Follow Slashdot blog updates by subscribing to our blog RSS feed


Forgot your password?
Operating Systems Software IT

VMware, Cisco Plan Data Center OS 83

Lucas123 writes "John Webster over at Computerworld says VMware and Cisco plan to develop a Data Center OS that would consist of a data center cloud populated by servers, storage, and Cisco's 'intelligent' networking gear, all managed by Cisco and its partners — starting with VMware."
This discussion has been archived. No new comments can be posted.

VMware, Cisco Plan Data Center OS

Comments Filter:
  • Imagine... (Score:1, Funny)

    by Jrabbit05 ( 943335 )
    a bewulf cluster of these!
  • this differs at all from OpenView?

    • When I first read this article, my immediate thought was that if they needed a mission critical kernel and/or mission critical hardware infrastructure to power the thing, then they could probably purchase OpenVMS and/or Nonstop Himalaya for pennies on the dollar.

      Or go ahead and purchase QNX from Harman-Kardon.

  • Linux. Just like VMWare ESX.
    • Re: (Score:3, Insightful)

      by Hatta ( 162192 )
      IIRC, ESX only ships with linux in the service console []. The actual vmkernel is not [] based on linux.
      • You have things a bit mixed up. ESX versions prior to 3 use a Linux kernel that runs all the time, with what VMware calls "vmkernel" on top of it, which is suspicious to say the least, and ESX3 loads "vmkernel" from the Linux initrd directly.

        Its more complicated than that for sure, and every description I've read of how ESX works makes me more suspicious of what VMware does. In particular their use of Linux drivers, which are and always have been GPL licensed, so either they are using the Linux kernel to ru
      • by Natales ( 182136 )
        That is correct. In fact, the recently announced ESX 3i has *no* service console at all. It is a pure vmkernel in 32MB of flash.
  • Confused. (Score:4, Interesting)

    by MrCrassic ( 994046 ) < minus language> on Monday October 22, 2007 @02:59PM (#21075453) Journal

    Is this implementation going to set up virtual servers aligned as a data center, for which virtual computers can access? Or is this an idea for a completely custom virtualization-based operating system that offloads one huge datacenter onto single computers?

    If either is the case, how is that any different than either setting up a test server (or servers) with VMware computers all connected to each other using physical connections, or just having multiple VMware sessions on one computer all interconnected using a single connection?

    • If either is the case, how is that any different than either setting up a test server (or servers) with VMware computers all connected to each other using physical connections, or just having multiple VMware sessions on one computer all interconnected using a single connection?

      Because this way you can get "a leading solution from the leaders in leadership" instead of paying in-house salaries.

      • by mindstrm ( 20013 )
        Everyone likes to knock the business world for provding proprietary solutions.

        Guess what.. hiring more staff costs more, and more importantly, takes far more time than purchasing VMWare, the commercial Xen thing, or any other commercial virtualization and management platforms.

        And you know what? it's not about the core virt tech... that's free. It's about the management tools.

        You are probably thinking, sure, why pay them to run a few real boxes with a few dozen virtual boxes. and right you are.
        But what about
    • Re:Confused. (Score:5, Interesting)

      by jhfry ( 829244 ) on Monday October 22, 2007 @03:12PM (#21075629)
      I think what they mean by "Data Center OS" is that they are developing an OS that utilizes an entire datacenter as it's resource... rather than a series of discrete systems that are linked by their OS's.

      Essentially I imagine it will be similar to the way a mainframe acts... a large number of resources, all controlled by a single OS. But instead of a single manufacturer implementation, this will be a proprietary "standard" that allows 3rd party components to be added to the cloud and become available to the "OS"

      It's actually a great idea... it's the next evolution from virtualization... Why should a data center admin need to ever concern themselves with the individual servers and storage devices... instead they just add another processor (server), storage device (NAS array), or external network, and the OS will utilize it as it chooses.

      I'd imagine that if it is properly executed, it will greatly improve utilization and make managing the data center infinitely easier as all you do is plug the device in and it is automatically assimilated into the system.

      All I hope is that they make it a true industry standard, it would suck if only a combination of Cisco, Dell, and EMC devices (for example) would work with the new "OS".
      • Re:Confused. (Score:4, Interesting)

        by jrexilius ( 520067 ) on Monday October 22, 2007 @03:32PM (#21075897) Homepage
        This is in essence what my company ( does. We take it a step further and "cluster" multiple datacenters into a single system. Although what we do is not based on virtualization more on underlying OS and altering programming language internals to make it transparent to the application.

