What the zEnterprise 196 can also include the zEnterprise BladeCenter Extension, or zBX, which is a standard server rack that includes a top-of-rack 10 Gigabit Ethernet switch, a BladeCenter chassis, and power distribution units for the blades. The zBX rack can house up to four chassis and links to the mainframe by two private 10 Gigabit Ethernet networks – one for management and the other for data being passed between the Power and x64 systems and the mainframe. The BladeCenter chassis has a separate switch for letting the outside world talk to the Power and x64 blades as application or infrastructure servers.
I had been speculating that IBM could use an InfiniBand link to lash the blades to the mainframe, but apparently this capability was not ready. There is some mumbling that tighter coupling between the machines is coming down the pike at some future time, but IBM did not confirm this at the announcement. Each zEnterprise 196 can have eight zBX chassis, for a total of 112 blades.
IBM says the whole shebang – with z engines and blades – can support up to 100,000 virtual machines. Something it claims no other machine can touch. (I guess it really depends on how you define machine).
At the moment, IBM is supporting only its single-socket, eight-core Power Systems 701 blade servers in the zBX chassis. (El Reg told you all about the Power7-based blade servers, which were announced in April, here.) IBM is only supporting AIX on these blades, not its proprietary i (formerly OS/400) operating system or Linux. Customers will be able to support Linux on x64 blades, but the x64 blades for the zBX are not expected to be available until sometime in the first half of 2011. The word on the street is that IBM will be using Red Hat's KVM hypervisor on these blades, and it is likely that the company is working to integrate KVM with the Hardware Management Console that IBM uses to control its z/VM on mainframe and PowerVM on Power hypervisors.
The one thing that the zEnterprise 196 will not have is a lower price tag compared to the System z10 it replaces. According to Karl Freund, vice president of System z strategy and marketing at IBM, the pricing per mainframe engine for the zEnterprise 196 is exactly the same as it was on the System z10. (IBM does not reveal its mainframe pricing, so it is hard to say what that number is, but it is hundreds of thousands of dollars per core). Each zEnterprise 196 engine has about 30 per cent more MIPS for that same money, which works out to a 23.4 per cent reduction in the cost of mainframe capacity.
IBM is not being totally stingy with mainframe shops. Freund did say that IBM had cut memory prices, maintenance, and Linux engine, and z/VM license costs by 35 per cent on the zEnterprise 196 machines compared to the z10s.
With the Power7 midrange machines announced in February, IBM held its system level prices on the Power 750 and 770 boxes about the same as their Power6+ predecessors, but the boxes had twice the aggregate oomph. This is a much bigger jump in bang for the buck.
The zEnterprise 196 machines are expected to start shipping on September 10. The z/OS V1.9 operating system, which goes off support on September 30 along with the end of sales for the System z10 mainframes, will run on the zEnterprise 196 server, but its takes z/OS V1.10 to exploit some of the new processor features and z/OS V1.12 to fully exploit it. If you want to use the zBX blades, you need to be at z/OS V1.10 and z/VM V6.1 at least.
z/VM 5.4 will run on the new mainframe, as will IBM's z/VSE V4.1 or higher and z/TPF V1.1 or higher operating systems. Novell's SUSE Linux Enterprise Server 10 or 11 will run natively on the z196 engines, as will Red Hat's Enterprise Linux 5. (Presumably RHEL 6 support is coming concurrent with that operating systems release within the next few months). On the Power blades in the zBX chassis, you need to run AIX 5.3 or 6.1, and presumably when AIX 7 ships around October, these will also be supported. The PS701 blades themselves will not ship in the zBX unit until November 19. ®
That would imply they can perform some sort of translation inside the CPU. Transmeta did this, but they could apparently not compete with "real" x86 processors. Also, their technology included some very sophisticated software performing the translation and I have never heard of that with Power or S/390 CPUs which are part of production systems.
My guess is that IBM reuses ALUs, caches, branch predicition logic etc from Power CPUs in S/390 CPUs and the other way around. Microcode can only be used for infrequently used instructions, as that makes them *really* slow.
What I suggest is very similar to Apple's CPU transition strategies. The successfully supported 68K code on Power and the Power code on x86 with acceptable performance penalties.
I heard IBM Mainframe cpus are a derivate of POWER6, is it true? Then it explains the high Hz of 5.2GHz and it explains why the Mainframe cpus are so dog slow. You need four POWER6 to match two 2.93GHz Intel Nehalem (not Nehalem-EX, just ordinary Nehalem). A Mainframe CPU is 5-10x slower than a modern x86 such as Nehalem-EX or AMD Bulldozer. If you have 16 Nehalem-EX then it gives more CPU performance than this newest IBM Mainframe with 96 cores. That is a bit funny. No, cpu performance has never been the IBM Mainframe strong side.
BTW, there are OpenVMS machines with 17 years of reported uptime. OpenVMS clusters are legendary for their availability and runs for decades, where you upgrade one machine at a time without any downtime.
I was under the impression that the z/Arch CPUs were modified POWER CPUs anyway, just with different decode front ends and/or microcode.
So, if they're already using a POWER core, but instead running z/Arch instructions in hardware or microcode... that's better than having to translate to POWER instructions, and then to the micro-ops of the real CPU, no?