Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Linux Software

Compaq sells Linux Clusters 54

Posted by CmdrTaco from the can-i-make-a-beo,-oh-nevermind dept.
sumdumgai noted that Compaq is now selling preconfigured Linux Clusters if you happen to have any weather patterns sitting around needing to be analyzed or something. But this is a pretty substantial bit of heft behind Linux as a valid Clustering platform, as well as an easier way to get one.
This discussion has been archived. No new comments can be posted.

Compaq sells Linux Clusters More

Compaq sells Linux Clusters

Comments Filter:
  • Take a look at www.linux-ha.org. Combined with linux-vs, it can do not only failover (N+N) but also hot spare (N+1).
    ___
  • The Linux virtual server project has open-source HA clustering. Good stuff.
  • I would never sell an x86 box with five nines guaranteed uptime (not that I sell these anyhow..)
  • And maybe you didn't know that Alpha was made by DEC, before Compaq bought them? Comaq had nothing to do with the quality of Alphas.
  • A few options:
    Open Source: Kimberlite [missioncriticallinux.com] from Mission Critical Linux, Inc [missioncriticallinux.com]
    With Support & FULL NFS (not fully Open Source: Convolo [missioncriticallinux.com].

    Kimberlite is the underlying technology, working today, designed from the ground up for HA on Linux (including some extremely robust data integrity capabilities). Convolo builds on it, adding full NFS support (state failover, host & netgroup failover, etc), and comes with 90 days support.

    They also sell support contracts [missioncriticallinux.com] for Linux and Convolo (platinum, "we need a kernel fix now" type contracts, not just "buy a bundle of calls" type).

  • > Unfortunately, the mirroring software is not open source or free software. :\

    Try SGI's GPL'd FailSafe [sgi.com]:

    (I have not tried it - all I know is from what's on the website.)

    SGI's Linux FailSafe is the premiere next generation High-Availability (HA) solution for business critical applications. It provides a simple way to make your applications HA without having to resort to rewriting or recompiling or the need to invest in expensive hardware solutions. Linux FailSafe provides a robust clustering environment with resilience from any single point of failure.


    Linux FailSafe is architected to scale up to 16 nodes in a cluster with the cluster members sharing storage. Linux FailSafe and shared storage allows multiple servers to assume control of data in the event of a failure. At the point of failure, applications are resumed on the remaining system{s} and filesystems are automatically made available to their applications.


    ---

  • I think these are aimed at the high availability, transparent failover, fault tolerant, load balancing type of clusters - not the supercomputer nuclear-blast weather simulating number crunching type clusters.
  • If anyone had actually taken the time to read the linked page at Compaq, they would have noticed that this is NOT a high-performance (beowulf) cluster.

    True that.

    But a 16 node ip failover turn key linux based web farm is still hotter than an over clocked P4 with a broken heat sink!

    --

  • LVS is *NOT* a high-availability offering. It is for load balancing.
  • This is a great concept and hopefully Linux has implemented it effectively (haven't used it yet). I have personally used Microsoft's clustering services and I think they stink. The HA side of it is horrid. In the time it takes for the service to gain control on another cluster machine you could have cycled the original machine that went down.

    Just my $.02
  • I don't want to analyze the weather of a nuclear blast. I have the weather patterns from Europe on June 5, 1944 and just want to find out what the weather will be like on June 6, 1944.
  • >How long would it take to get a replacement PS from Compaq?

    Depending on your maintenance contract anywhere from 2 hours to 2 days. This is one of the reasons this isn't such a bad solution if you want a Linux cluster. Compaq has a excellent service unit. It's about the only thing left from Dec other than Storage works [now gutted, mostly], and the Alpha line [though nothing new has been released that wasn't already being designed at the time of the purchase]. Honestly I don't see VA Linux beating Compaq on this point.

    Of course, I'd also like to see this product with Alphas, though that would never happen lest TruCluster sales would slip further.
  • While this is a good thing and all for Linux advocacy to the unwashed masses of PHBs, it really is only as good as the announcement.

    I won't consider Compaq, Dell or any others to be truely supporting Linux until:

    a. Linux is an option in paper catalogs.
    b. A server w/Linux pre-installed costs less than the equivalent NT version.

    I only say this because I would think PHBs and CTOs look more at paper catalogs than anything online (ooh - glossy paper)...

