Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
IBM

Sendmail On IBM Mainframes Running GNU/Linux 132

Posted by timothy from the sends-mail dept.
raffe writes: "Cnet reports that Sendmail has released a version of its e-mail server software that can run on Linux-powered IBM mainframe computers. In one benchmark test, IBM found that it was possible to house 2 million e-mail accounts on a single server, with 10 percent of the users accessing their mail at any given moment" For some reason though, IBM zSeries machines aren't listed at pricewatch ;)
This discussion has been archived. No new comments can be posted.

Sendmail On IBM Mainframes Running GNU/Linux More

Sendmail On IBM Mainframes Running GNU/Linux

Comments Filter:
  • its just less buggy than sendmail.
    • Qmail is stronger only because it doesn't run as root. Sendmail can do the same thing [coker.com.au].
      • sendmail is one big monolithic piece of software, qmail is not, and qmail is much simpler than sendmail, too. sendmail doesn't suddenly become ultra-secure just because you don't run it as root.
        • I'm not going to get in a flame war with you but SENDMAIL IS BETTER.
          • I didn't want to say something else. IMHO, qmail and sendmail are so different that you can hardly say which one is better in general, only for specific purposes there might be a clear winner. (If your main criterion is flexibility, sendmail clearly wins hands down.) If you think qmail can do the job, go with it, if you need a development environment for email address parsers, use sendmail.

            Personally, I use Exim where I can; another large, monolithic piece of software...
          • LOL! Obviously you've never used qmail.
      • Qmail and Postfix are stronger for a variety of reasons. As well as a more modular design, both are designed and configured for modern environments - no 300k configuration file full of UUCP / X400 configuration stuff people won't use. More things to configure mean more things to misconfigure and go wrong. Sendmail also doesn't support maildir AFAIK.

        Not to mention violating Unix philosophy: text should be a common interface. Sure apps like LDAP and RPM use databases to keep their configuration data in (and simply allow interaction via text), but this is for performance reasons rather than legacy compatibility (i.e, just that should have been disposed of or made optional some time ago).
    • I use nothing but Qmail. Its own programs don't even trust each other, much less the outside world. Qmail even offers a Security Guarantee [cr.yp.to] - I'd like to see Sendmail do that...

      But on topic, I think the port is cool anyway. I'd personally love to get ahold of an S/390, and run about 40 virtual Linux boxes within it. If someone owns one of the virtual boxen (via a Sendmail sploit for example), I suppose it'd be easy enough to clean up...
    • No, no, no.
      If you want to go the "Sendmail [sendmail.org] is buggy" way, well, at least, try to be informative where the alternatives are concerned.

      For those who wish to try another MTA, the three big ones, not counting Sendmail, are Exim [exim.org] (small and easy, good for your home net), Qmail [qmail.org], and Postfix [postfix.org] (fast and powerful, my personal fav). All four have their good points, and all four are certainly worth checking before you decide on one.

      See? I mean, if Sendmail is still so widely used, there is a reason, you know... :)

      • No, no, no.
        If you want to go the "Windows [microsoft.com] [microsoft.com] is buggy" way, well, at least, try to be informative where the alternatives are concerned.

        For those who wish to try another OS, the three big ones, not counting Windows, are BSD [bsd.org] [bsd.org] (four powerful, secure, and robuts variations), Linux [linux.org] [linux.org] (more distributions than you can shake a CAT 5 at), and Solaris [sun.com] [sun.com] (The premiere commercial *Nix). All four have their good points, and all four are certainly worth checking before you decide on one.

        See? I mean, if Windows is still so widely used, there is a reason, you know... :)

        • Funny. :) And good point, too. If Windows is still widely used, there IS a reason (we may or may not like it -- as I pointed out, my fav MTA is Postfix, not Sendmail). Try having your grand'ma understand BSD, Linux or Solaris. :)

          Bottom line: Sendmail works for a lot of people -- and some of them even know what they are doing. :) Since all MTAs are interoperable, as opposed to OSes, it is not a problem for you if other people prefer other MTAs, so let's just all be happy that we have a good choice of them.