        In essence they get to the same place but ours is more limited and geared towards a particular class of application running on the LAMP stack. Also the problems we were trying to solve when building it was not utilization but scaling, availability and performance. We are actually kind of wasteful of hardware currently but as we grow we get more and more efficient. Economies of scale thing.

      • it will greatly improve utilization and make managing the data center infinitely easier as all you do is plug the device in and it is automatically assimilated into the system.
        We are the BOS*. You will be automatically assimilated into the system. Resistance is futile. Welcome your new Cloud Computing overlords.

        *BOS - Borg OS.
      • Re: (Score:3, Insightful)

        by nine-times ( 778537 )

        this will be a proprietary "standard" that allows 3rd party components to be added to the cloud and become available to the "OS"

        It's actually a great idea...

        The part that isn't so great about it is the "proprietary" part. I guess it depends on the implementation and how open the "standard" is. But having worked in IT for several years now, I've become increasingly convinced that closed standards and proprietary systems just aren't acceptable. What inevitably happens is that this terrific idea and great

        • Remember that actual standards are in many cases based on proprietary standards, and those that aren't can take an very long time to complete and become adopted, because they sometimes lack financial backing and motivation. Bleeding-edge technology will often use proprietary standards at first, with an actual standard emerging with time and popularity. It's also cheaper and faster to introduce features into a proprietary standard instead of pushing widget X through a committee (where your competitors all ge
      • Looks like Plan 9 may make it to the real world after all... this kind of application is essentially what it was built for, right?
      • So it's clustering, then.
        • by jhfry ( 829244 )
          A cluster is still a set of discrete OS's working in concert... I believe what they are intending to do is use a hardware layer to allow the "OS" to utilize the hardware independent of an operating system on the individual components.

          The closest parallel I can draw is that of a Fiber Channel SAN... any machine that is physically/logically connected to that SAN can access the storage as though it is local to that machine.

          Imagine a protocol, implemented purely in hardware, that allowed distributed processing,
      • by vidarh ( 309115 )
        There's several different technologies already doing similar things to this.

        The oldest being "single system image" clusters. DEC/Compaq/HP had an SSI project for Linux a while back, see [].

        There's also a number of platforms such as 3Tera's AppLogic [] that abstracts away the invididual servers using virtualization. AppLogic looks fantastic in many respects - you get a nice GUI to connect together "components" to applications that can span hundreds of CPU's, and you can then deploy applications on

      • All jobs at work station WKS008B ended.
        All jobs at work station WKS010B ended.
        All jobs at work station WKS041A ended.
        All jobs at work station GIBSON ended.
        The call to *LIBL/QCMDEXC ended in error (C G D F).
        Reply . . : C
        All jobs at work station CONSOLE ended.
        All jobs at work station WKS009A ended.
        All jobs at work station WKS010A ended.
        All jobs at work station WKS014A ended.
        Device DSP03 no longer communicating.
        Device DSP03 no longer communicating.

      • So what's the difference between this and an IBM mainframe?
    • by t0rkm3 ( 666910 )
      I imagine one of the key components is the MDS9000/MDS9500 or a switching device like that. One of the features that was supposed to become part of the switch (I was a beta tester for a very large storage company) was disk virtualization. FC connected disk could be presented via FC-IP, ISCSI, or CIFS/NFS from the switch. If the volume was CIFS/NFS this could be presented from a small scale special purpose server built into a slot in the chassis.

      Take the above piece of hardware and add a FWSM (Firewall Servi
    • This is IBMs VM OS from the 1970s. We are just now catching up to just over 30 years ago. Good work team. You almost missed it. Any later and it would have been from 40 years ago.

      I'm both impressed and saddened at the reinvention of this wheel.

  • UTILITY Computing, IBM's wet dream. Also, game hosting in the cloud. No more installing clients, you play in virtual clients. Game houses wet dream. Thin clients, Oracles wet dream.
    • > "Game houses wet dream. Thin clients, Oracles wet dream."

      ... guess its time you all started stocking up on Depends, because next you'll be sh*tting yourselves over it ...

    • What the fuck is that, seriously?

      Are we to stream video from a server somewhere that has a beefy video card?

      Because if so, that's the dumbest idea I've heard in awhile, though it may make cheating harder.

      If not, I don't see what you mean by a "virtual client" or what it has to do with this concept.
  • License costs? (Score:4, Interesting)

    by Colin Smith ( 2679 ) on Monday October 22, 2007 @03:11PM (#21075611)
    I keep looking, going "holy fuck", then shelving the idea for another year.

    I know the architecture I want. Just can't justify it... Xen might.