    Worldcom [worldcom.com] - Generation Duh!
  • I agree... the further you get away from the server end, the more Compaq machines generally suck. I think that a big part of their problem is that they treat all their machines (tech support wise) as if they WERE servers. It's great to have a riser board in a server, as well as little sliding drive bays, and rompaqs / softpaqs galore so you can fix all those little incompatibilities.

    However, on the non-server side, these things drive me NUTS! When user X can't get application Y running on his bastardized, Compaq patched-to-hell version of Windows 98, things aren't quite as easy to debug. And of course, user X can't be damned to follow all the little steps he needs to in order to fix the problem. Ah, hell...

    In short, I think this applies to Compaq more than it does to any other computer manufacturer selling home systems: Keep It Simple, STUPID!!
  • I think that was supposed to be funny. 'Reliability' and 'scalability'? 'Tarnish its image'? Compaq?? Ha! The whole thing reeks more of 'Funny' than 'Troll'. ;)

  • Re:How about a price? (Score:3)

    by vectro ( 54263 ) <vectro@pipeline.com> on Monday January 22, 2001 @11:10AM (#490307)
    If you have to ask, you can't afford it.

  • Not a beowulf cluster (Score:5)

    by vectro ( 54263 ) <vectro@pipeline.com> on Monday January 22, 2001 @11:05AM (#490308)
    If anyone had actually taken the time to read the linked page at Compaq, they would have noticed that this is NOT a high-performance (beowulf) cluster.

    This is a high-availibility cluster, the idea being that one of the machines can die without losing anything. Kinda like RAID taken to the systems level.

    That linux lacks this has been one of Microsoft's marketing points, so this is a really good thing. Unfortunately, the mirroring software is not open source or free software. :\
  • VA Linux.. They're 90% 'standard' hardware. (In some cases the mobo isn't exactly what you'd see on the component market)

    Oh, and Compaq usually says 3-5 business days, unless you have one of the 'hella support' contracts.
  • Keep in mind, nothing (single system) based on the PC arch can claim five nines. As we move out of the PC realm to DEC/Compaq, SGI and Sun, well they can't do it in a single 'system' either. (although the issue is foggy on what a single system entails as you move up the product offerings, and they do come close) I think Sequent used to claim five nines for a few products, but then again they weren't *exactly* single systems.

    In fact, the only current, conventionally single, 'system' I can think of that can claim five nines is the S/390.

    And it runs Linux.

  • Re:Reliability? (Score:3)

    by technos ( 73414 ) on Monday January 22, 2001 @11:58AM (#490311) Homepage Journal
    Nope. The best HP will guarantee with Win2K on a single server is 99.9, with a 24-hour on-site tech. That's THREE nines, not five. They will also be happy to sell you a Win2K HA cluster, but there again they will only guarantee you 99.99, once again with a 24-hour on-site tech. That's FOUR nines, not five. HP does sell a couple systems that claim five, but they're MPE/IX and don't exactly fit what one would call a 'single' system.
  • Yep, while this may seem like just one of many recent announcements of clustering products for Linux, LifeKeeper is a little different because it is based on proven high-availability (HA) technology that many Fortune 500 companies, including Chase Manhattan Bank and AT&T, use on a day-to-day basis to insure that business-critical applications keep running under any circumstances. LifeKeeper was originally developed by NCR [ncr.com] for its MP-RAS UNIX, where it competed head-to-head with UNIX-based HA clustering products from heavy-weights such as IBM and HP. NCR later ported LifeKeeper to Windows NT and Solaris, and when the company got out of the general systems business to focus on the retail market, it sold the technology off to Steeleye, a start-up founded by a group of industry veterans which is now doing the port to Linux. Steeleye's functional overview of LifeKeeper for Linux is here [ncr.com], and NCR still maintains detailed documentation for LifeKeeper here [ncr.com].

    The availability of LifeKeeper directly from Compaq is a big step forward for the deployment of Linux systems in enterprise environments because HA clusters need to be tested and supported on a system basis, i.e. HW and SW together.
  • At the University where I work, one of the research groups ordered one of these things. It seemed to me that they spent way too much money on something they could have put together, but it does look nice. Oh.. and someone didn't noticed that you have to order the OS for these, so the whole cabinet full of equipment sits idle with a "system is down" message on the led panels.

    Have to love compaq.