          • And Postfix is my favorite, too, after having administered Sendmail for 8 years and Qmail for 1 year. I wouldn't go back. It's quite close in many respects to how I would make an MTA.

      • Widely used != Good

        See: Microsoft Windows

  • Wow! (Score:3, Funny)

    by roguerez ( 319598 ) on Monday August 27, 2001 @09:03AM (#2220869) Homepage
    This makes the ultimate spamrelay.. ;)

  • The mainframe's not dead... (Score:5, Interesting)

    by Jay Maynard ( 54798 ) on Monday August 27, 2001 @09:05AM (#2220874) Homepage
    This is the kind of thing that mainframes do well: information processing with little or no actual computation. Their I/O abilities really make the difference here.


    Hopefully, this kind of result will show the skeptics that there's a real purpose for the big boxes.

    • Anyone got some numbers on running something like qmail on the same hardware?
    • There was an article in the April 98 Byte that went into the advantages of mainframes over PC's. Too bad the print archive aren't online anymore. When you buy a mainframe, you are buying support. Having the OS crash on a PC is an event that, while rarer than in the recent past, is still fairly common. Yes, even with Linux. When it crashes you reboot and, if you are unlucky, reload from backup. Having the OS crash on a mainframe is a dire event that results in a team of engineers being put on the next flight out to your site. The same applies to hardware problems. PC's have uptime measured in months ,and sometimes years. Mainframes have uptime measured in years, and sometimes in decades.
      • A recent issue of the IBM Journal of Research and Development was devoted to reliability issues in the G5 and G6 series chipsets used in recent IBM big iron.

        Let's see - one stray alpha particle can cause a Pentium III to crash. It's not the chip's fault, it's just the way it is.

        On the other hand, here's what the IBM mainframes call error recovery:

        1) Each CPU chip is actually 2 complete CPU's running in lockstep with a "tell me twice" comparator.

        2) At the end of each instruction, the entire internal states of both sides are compared, and if they match, the state is latched out for safekeeping, and the next instruction is started.

        3) If the two sides *dont* compare, this is a "soft" error. The current state is latched out for offline analysis. The saved state from the latch-out is reloaded, and the instruction is retried. The reload will clear any corrupted bits due to alpha hits or the like., so this is all that's needed for recovery.

        4) If after a retry the two sides still don't agree, the known-good latch-out is then loaded into an entirely new spare CPU chipset (a common configuration is 12 CPU and 2 spare, from what I read) and execution is resumed on the new CPU, with no impact on processing.

        5) You don't get a actual "CPU failed" error until it's done a soft retry and then moved to a spare repeatedly, and run out of spares. THAT is why you end up with engineers on a plane - to get to that point the machine has to be seriously sick.
        • I met a guy who worked in a mainframe shop. He said that one day an engineer showed up to fix the machine. They didn't know it was broken. Apparently it reported a problem to the home office, which dispatched a tech to fix the problem.
          • My friend worked in a similar office. (unfortunately, he isn't in right now otherwise I would get the name of the machine. Something starting with an S. It had redundant EVERYTHING on it.) Not only did he have engineers show up but every now and then a FEDX would arrive with a new part for the machine. They didn't even know it was down. Sure enough they would take the part out back and there would be an error message for the affected part. This thing called home all the time.

            A very cool box. They don't make them like that any more.
            • Yes, they do. We run an AS/400 which does exactly the same thing. Twice we've had a hard drive 'about to' go bad. IBM called us up and arranged the replacement part and the technician. No down time, since the system is all about RAID.
          • The IBM 3090-300J processor controller would phone home. The part that floored me was when it put up a dialog box on the hardware control console, asking for permission to make a long-distance phone call...

            Definitely makes you pause when it happens at 2AM.,
        • This may not be the same article, but it has some interesting descriptions of reliability enhancing features in mainframes.

          IBM S/390 Parallel Enterprise Server G5 fault tolerance: A historical perspective [ibm.com]
          by L. Spainhower and T. A. Gregg
          IBM Journal of Research and Development
          Vol. 43, No. 5/6 - IBM S/390 Server G5/G6

    • Re:The mainframe's not dead...