    • except, now that Citrix has purchased Xen they raised the cost by 300%
    • They have a cloud like this, running on Xen.

      But it's more like a collection of virtual machines, and a collection of servers. You buy the virtual machines, they figure out where to put em -- but nowhere near as flexible as a real cluster would be.
  • ... to place the information that the upcoming world leading DCOS is very green [].

  • There are better virtual I/O solutions out there, several new startups.

    check out []
    • by dosguru ( 218210 )
      Our cisco guy won't even sell it to us, he said he wants to keep us as a customer. Without naming names, we're a Fortune 100 company and are the target market for this.
  • If I have a big pile of windows machines I can employ cheap admins to point and click all day long to keep it up. With a solution like this I have to have specially trained staff, i.e. expensive. There would have to be an enormous saving on staffing levels to make it worthwhile. Yet, then I'm in the situation of say one staff member leaving causing a major headache when it comes to finding a replacement.
    • by umghhh ( 965931 )
      Well done. And if you want something from your staff (other than what is done on 'normal' day) then then you learn Cantonese?
  • by machinecraig ( 657304 ) on Monday October 22, 2007 @03:24PM (#21075781)
    TFA was pretty short on details - but coupled with this release from the folks at BEA [], which basically allows a Weblogic app server to run directly under VMware (no other OS required); it may give a clue as to future direction. I'll take it all with a grain of salt.
    • by larstr ( 695179 )
      The statement "directly under VMware (no other OS required)" is utterly wrong and their marketing leads you to get this wrong. BEA has written their own OS that supports the limited set of hw devices that exists under VMware. They've named this OS Bare Metal, and they now market this with that you now can run your Java apps on bare metal under vmware esx. Sure.
    • Re: (Score:1, Informative)

      by Anonymous Coward
      Ordinarily BEA's stack would look like this:

      HW => OS => BEA => JVM => your code

      With virtualization, it looks like this:

      HW => ESX => OS => BEA => JVM => your code

      They essentially wrote their own OS, so now it looks like this:

      HW => ESX => BEA OS => JVM => your code

      The ironic thing is that BEA called their OS "Bare Metal", but it is only capably of running as a guest in a VM (i.e. it can't actually run on bare metal). Since it only runs in a VM, it doesn't have to concern
      • by larstr ( 695179 )
        Well, if you have the right hw, you can get it running on a physical box too. I've seen it run. :-)
    • by Natales ( 182136 )
      This is pretty real stuff and this is what the trend seems to be these days. BEA's LiquidVM is the most clear example of this. VMware calls this concept JEOS (Just-Enough OS). The idea is to leverage the hypervisor's capabilities for all the underlying operations related to access to the real hardware. That allows you to develop your OWN OS based on the needs of the application. You no longer need to support thousands of devices on the lower level or thousands of software facilities and APIs in the higher l
  • by Anonymous Coward
    Take ZFS for filesystem clustering. Add Beowulf for processor clustering. Tip hat to 20-year-old VAXcluster technology, and 40-year-old IBM utility computing.

    What am I missing, please? Apart from buzzwords like "cluster computing" and "intelligent gear".
  • Why not... (Score:3, Interesting)

    by symbolset ( 646467 ) on Monday October 22, 2007 @03:28PM (#21075849) Journal

    A vm on every desktop for serving stuff, with some management glue to make it look like the vm is running on a server in a rack?

    Is it not time for that yet?

    • A vm on every desktop for serving stuff, with some management glue to make it look like the vm is running on a server in a rack?

      Is it not time for that yet?

      I have liked this idea for years. Its like the Seti@home project for under used desktops. Very few professions use a significant part of the CPU of the computer sitting on a desk, and those that do, probably do so only occasionally.

      However, while simple in concept I suspect it would still be hard to implement.
      • Is code for high margin opportunity.
      • by mindstrm ( 20013 )
        It's an interesting concept.. but...

        No, i'ts not time. We don't care, and the resource utilization is too hard to manage and marginal.

        Our main problem isn't computing resources, it's configuration management. Servers are cheap.. but we have a lot of servers sitting around 98% idle because we don't want to mix applications on the same server. We get held up launching a new application because we don't really want to keep buying new dedicated hardware for each project.

        VM tech lets us cleanly separate resour
      • An under-used server is a waste, I'll agree, provided you have more than one of them. That's because one server at 50% CPU is likely consuming less power than two servers at 20% CPU each. Even the CPU is probably more efficient, but there's the rest of the box to consider.