  • Comapq has been doing this w/Alpha Beowulfs for a long time through CSSD - a /. reader just found it so it's news??? PS - try to find a Compaq supported tape backup solution for it - HA! Compaq - we put the "HaHaHa" in HA.....
  • ...but it has to be done ;) "can you imagine a beowulf cluster of these?"

  • Its good to finally see that the main players are starting to be a little more active about their Linux servers and not just making idle promises like so many have done in the past. This promises to make high end servers that are truly reliable a much easier and affordable option for many companies out there.

    As an aside, now that Sun owns Cobalt, does anyone know if the Cobalt servers are going to have Linux replaced by SunOS or whether Sun intends to keep Linux as the primary OS for the Cobalt servers. In all honesty, I cannot see devices like the Cube and so on remaining competitive if a proprietry/commercial OS like SunOS is used. They would then have to directly compete with Windows NT/2K and HP's servers.

    With Compaq up'ing the stakes like this, now what you truly are paying for is the server and the technology. Most excellent. This will truly help the Linux community increase its reputation as well as put some real servers out there that can be used to directly compare stats against the other more established server OS's.

    Kudos Compaq, Kudos IBM

  • As far as I'm aware, IBM does all their own support... As for Compaq, I'd say the majority of their Linux support would be external. Based on what I know of the Presario & Deskpro support structures, I'd say that Compaq will support the hardware and Compaq specific software, but the Linux support itself will probably be outsourced. Either that or Compaq is going to be doing some serious overhaul of their Proliant support teams.
  • Now i can finally fill up all that unused closet space. How many nodes should i get? Oh wait, i forgot, i can't even afford _one_ server...maybe some other time =)
  • I'll give you the benefit of the doubt in that you may be an experienced computer user, but as for linux's alleged unreliablity, all i can say is, prove it.

  • LifeKeeper and compaq (Score:3)

    by drshannon ( 105510 ) on Monday January 22, 2001 @11:49AM (#490320)
    I work for a company that had puchased several compaq DL580's and evaluated a few different cluster/failover/HA solutions. The decision to to with LifeKeeper [lifekeeper.com] was made just before I joined the company.

    It has been part of my job to get the system implememted with RH6.2, Apache, Informix and a shared mass storage device.

    A bit of background on LifeKeeper first. Lifekeeper is a process monitoring tool that monitors a number of services and its dependencies, and manages the re-start and failover of these services to a group of defined servers.

    If you can imagine a stacks of servers with, for example, a firewall, a web server, and a db server as a unit, you can implement failover in a N+N configuration.

    primary.......backup

    firewall......firewall
    apache........apache
    informix......informix

    Each layer has a virtual Ip that flips back and forth between primary and failover so that each layer knows how to talk up and down to the other layers blindly.

    We found that we had to do a lot of custom work to get everything implemented the way we wanted it.

    The Informix kit was a bit messy to failover, however apache and tomcat failed over very well. we had some issues with the NFS kit, as our Informix DB was mounted on the shared mass store.

    We had and still have some issues with our firewalls. We are using gShield (ipchains) for our firewalls, and the gshieldconf tool allows for only one IP per box, and we had the real IP's plus a Virtual IP that floated between the firealls. (our production url resolves to this virtual IP). We loose connectivity to our primary firewall on its _real_ ip, and must access it from the VirtIP. It went on to cause us problems with routing mail up thru the stacks, with accessing customer sites from behind the firewalls, and a few other things. Bottom line, Lifekeeper doesnt work so well to fail firewalls over.

    It is pretty neat to watch apache or informix or tomcat (or our firewalls for that matter) to flip over from the primary to failover box.

    It ends up that all services that LifeKeeper Monitors need to be started by lifekeepr with some command such as: perfome_action -t apache -a restore, instead of apachecrl start.

    we are putting ssl on the boxes right now and will see how Lifekeeper handles mod_ssl!

    Neat product, would love to see it open.

    Its also too bad, because the dl580's compaq sent us took about 3 weeks to get up and running, lots of wierd hardware problems with the pci bus.
  • Sure, this isn't the best choice for the small cluster-building market, as pointed out, but for those companies that have the money and can ramp up a project faster -- the benefits will trickle down to individuals pretty quickly.

    ... from design improvements to better customer service to anything that isn't already sequencing genes or beating Garry Kasparov. This sounds cool.