      That's right IT'S JUST RESTING!!
  • If only I could convince our sysprog to give me an LPAR on ours.

  • The rest of the article... (Score:3, Funny)

    by Anonymous Coward on Monday August 27, 2001 @09:11AM (#2220896)
    In one benchmark test, IBM found that it was possible to house 2 million e-mail accounts on a single server, with 10 percent of the users accessing their mail at any given moment..."


    ...And in another benchmark test, IBM was also proud to announce that the massive I/O and processing infrastructure of its flagship zServer range was able to sustain 2 million Sendmail security holes, with 10 percent of the holes being exploited at any given moment...truly a new world record.

  • ...when I see mainframes running Exchange.
    • So would I - but I don't think MS are able handle the task.

    • I'll be impressed when I see mainframes running Exchange.

      Aw, now, why would you want to go and do that to a nice mainframe? :-) If you really want to "run Exchange" on a mainframe, give Bynari [bynari.net] a call. They've ported TradeServer [bynari.net] to Linux/390. So yes, you can move your MS-Outlook users to Linux/390-based email. Today! Just ask Winnebago - they're doing it. [ibm.com]

    • that would be awesome to see the first piece of software ever to bring a mainframe to its knees
  • OS/390 USS ("Unix System Services") has had a sendmail port for a long time. Can't speak to its performance -- we only use it for low volume outbound mail -- but it *is* sendmail.

  • Sendmail's an MTA not a MUA (Score:3, Insightful)

    by swb ( 14022 ) on Monday August 27, 2001 @09:34AM (#2220964)
    No end user can access their mail with Sendmail, it's a mail transfer agent for relaying mail, intra or inter-node.

    Mail access means reading the end-user spool through the usual MUAs and support daemons: Pine, Elm, mail(1), imapd, pop, etc.

    End users do use sendmail to relay mail, but they can't access their own mail that way.

  • Misleading (Score:2, Informative)

    by Syberghost ( 10557 )
    The test was of 400,000 users, not 2 million; the 2 million number is a projection that has not been tested.

    If we're going to pretend we're journalists, let's pretend we took at least one semester of it, shall we?
    • whatis 10% of 2 million ??

      i've hate to break it to you .. they said.. the system supported 10% of 2million concurrently

      ofcousse 2 million was never tested

      it was the amount of storage that could be given away .a.k.a. hard disk space..and the limits on sendmail.

      i'm sure this isnt' a mindcraft benchmark

      and IBM has every right to blow their trumpets

      • whatis 10% of 2 million ??

        An irrelevant, untested number.

        i've hate to break it to you .. they said.. the system supported 10% of 2million concurrently

        No, jackass, they said it supported 10% of 400,000 concurrently.

        Next time, read the fucking article before you go correcting your elders.
  • This thing costs 1.2 Million dollars! [ibm.com] Image what I could build with 1.2 million in cheap clone hardware... I think I could do at least twice as much processing, and include the pop/imap servers.
    • You certainly could buy a lot of hardware for that money, but who is going to want to admin 1.2 million worth of pc hardware all running in one cluster? You would also need a very large warehouse to store all of those boxes, and the power consumption would be a bitch too. In the long run you would be much better off with the mainframe if you really need that amount of processing power.

    • Re:1.2 Million dollars! (Score:5, Insightful)

      by Jay Maynard ( 54798 ) on Monday August 27, 2001 @10:28AM (#2221153) Homepage
      Image what I could build with 1.2 million in cheap clone hardware... I think I could do at least twice as much processing, and include the pop/imap servers.

      Then you get to maintain and run those thousand boxes. Consider power, floor space, and most importantly, people requirements. (Are you going to maintain those systems yourself? Two or three people, maybe? I don't think so.)


      Sometimes you do get what you pay for.

