IBM shrinks z16 and LinuxONE systems into standard rack configs Big iron is not always beautiful. Just ask IBM, which has updated its z16 mainframe and lifted the covers off a shiny new LinuxONE 4, both of which are built for rack mount and single frame configurations. The systems are intended to appeal to a broader audience due to smaller size and reduced energy consumption. The servers …
> a single Rockhopper 4 would let customers replace at least 36 x86 servers, reducing energy consumption by 75 percent and space by 67 percent
So if it can replace 36 x 1U servers and use 67% of the space, does that mean this beast is a 24U rack mounted server? Does it come with a free forklift to get it into the rack?
So if it can replace 36 x 1U servers and use 67% of the space, does that mean this beast is a 24U rack mounted server? Does it come with a free forklift to get it into the rack?
Probably. Based on the spares and tools that came with a p5-595 a long time ago it's likely an optional FRU. Besides, don't you already have a server lift in your DC? :)
The p5-595 process included an IBM engineer that came ahead of time to verify the installation prep like power (redundant 3-phase 60A), floor capacity, and they path from the dock to the DC were all up to snuff. I think the 24" rack was in the neighborhood of a ton (US short, long, or metric I don't recall).
I can't say for certain whether they ship a forklift, but the IBM P7 775 HPC systems that used to use the same 26" racks as whatever was the then z system (z196?) came with a device called a 'lift tool" (which was basically a large trolley with a scissor lift incorporated in it) , and also had wind out rails to allow you pull the drawers out of the rack more gently than just tugging them. But removing the drawers from the racks was a relatively unusual operation, as much of the work could be carried out with the draws in-situe, as may of the replaceable units pulled out from either the front or rear of the drawer.
These were both powered using a meaty Makita rechargeable hand-held power drill which was actually part of the maintenance kit.
When we got the original details through, there was supposed to be a powered tow unit to move the racks, but when they were delivered, they resorted to old-fashioned muscle power to move the racks (although 3.5 metric tonnes per rack, without the water, was quite heavy, and creased the aluminium load spreading sheets that were used to prevent damage to the suspended floor).
I'm sort of interested in how they power these new systems. For full-blown z-systems, there is a thing called a "Bulk Power Unit" at the top of the rack that conditions the power and provides multiple-redundant power conversion and regulation from whatever is the local supply voltage (normally three-phase, of course) to what the system actually needs (I think it was 48 or 64 volts DC). This BPU is 'intelligent' in that it had it's own redundant service processor pair in it, and is controlled and monitored from the Hardware Management Console.
I wonder what they do for these 19" rack-mount units.
Last time I saw a comparison between X86 servers and IBM tin (albeit I think it was for Power rather than z), it specifically excluded any form of virtualisation on the X86 side for some spurious reasoning. At that point, any comparison of 32 physical servers vs 1 virtualised host was going to be a win for the virtualised host.
Also, unless they've reduced their costs significantly, Linux IFLs don't come cheap. You could buy a 4 socket (not core) server for about the same cost as a single Linux IFL last time I saw a price sheet.
So yeah, be wary of marketing fluff from IBM comparing their kit to x86.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