  • Doesn't sound like anything that the already open-source Linux Failsafe [sgi.com] won't do. Not that there is anything wrong with lifekeeper.

    -dB

  • imagine a beowulf cluster of these rendering images of natalie portman being upset that the goatse.cx guys are more of a popular download than she is now. :)
  • I'd rather have off the shelf HW. Proliant servers are what I what call commodity HW. These things are a pain in the rear. Anyone offering pre-configged/built clusters using "standard HW"? I'm glad to see Compaq doing this, but I would like to see use of HW that if it takes a crap, I can run to the local Computer store and be back up before lunch. How long would it take to get a replacement PS from Compaq? (Yes, I realize they should be redeundant/hot-swappable)
  • It shows how compaq really views linux. The graphic on the page linked shows a tiny R/C car with "TUX" behind the wheel. Hah.
  • "commodity hardware" is the crap that you buy from computer stores, not what you get in a server from a large vendor like Compaq or HP.

    I've seen far more "servers" built using PC-grade parts fail, than I have seen servers (and I admin 280 HP servers) go down due to failed power supplies. The amount of cooling, as well as the quality and size of the power supply components is vastly better in those custom power supplies you disklike so much.

    If your server availability is important enough to you to have redundant power supplies, then it should be important enough to get at least a 24-hour service level agreement in place with your hardware vendor (you could also stockpile a few spare parts).

  • Not really, but combined with the HA stuff at Linux-HA [linux-ha.org], we can get a load-balanced, HA system. We have a running hack at Virtual Malaysia [virtualmalaysia.org] that does this. Basically, we put a Linux box at the front of a couple of NT web servers with monitoring software to check for downed boxes. We just have to add another box for failover and LVS + HA should be complete.

  • Argh... Link to Virtual Malaysia was wrong. It's a dotcom thingy.

    Here's [virtualmalaysia.com] the right link.

  • this is NOT a high-performance (beowulf) cluster.

    Umm, sorry? Are you implying that the only high performance clustering solution available is beowulf? Stop smoking the cheap crack, it's messing with your head. Most clusters are not focused on computation, but providing services. Often the bottleneck that's being overcome is not the CPU.
    Just because it's not plotting weather patterns doesn't mean that it's not high performance. The performance is just measured in other ways (HTTP req's /sec, FTP throughput, etc. etc.) This is an infinitely more common, and to most, more useful type of cluster. Sometimes I think the /. crowd is a little _too_ fixated on beowulf clusters...

    That linux lacks this has been one of Microsoft's marketing points, so this is a really good thing.

    That you _believe_ that Linux lacks this is testimony to the success of Microsft's marketing...
    It's here, and it's been around for a while. http://www.linuxvirtualserver.org [linuxvirtualserver.org]
    is where you can read about it. Not only does LinuxDirector exist, but it scales farther than MS's offerings, is GPL'd, and in typical Linux fashion, doesn't require fancy, expensive, matching hardware...
  • Two weeks ago I got a Dell Precision 220 with RedHat 7.0 preinstalled.

    It actually cost $300 more than the same hardware with NT (: But I did it to save dealing with drivers for state-of-art stuff which burned me in the past; saving a few days on company time is worth it.

    The sad thing is that it had only the original obsolete 7.0 without security patches available 6 months ago. Last week, within hours of hooking it to the Internet to install security patches, it was wiped out by the Ramen worm. Chicken and egg problem: how do I get the security patches without hooking to the Internet, and how do I hook to the Internet without the security patches? (OK, I could have used another computer but you see my point...)

    Anyway the OS is reinstalled, and it rocks... :)

    Moral: If you get a preconfigured Linux, check the security updates status pronto.

  • Compaq has a product that has achieved better than five nines throughout it's entire installed customer base. A recent audit of the Compaq/Tandem Himalaya system showed that they achieved almost six nines of reliability, including "minimal" systems. Of course the minimal system is fully fault tolerant, including 2 CPU units, each unit using multiple CPUs.

    As a former Tandem employee, what I find interesting is the wide range of HA technology being used by Compaq. In addition to the Nonstop Guardian systems (Tandem), they have DEC Tru-Clusters, Microsoft HA clustering, and now the LifeKeeper solution (this thread). They also recently announced they were licensing unspecified "stuff" from Stratus (formerly the 2nd banana in the Fault-Tolerant market behind Tandem). I know that no one technology works best for all situations, but this is getting ridiculous.