      • The problem is that if you buy 1.2 million in cheap clone hardware, is that you get cheap clone hardware. The 390 mainframe is setup so that if you set it up right it will go years without a reboot. The 1000 pc clones will need a small army of people to look after them etc. Go with the mainframe!
      • Then you get to maintain and run those thousand boxes. Consider power, floor space, and most importantly, people requirements. (Are you going to maintain those systems yourself? Two or three people, maybe? I don't think so.)

        It's only hard if you don't know how to prepare the setup required to maintain $1 millon in PC hosts:

        http://www.infrastructures.org [infrastructures.org]

        Computers are amazingly good at automatic repetetive logical tasks. 99% of all systems administration involves repetitive logical tasks. The trick is to make the machines do all the work of maintaining themselves that they can programmatically handle.

        So if you're a clueless "paper" sysadmin, yep, it's impossible -- can't be done.

        Still, I'd rather have the z-machine mainframe!

    • You might be right, you might have more raw computrons with 1.2 M $ of PCs than a z/390. But not everything lends itself well to distributed computation. So depending on what you run, you could see lower performance and hellish administration.

      You're also forgetting the point of a mainframe - HA. PCs aren't designed for reliability. When you administer a cluster you expect a certain rate of hardware failures. A mainframe is expected to have 99.999+ percent uptime. The hardware is fully scoped, localized, and hot-swappable - right down to the processors. A company that's looking for that kind of uptime really has no other options. What all this Linux/390 stuff is about is selling Linux to groups who won't compromise on the uptime.

    • 1.2 Million dollars is the baseline system with 1 processor and 5GB memory. After you add the maximum 640 processors, parallel sysplex, a couple of TB of DASD I'm sure you could get over 10 mil. Any 390 admins around here? How much is the StorageTEK silo to back it up :)

      Of course you're looking at the suprime icon of reliability. Why do you think that banks rely on them for all their processing? A bank can't afford downtime or lost data...

  • whaa-? (Score:2, Funny)

    by guest12 ( 248543 )
    oh. for a moment I read "house" as "hose" million email accounts.
  • Comment removed based on user account deletion
    • If microsquish could get over the embarrassment of admitting that NT will never be able to do the job, they could redeploy hotmail on IBM equipment, and make it worth using.

      The back-end already lives on Sun boxen (Enterprise 4500) - not quite the big-iron of IBM's zSeries/S/390, but a far cry from an NT system... Only the front end - the WWW servers and CGIs - run on NT, on the PC hardware which ran FreeBSD earlier.

  • Uh oh! (Score:2, Funny)

    by jpmorgan ( 517966 )
    IBM announces the Sendmail MTA has been ported to Linux390. For the first time spammers will be able to send mail to everyone in the world at once.

    It brings a tear to my eye. *sniff*

  • Hey, cool! Microsoft now has a solution to their Hotmail scalability issues!

  • GNU/Linux?? (Score:1, Offtopic)

    by duffbeer703 ( 177751 )
    Maybe I've been smoking too much crack, but I though CmdrTaco stated one short week ago that Linux should not be referred to as GNU/Linux?

    Fuck the lameness filter, here's something really lame, courtesey of Slash:

    # interpolative hash ref. Got these figures by testing out
    # several paragraphs of text and saw how each compressed
    # the key is the ratio it should compress, the array lower,upper
    # for the ratio. These ratios are _very_ conservative
    # a comment has to be absolute shit to trip this off
    if (!$bad) {
    my $limits = {
    1.3 => [10,19],
    1.1 => [20,29],
    .8 => [30,44],
    .5 => [45,99],
    .4 => [100,199],
    .3 => [200,299],
    .2 => [300,399],
    .1 => [400,1000000],
    };

    # Ok, one list ditch effort to skew out the trolls!
    if (length($$comm) >= 10) {
    for (keys %$limits) {
    # DEBUG
    # print "ratio $_ lower $limits->{$_}->[0] upper $limits->{$_}->[1]<br>\n";
    # if it's within lower to upper
    if (length($$comm) >= $limits->{$_}->[0] &&
    length($$comm) <= $limits->{$_}->[1]) {

    # if is >= the ratio, then it's most likely a
    # troll comment
    if ((length(compress($$comm)) /
    length($$comm)) <= $_) {

    # blammo luser
    $$error_message = slashDisplay('errors', {
    type => 'compress filter',
    ratio => $_,
    }, 1);
    editComment('', $$error_message), return unless $preview;
    $bad = 1;
    last;
    }

    }
    }
    }
    }

  • What did they use to generate 400k users? (Score:1, Interesting)

    by Anonymous Coward
    Did they use Specmail2001 (from http://www.spec.org) to generate the load on their MTA? If so, I don't think that 400k is that special.

