and the cost of the software licenses?
About as much as one of the skyscrapers in the picure probably....
Paris because even she'd blink at thost costs.
At least twenty years after pundits first pronounced the death of the mainframe, IBM has released a new one. Of course, the proof of the pudding will be in the market, but IBM will be hoping that the billion dollars it's poured into developing the new z13 mainframe will get the big end of town as excited as Big Blue itself is …
I dont get it how IBM can claim that their latest Mainframe can emulate 8.000 x86 servers? Considering that a Mainframe cpu is much slower than a decent x86 server, something does not add up.
This z13 Mainframe gives 110.000 MIPS in the largest configuration z13-NE1, costing millions of USD.
http://www.itjungle.com/tfh/tfh011915-story04.html
This is only 40% faster than the even slower z12. Since z13-NE1 only has 21 sockets, and each socket is much slower than a decent x86 cpu, how can it replace 8.000 x86 servers? Well, it turns out that IBM assumes all x86 servers idle at 1-2% load, and the Mainframe is loaded to 100%. Now, imagine 21 of the x86 servers start to do some work, how can 21 slower z13 cpus keep up with the work load? It is impossible. IBM marketing division is over ambitious again.
Also, in the link above, the IBM die hard Timothy Pricket Morgan explains that an old IBM P795 Unix server gives 1.6 million CPW. Whereas this z13-NE1 gives 735.000 CPW. This means that the old Power7 server is much faster than this z13 cpu.If you normalize, 32 sockets POWER7 cpus, vs 21 sockets z13 cpus, you see that the old POWER7 cpu is 43% faster than this brand new z13 cpu.
Ergo, the old POWER7 cpu is about 50% faster than this spanking new z13 mainframe cpu. And we also know that the latest x86 cpus are much faster than the POWER7, and even faster than the latest POWER8 cpu. So, tell me how 21 slow z13 cpus replace 8.000 x86 servers?
BTW, this z13 mainframe has 10TB RAM. That is chicken sh-t, considering that the latest x86 servers with 8-sockets has 12 TB RAM. And Oracle SPARC M6 server has 32 TB RAM. And this year Oracle will release their M7 server with 64 TB RAM.
IBM may have done it again, too early to tell if this is as big an architectural breakthrough as the System 360 was in its day, there is very much a market need for safety first with security built-in and persistent, protected and audited real-time transaction analytics available end-to-end.
Mainframe concept is still very much with us, from large flagship first-rate systems to writ small on a chip and wafer, due to the need for absolute security and transaction data integrity for the new internet and of things.
Perhaps true now more than ever in today's tough marketplace and volatile world? Safety first.
Fujitsu, with their partner Oracle, is a major mainframe maker and this development is likely to raise the protected performance bar for both.
Interesting and fascinating to watch how this market reacts and develops...
There is a market need for "safety without investing into software failover and clustering".
These require qualified software developers, testers and in most cases fairly complex test setups to verify failure paths. That costs money and quite a bit of it too.
This is what mainframe addresses and it does it pretty well too.
Probably, if you think you have the space.
This looks like the same racks that the 9125-F2C P7 775 system is packaged in (they're both products from IBM Poughkeepsie, NY), an if so, this is 2 racks side-by-side, with each rack over 2.10 metres tall and 1.8 metres deep. Both racks together would be around 2 metres wide.
In addition, they will not take standard 19" wide rackmount devices without some additional mounting hardware as the 'gap' is 26" IIRC (sorry, I realise I've mixed measurement units).
IBM actually have some quite fancy doors available for their standard T-series racks, if you want to pay for them!
Pretty much all racks have castors now, including supercomputers and mainframes.
I can check, but I think that all of the IBM P7 775 and z196 and the Cray XC40 frames that I can see in the machine room here have castors.
They also have wind-down feet and load-spreader bars when they are in their final position, so that they don't move.
This post has been deleted by its author
Even back in the 360 days we had a Johnson bar to lift the boxes on one side to retract the feet so it could be rolled about as needed.
And transistors weren't just a feature of the 360. The earlier 7080/7090 used transistors in their memory (and elsewhere). Of course, they were as big as your pinky fingernail, but they were still transistors.
... I planned the installation of a full height 9076 SP/2 into a normal office space in an IBM building in Basingstoke.
When installed, it was about half full, and did not quite exceed the floor loading weight.
After I left, I heard it had been filled. I had visions of it descending through the 11th floor, then the 10th, the 9th and on to the ground!
@Peter "After I left, I heard it had been filled. I had visions of it descending through the 11th floor, then the 10th, the 9th and on to the ground!"
Indeed it was filled up and we later updated the switch on it. It sat there for quite a while being used by both HW and Call-AIX members.
We also had a S70 with separate disk enclosure next to it. So we must have been way over the floor limit at that time.
Many (many) years ago we had one of the first VAX 11/780's installed in an office with no goods lift. Those systems weighed around 3/4 tonne (and were less powerful than my current mobile phone!). Floor loading was OK, but when it was delivered it was discovered that the only way to get it in the lift was to uncrate it in the lobby, slide it in (1cm clearance), reach round to push the 2nd floor button, and then run up the stairs.
It was about 50% over the weight limit for the lift, but people assumed that the lift probably had a factor of 2 for safety, and there's no-one in it, so it'll be OK. It was, almost. The lift stopped about 5cm below floor level. Ever tried to get a 3/4 tonne box up a 5cm step when you only have 1cm room at the sides to insert fingers, levers, etc.?
They did it, eventually. The DEC service guys said that it went better than a previous delivery, that one had to go in via a window, and just as it reached the 2nd floor it slipped out of the crane's webbing...
A similar lift story from after I left IBM, but not as interesting.
We had a Power 4 system delivered in a T42 rack to a site I was working at in Poole, and to keep it under the weight limit for the lift and to get it through the doors (it was too high for the lift doors) we stripped the drawers out of the frame in the loading bay, tipped the frame on it's side, and then re-installed the drawers in the frame once it was on the machine room floor. All without telling the IBM hardware engineers!
The only problem we had was that the SPCN (Sequenced Power Control Network) cables were put back in the wrong locations, which gave us problems with the I/O drawer identification for the remaining life of the systems, even after they were connected correctly.
It should be noted that IBM pretty much invented virtualization with the 370 mainframe systems in the early 1970's. About the same time, Intel were making 4 bit microprocessors and TTL chips.
The virtualization will be performed either by the PR/SM type 1 (hardware) hypervisor or z/VM.
Read up on Type 1 hypervisors. There does not have to be a host OS, at least not as I think you understand them.
"Read up on Type 1 hypervisors. There does not have to be a host OS, at least not as I think you understand them."
Well presumably something has to control the physical hardware or nothing would be able to be loaded from disk. I'd call that the OS or at least the kernel.
Yes, that's quite true, but if you look at PR/SM, the IBM Power Hypervisor, or Amdahl's MDF (the bare-metal hypervisors I've had experience with), they are deliberately very limited in function. The name Hypervisor (derived from an old alternative name for an operating system, the Executive Supervisor) was coined to indicate that it was a supervising program that was not an operating system. It was very deliberate to not call the hosting environment an Operating System.
It's only relatively recently that you've had Type 2 or 'hosted' hypervisors that sit on top of what one would describe as a normal operating system like Linux or Windows. Examples include the original incarnation of VMware, Xen, KVM and Parallels. I understand that HP's Integrety VM sits on top of HP/UX, although I have no experience.
And then you have things like VMware ESXi, which is classed as a type 1 bare metal hypervisor, but is really a canned Linux stripped of all functions that are not required to host other systems. Mind you, you could probably say the same about IBM's Power Hypervisor, but that is so deeply embedded in the firmware of Power systems that it's relatively difficult to see that it is Linux at heart.
Complicating it still further are Oracle/Sun's containers and IBM WPARs, which are not true VMs but still allow you many of the advantages of partitioning.
It's all getting complicated.