  • In the loosest most overhyped sense of the word, you MIGHT be able to say this is a cluster. With a maximum of only 2 computers / "cluster", it might as well be Wolfpack. You know, that NT Enterprise cluster. You have quorum devices, etc. What damn use is it. You can't even use it for a high performance High availability webfarm. -SB-

  • Compaq Linux Clusters - stay crispy in milk and give you your RDA of iron!

    Brian
    Your Tech Humor Source [bbspot.com]
  • ...but just wait 'til Compaq becomes an ASP for Beowulf Clusters. Wow!
  • Okay, I generally HATE Compaq, I will acknowledge several truths about them and their products:
    • Presario == much suckage
    • Prolinea == suckage
    • Deskpro == borderline suckage
    • Proliant == not that bad, actually
    the reason they're not that bad is that the methods to maintain the systems are well documented, and they do actually seem to work. I once was upgrading from a PII-400 to a PIII-550 in a Proliant 1600-R, and the steps were very simple... download a new ROM image floppy, boot the server with the floppy, burn the ROM, power down the box, and swap the processor/fix jumpers. It worked. the old bios was something out of 1997 or 1998, and the new one was from late December 2000, and it didn't fail. I'd imagine that a new image will let me put any non-XEON PIII in the case, without having to make drastic parts changes.

    This may sound a little roundabout, but when one has to support 50+ of these, having a working method is a godsend...

    "Titanic was 3hr and 17min long. They could have lost 3hr and 17min from that."
  • ... and so far I've been quite nicely surprised with the Proliants. I had one ML/370 that had problems with a power outage that fried the mainboard, but that was all that has gone wrong. We deal with about 50 or so of these beasties.

    We do have one of those paranoid service contracts from hell where the Compaq paid technician will come out the next day and fix it...

    "Titanic was 3hr and 17min long. They could have lost 3hr and 17min from that."
  • Compaq can't even make decent PCs. What makes the company think that it can build Linux clusters?
  • Don't forget Dell. They've been offering servers with RH 6.2 for at least 6 or 8 months now, maybe even longer. I know for a fact that Dell outsources its linux server support. I doubt IBM does that but do you think that Compaq does?
    "Me Ted"
  • This is a high-availibility cluster, the idea being that one of the machines can die without losing anything. Kinda like RAID taken to the systems level

    Personally, it's my belief that the vast and overwhelming majority of /.ers are much more likely to need high-availability fall-over systems.

    The main question most of would have now pertains to security. Granted, it's not BSD, or SecureLinux, but is this a buttoned-down failover system or does it come in the vanilla wrapper with all the ports open and services ready to be exploited?

    Let's hope it's fairly buttoned down with some easy admin tools.

  • True, but at least the media and every other lame brain out there will at least take note of the fact that 'Hey! Compaq is
    now selling Linux with its computers. Maybe that's a good operating system.'
    Not quite. What journalist would be looking at anything that would relate to these clusters that would not already know the power of Linux, that would be swayed by such a small quantity?
  • It's like a frickin' card game! "Hey there pal, my 5 9's beat yer 4 9's any day where I come from!"

    Heck, based on that little marketing strategy, Linux just better tuck their little wagger between the legs and head for the porch.



    .
  • This doesn't surprise me at all. I used to work with a guy that was an assembly programmer (later on a tester) with NCR for years back in the 70's. Their corporate strategy was simple; figure out what IBM is doing, then copy it.

    Of course, IBM was something of a fuckup themselves a while back, and with NCR making cheap knockoffs of everything IBM was doing, I suppose it could only be for the better of that software to be placed in the hands of some other company.

    The same could be said for the products of my company. We were one of dozens of fragments that were cast from Control Data Corporation in its tailspin before they crashed and burned -- most of them are doing OK, but AFAIK what remains of CDC is gone, at least by name.

    --

  • Yet another step to getting Linux more mainstream. I'm glad to see that large companies like Dell, Compaq, etc. are finding that M$ Windows isn't the only solution for users needs these days.
  • True, but at least the media and every other lame brain out there will at least take note of the fact that 'Hey! Compaq is now selling Linux with its computers. Maybe that's a good operating system.'
  • I just scanned the compaq site, best i could see there's no mention of price... anyone got any leads?

Slashdot Top Deals

"The medium is the massage." -- Crazy Nigel

Close