC. As part two of the z10 BC announcement I’m going to comment on some of the features that are announced simultaneously for the z10 BC (‘IBM System z10 Business Class - The smart choice for your business. z can do IT better (ZG08-0806)’) and the z10 EC (‘IBM System z10 Business Class - The smart choice for your business. z can do IT better (ZG08-0843)’. As you’ll notice on the picture below, there’s much more than what I can discuss in one post, so I guess I’ll talk about some of the other features in the near future.

Click on image for larger version in new window

OSA-Express3
With the z10 EC we saw the first announcement of one OSA-Express3 card (10GBe). The range of OSA-Express3 cards has now been largely extended. The cards also have twice as much ports as on the OSA-Express2 cards. This means less CHPIDs, less I/O slots and perhaps less I/O cages (z10 EC) or I/O drawers (z10 BC). An overview :
  • The z10 OSA-Express3 – 10 GbE card is either Long Range (LR) with single mode fiber or Short Range (SR) with multimode fiber. There are 2 ports per card.
  • The z10 OSA-Express3 GbE comes in a Long Wave (LX) and a Short Wave (SX) version. The cards have 4 ports and for the z10 BC there’s a Short Wave (SX) card with 2 ports.
  • The z10 OSA-Express3 1000Base-T card keeps the same functionalities as its predecessor (e.g. OSA Integrated Console Controller). The card has 4 ports and again there’s a 2 port version for the z10 BC.
Z10 High Performance FICON for System z (zHPF)
If you thought that the FICON performance wizards had pulled the last trick out of their sleeves with MIDAW, then think again. They now have come up with z10 High Performance FICON (zHPF) in order to substantially reduce channel overhead once again. It is mainly with small block sizes that the performance is largely improved, in other words mainly for online transaction processing.
It’s only available on the z10 and you have to be at least on z/OS 1.8. There’s a ZHPF=YES/NO parameter in the IECIOSxx member in the SYS1.PARMLIB which has to be activated (default is NO). For the moment it will only work with the DS8000 with R4.1 as it’s the only Control Unit which support this feature for the moment. This DS8000 enhancement was also announced yesterday : IBM System Storage DS8000 series (Machine type 239x) function authorization for zHPF and RMZ resync (ZG08-0908).

Click on image for larger version in new window

Capacity on Demand
‘Capacity on Demand’ has had many enhancements over the last couple of years. The z10 BC offers the same possibilities as the z10 EC : Capacity Backup (CBU), Capacity for Planned Events (CPE) and On/Off Capacity on Demand (OOCoD). Still some new functionalities are introduced for both systems.

There’s a whole Redbook describing every variation, but put simply you can order one or more records for each of these offerings. That way, you can have up to 200 records at your disposal on the Support Element. At that moment the records are said to be ‘staged’. In order to be able to actually use them, you have to install them. You can have a maximum of 8 installed records or tokens. This can e.g. be a combination of CBU tokens for CPs, CBU tokens for zIIPs and one token for On/Off CoD. This way of working was already introduced with the z10 EC, but the number of tokens is extended from 4 to 8.

The major change has to do with the way you are charged for your ‘capacity on demand’. There is no change for CBU and CPE : there’s a hardware charge but no additional software charge. For On/Off CoD you still have hardware as well assoftware charges. So what’s the big difference now ? Well, until now when OOCoD was activated, the charges were calculated according to the number of days it was used. In the new system, this is called ‘Post Paid OOCoD’. Completely new is that you can now also order ‘Pre Paid records’. You buy OOCoD records for a certain capacity in advance e.g. at the same time you order your new z10 BC. When you activate these records afterwards, there’s no payment to be done any more. It’s quite obvious (as opposed to the post-paid records) that these records have no expiration date.

There’s one other new feature I only come across in the z10 BC announcement : you are entitled to one (1) On/Off CoD Test per server with a maximum duration of 24 hours.

Other
From now on you can order Plan Ahead Memory. It means that you order more memory than you’ll be needing at this very moment. You pay immediately for the memory that you are actually using and you partially pay for the ‘plan ahead memory’, which you’re not yet using. When this memory is activated, in one or more steps, afterwards, you pay the rest of the charges for it. This can help you plan for nondisruptive memory upgrades.

The use of PSIFB (Parallel Sysplex using InFiniBand) has also been extended. From now on InfiniBand can also be used as a replacement for ISC-3 links. For the moment it’s exclusively available on the z10 with an unrepeated distance of 10km. Unlike the ISC-3 links, it does not use an I/O slot in the I/O drawer.

The HiperDispatch feature should have little or no impact with the z10 BC.

And there’s one other little sentence I’d like to mention that might perhaps slip your attention when browsing through the announcement. Just like on the z10 BC there's a “62% price reduction on memory for z10 EC when purchased after October 21, 2008, and with a specialty engine for new workloads (…) limited to 16GB per specialty engine".

Availability dates
October 28, 2008 : z10 EC New Features and functions, STP enhancements on z10 EC
January 28, 2009 : STP enhancements on z9 EC and z9 BC
January 28, 2009 : OSA-Express3 1000BASE-T and 10 GbE SR on z10 EC
January 28, 2009 : HCA2-O LR fanout card for Long Reach 1x InfiniBand on z10 EC
First Half 2009 : HMC feature #0090 on z10 EC, z9 EC, z9 BC

No comments: