Microsoft-Backed Firm Says IBM Is Anticompetitive 174
BBCWatcher writes "Microsoft has long claimed that the mainframe is dead, slain by the company's Windows monopoly. Yet, apparently without any mirror nearby, Microsoft is now complaining through the Microsoft-funded Computer & Communications Industry Association that not only are mainframes not dead, but IBM is so anticompetitive that governments should intervene in the hyper-competitive server market. The Wall Street Journal reports that Microsoft is worried that the trend toward cloud computing is introducing competition to the Windows franchise, favoring better-positioned companies including IBM and Cisco. HP now talks about almost nothing but the IBM mainframe, with no Tukwila CPUs to sell until 2010. The global recession is encouraging more mainframe adoption as businesses slash IT costs, dominated by labor costs, and improve business execution. In 2008, IBM mainframe revenues rose 12.5% even whilst mainframe prices fell. (IBM shipped 25% more mainframe capacity than in 2007. Other server sales reports are not so good.) IBM mainframes can run multiple operating systems concurrently, including Linux and, more recently, OpenSolaris."
I Don't Quite Understand (Score:5, Interesting)
T3 contends that IBM pens in mainframe customers faced with a high cost of switching to other systems, while refusing to share blueprints necessary to offer a cheaper alternative.
I'm not a hardware guy and I'm no fan of IBM but I must be missing something here: what is it about mainframes that makes them so different from servers?
Tampa-based T3 develops mainframe technology compatible with IBM software that is designed for small and midsize enterprises.
Maybe they can't release details but I'm guessing that there's some proprietary chipsets and microcontrollers inside these things to run the (what are they at 32 or 64 processors) CPUs stacked on top of each other and banks of memory and storage and database crap. So what you've gotten software written specifically to take advantage of this stuff? And it's going to be hard to move to another mainframe or standardized servers with that stuff? Are you surprised? It'd be like if I wrote something for Windows and then complained I couldn't get the blueprint from Windows of how the API works so I could move to a "cheaper solution" like Linux.
So if T3 wins this case, what's the ideal outcome? IBM open sources the software that runs on these mainframes? IBM releases detailed chipset information? Both are laughable. And if you're going to argue that, you might as well argue that Microsoft open up Windows or Intel layout the insides of its Atom processors for the world to see.
I wish I didn't find myself defending IBM (I hate their software and these mainframes sound like a scam) but you have to draw the line somewhere or apply to everyone. My advice to the poor companies still at the hands of IBM: get out. Of course that's my advice to anyone foolish enough to buy into vendor "lock-in" software like Flash. Lesson learned: An extra layer of well defined and thought out abstraction will add a bit of overhead but in the end it might save your ass when you need to switch technologies.
Buh buh but.... (Score:3, Interesting)
The mainframe is a dead relic of times past surely?
I love the cyclical nature of all this stuff.
Re:a case of sour grapes? (Score:5, Interesting)
Microsoft is now complaining ......... IBM is so anticompetitive that governments should intervene.
Talk about the pot calling the kettle black!
Well, I think it's more a case of "we got hosed in Europe so let's see if we can turn this same sword on our competitors." From the article:
The CCIA now has added encouragement from a tiny firm backed by IBM rival Microsoft Corp., which has lodged an antitrust complaint in Europe, while pressing a related lawsuit in federal court in New York and sounding out U.S. regulators.
Microsoft's been picked over with a fine toothed comb by the EU recently and I think their strategy now is to make sure everyone else is too. If you look at it that way, Microsoft has nothing to lose. They've been scrutinized to the fullest extent and you should expect them to turn this same scrutiny over to other companies in other fields. I wouldn't be surprised to see a sort of anti-competitive gaming lawsuit aimed at Nintendo come about one of these days in the EU.
Re:I Don't Quite Understand (Score:5, Interesting)
Once upon a time, a company started selling an emulator for IBM's OS/360-derived line. IBM used various legal tactics to make them stop. This line (it keeps being renamed. I think it's z/OS these days, but I could be wrong) has been backwards compatible since 1960. Any of IBM's customers who bought binary-only software for this platform at any point in the last 50 years is locked in to buying IBM mainframes.
Any customer who insisted on receiving the source code and porting rights to the code is able to move to a new platform. It therefore sounds like Microsoft is arguing against proprietary, binary-only, software. If this goes to court, I imagine the Nazgul will point out that IBM recommends that their customers invest in an open-source software stack, which frees them from lock-in. If people choose to be locked in when their supplier is recommending solutions which do not involve lock in then that's their problem. If they win arguing this strategy then it could backfire on MS quite badly.
Silliness (Score:2, Interesting)
IBM in no way forces a customer to use their systems. At any time, a customer could leave and move to another setup.
It sounds like the issue is competitors want IBM to release more details on how things are engineered, so they can design solutions for people who want to transition from IBM to other products.
IBM stuck with their investment and are positioned to make some great cash of this. The other companies need to make their own solutions that are good enough to win over customers. Lawyers have way to much time on their hands imo.
Sigh... (Score:3, Interesting)
A system Z mainframe is always run in virtualization. That's been one of their big features.
In terms of 'cloud', the term is so ill-defined and buzzed it's hard to say much, but generally speaking, a 'cloud' on a mainframe is not any less possible than a 'cloud' on disparate x86 rackmount servers.
Re:I Don't Quite Understand (Score:3, Interesting)
what is it about mainframes that makes them so different from servers
Most servers are PCs in pizza boxes. This is from Wikipedia:
Released on February 26, 2008, the System z10 Enterprise Class is available in five hardware models: E12, E26, E40, E56, and E64...The number of "characterizable" (or configurable) processing units (PUs) is indicated in the hardware model designation (e.g., the E26 has 26 characterizable PUs). Depending on the capacity model a PU can be characterized as either a Central Processor (CP), Integrated Facility for Linux (IFL) processor, z Application Assist Processor (zAAP), z10 Integrated Information Processor (zIIP), or Internal Coupling Facility (ICF) processor. (The specialty processors are all identical and IBM locks out certain functions based on what the processor is characterized as.) It is also possible to configure additional System Assist Processors...The Enterprise Class PU cores (four per chip) operate at speeds of 4.4 GHz, still (December, 2008) the highest clock speed of any processor with more than two cores per chip. The processors are stored in one to four compartments referred to as "books". Each book is comprised of a multi-chip module (MCM) of processing units (PUs) and memory cards (including multi-level cache memory).
Not quite the same as an x86, a disk, and some memory.
Re:Buh buh but.... (Score:5, Interesting)
A good point. The z (mainframe) is wonderful for something that requires its reliability. Some actual examples that I personally have had (I work on the mainframe). We had a CPU fail in our z. The __hardware__ transparently moved the running work onto a new CPU that was in "standby" mode. No outage of any sort. We didn't even know of the failure until IBM support called us that they had received a "call home" alert. We have two OSA (they are like NICs for the z). One OSA failed. The other OSA did an "arp takeover" and all TCPIP sessions continued with __NO__ outage. Again, we got messages about this, but the hardware/software recovered with no action on our part. The CE (repair man) came in, took out the bad OSA and put in a new OSA. We then issued commands and the new OSA simply started working. No down time. No interruptions to any work in progress.
If you don't mind downtime, then by all means, use Intel or AMD with Windows or Linux. If will be cheaper. And maybe even more cost effective. But, based on what happens with MS Exchange goes down around here, don't do it for anything that will make people scream if the service is done for very long. But perhaps we aren't using MS Exchange properly - I wouldn't know.
--
John
Re:Microsoft...the model of competitiveness (Score:3, Interesting)
My best guess is that their lawyers could use a lack of monopoly proceedings vs. IBM as a pretext for getting out of their own problems in that regard.
"But IBM's mommy lets him go out and play after dark" type thing.
Re:Microsoft...the model of competitiveness (Score:5, Interesting)
Let's ask one of the biggest computer buyers in the world if they are being forced to use IBM mainframes, or, if maybe they have satisfactory and competitive alternatives.
http://news.cnet.com/8301-1001_3-10209580-92.html [cnet.com]
I suspect that one of their datacenters has more computing capacity than most mainframes...
Re:I Don't Quite Understand (Score:4, Interesting)
http://searchdatacenter.techtarget.com/news/article/0,289142,sid80_gci1266438,00.html [techtarget.com]
Re:Sigh... (Score:3, Interesting)
A system Z mainframe is always run in virtualization. That's been one of their big features.
And running a VM on such a system is cheap. You can run thousands of Linux instances on a single mainframe. They won't be fast running computationally-expensive things, but they will work and be load-balanced. Try running a thousand Linux instances on a small cluster of PCs and have random load spikes in them and watch your network load jump to 100% and stay there while it tries to re-balance the load across the cluster (if your hypervisor supports this, of course, otherwise wait until all the VMs on one node spike at the same time and watch 90% of your hardware sit idle while a load of VMs are CPU-starved).
Migrating a VM from one cluster node to another using something like Xen or VMWare takes several seconds, more if the network is slow or saturated. GigE does not count as fast in this situation; you're transferring at least 512MB of data (the VM state) which takes 4 seconds at the theoretical maximum throughput for an uncongested GigE link, more like 10-15 seconds in the real world. On a mainframe, you're just rearranging around CPUs, which has a response time measured in milliseconds (or nanoseconds, depending on whether the CPUs are on different boards). This is makes a massive difference to the load balancing, and good load balancing makes a huge difference to your utilisation. If it takes 10 seconds to move a VM then it's very difficult to respond to unexpected demand and so you need to over-provision by a lot.
Of course, you can just buy a really big SMP machine to host your VMs. A 64-Core i7 would be similar to a mainframe, except that you'd also need a load of PCIe slots filled with network and disk controllers to get close to the I/O bandwidth of a cheap mainframe. You'd also need some custom hardware or software to handle CPU failures (I can recommend a company that does this for Xen, but they're not cheap), and when you're adding that many commodity components to a single box you're going to have a lot more frequent failures than for smaller systems. By the time you've done this, you've probably paid more than you would for a mainframe, effectively built a mainframe, but paid more and not got a decent support contract. IBM could probably build a mainframe with Intel chips without much effort (although, if you look at the error recovery stuff in the z10, you'd be crazy to prefer an i7), but there's not much point.
One thing that surprises me is that there aren't hosting companies selling Z/Architecture VMs.