        But, desktops? Hell no. Aside from reliability issues -- figuring out which ones are on and available, and where to route a request -- there's security -- just who's computer gets to process your credit card information? What if it's not m
    • Not until ECC RAM becomes common in desktops.
    • by y86 ( 111726 )
      If only we had some sort of windowing system that could use this.... oh WAIT, xorg was built for this. :-)

      Have SSH port forwarding, will remote in.
  • Xen (Score:3, Interesting)

    by athloi ( 1075845 ) on Monday October 22, 2007 @03:41PM (#21076007) Homepage Journal
    A new battle's shaping up. Citrix, known for remote management software, has acquired XenSource. Symantec has a management utility. So does Microsoft/Novell. Should be a good fight.

    I'm looking forward to a return to big iron or something like it. The quality of hardware, and the amount of thought that went into the operating system, software and configuration, was much higher. Big Iron is like the aristocracy of computing.

    An interesting article from last year on this topic,1759,2004075,00.asp []
  • forget all that silicon... such a waste
  • by SoupIsGood Food ( 1179 ) on Monday October 22, 2007 @04:21PM (#21076631)
    Since this is Cisco we're talking about, I'm looking forward to long nights memorizing oddball commands to pass the certification test. I can almost feel the one-coffee-too-many burning a slow ulcer into my esophagus while puzzling over the two-and-a-half bibles thick study guide.

    Cisco makes mad, crazy money from certification tests. It's a way they can squeeze another dime from both out-of-work and desperate tech workers as well as companies confused as to exactly what their CIO bought when he went to play golf and came home with the Cisco polo shirt (and, oh, yeah, some contract or something. My name is going to be in Business week, and I got a shirt!)

    Money all around, and all they need to do is pretend the advances in modern GUIs, scripting tool, shells and command line utilities the rest of the industry uses don't exist.

    Now they want to take this esoteric-for-esoteric's-sake aesthetic across the entire enterprise! On the one hand, having that certification will mean a huge pay jump, as no-one will be able to design, deploy or maintain the sumbitch... I won't either, but I'll be making lots of money calling in Cisco consultants to do my job for me. I might get them to bring me a polo shirt. On the other, you will never be able to bring into the server room a new technology that Cisco/VMWare doesn't want in the server room, regardless of whether or not it's the right thing for your organization. It's like Bad Old IBM all over again if this thing gets any traction.

    • I've always thought that Cisco was the most over-hyped item in an industry full of hype. Their quality control is a long way from what it should be if they want to charge a premium for what is really a commodity product. I've had dead ports, dead switches, flaky fiber ports, POE switches that drop power to equipment at random intervals, POE switches that don't feed the required voltage, routers that regularly corrupt their routing tables, and more. Worse yet are the the CCNE's that I have to work with lo
      • Took me two months and divine intervention to drag one kicking and screaming down into the basement of the same fucking building where she worked so that she could see that there really **WAS** a problem and it really **WAS** with her equipment.

        You probably should have just killed her, removed her head and placed it on a pike in front of Cisco's main office. Oh, I'm sure it would have little effect on Cisco's support policies ... but just think how satisfying it would have been.
  • Datacenter 1 is going down due to a license violation. Please contact or rerun the genuine datacenter advantage tool.

    Cisco might still be able to get away with having proprietary networking gear, but there is no way most organizations will move to proprietary for entire data centers.

  • This sounds strangely like Plan9 []

  • by ejamie ( 765128 ) on Monday October 22, 2007 @10:34PM (#21080453) Journal
    Shortly after the first Cisco Datacenter OS is brought online, it becomes self-aware alarming its creators at its newfound abilities. When the lab rats attempt to shutdown the datacenter, "CDOS" defines all tape operators as its enemy and decides to terminate their mount tape requests. 3 billion digital bytes are destroyed within milliseconds.
  • those of you who are better informed on this subject than I, or for that matter anyone with a reasonable-sounding opinion: how does something like this compare to Google's GFS and all the tools that Google has for large-scale deploments?
  • ....will it be backwards compatible with web 1.0??
  • I can't help thinking that now that hardware is finally outpacing our needs we need yet another level of software abstraction to justify buying more and more hardware. First there was library bloat, then feature creep now every application has to be run inside its own virtual machine. I remember when "Hello World" compiled down to about 850 bytes on a DG Eclipse machine. Of course that machine had 32K RAM so it had to be efficient. Now I compile "Hello World" on my Linux box and it's well over 30K but my Li

"No, no, I don't mind being called the smartest man in the world. I just wish it wasn't this one." -- Adrian Veidt/Ozymandias, WATCHMEN