    I've generated very close to that number (minus 100k) with a dual 1Ghz cpu w/1Gb ram system, and a Netapp F820 for storage with the Syntegra Intrastore product. I even had the close to the same CPU utilization they had. This is user space CPU of course. The system CPU time is higher due to the NFS overhead.

    Once I get around a linux 2.4.x virtual memory problems, I'll publish the results to the spec.org site for everyone to see.

    http://intrastore.us.syntegra.com
    • You did that with ONE system? Interesting - the Mirapoint [spec.org]results for 400K users needed:
      • 2 POP server machines
      • 5 SMTP router machines
      • 10 message store machines
      • 1 benchmark manager machine
      • 1 mail sink machine
      • 5 load generator machines

      Dealing with 300K outbound postings is no biggie - I've been able to deal with that level on an old IBM RS6000-F30 (166mz 604). You don't need really big iron for outbound mail until you have more than 500K or so RCPT TO's on one piece of mail. It's mostly a matter of good queue management, and Sendmail 8.12 has new queue management code that makes it even easier (I should know, I tested it). The only real magic is not getting logjammed due to DNS waits and unreachable destinations.

      On the other hand, having 40K people doing POP accesses while you're dumping mail into their mailboxes is trickier. Some of the more obvious issues:

      • The obvious popd solution leaves you 40K processes running at once. This could be bad.
      • Locking issues get interesting.
      • Even with a journaling filesystem, you can get killed on the I/O. Remember that writing to a file means you also need to do stuff with the inode....
  • I got an account on the z series developer box where I get my own virtual server and compiled qmail, bind, etc, etc. Why did they have to release a special sendmail to run on it? Did they simply cross-compile it?
  • Now if they'd only do that internally, instead of that !@$#%! Lotus Notes (Bloated Goats.) Maybe IBM should practise what they preach...


    While we're on the subject, if IBM is so gung-ho about open standards, why haven't we seen any Lotus file formats documented? It sure would be nice if I could load up all those WordPro and 123 documents using Abiword and Gnumeric.

  • I read an article that Sendmail, Inc., uses IBM Mainframes for their development anyway.... Sounds to me like IBM just convinced them to sell/support the product natively on that platform.

  • How come IBM doesn't at least try to use Postfix [postfix.org]? I mean, Postfix is an IBM-funded thing, and was developed to be the, quote, "IBM Secure Mailer"...
    • How come IBM doesn't at least try to use Postfix [postfix.org]? I mean, Postfix is an IBM-funded thing, and was developed to be the, quote, "IBM Secure Mailer"...

      Probably because it is so damned hard to even get access to a S/390 or zSeries virtual machine account to do anything serious with. I'd love to port, test, and package my stuff on more platforms, including mainframe, but an account that is limited to 3 months doesn't work for ongoing projects that never end. And one of my projects needs 2 with dual shared DASD. And those guys at IBM never responded to any of my email. So as much as I'd love to work on the mainframe, I'll just stick with Intel, Sparc, and maybe soon PPC.

  • Remember back when Unix was 12 vendors all yelling about each other that their competitors' Unix sucked and that theirs was best? Who won?

    NT

    It came up behind while the big boys of Unix were standing in their circle peeing at each other.



    In corporate-land, the ones that have mainframes already and are facing huge IT costs and a recession, the ones who are winning the mailboxes are Exchange and Notes. They had virtually no share 10 years ago, now they have lots of network share. They also cost a lot to run (Gartner says $25+ per mailbox per month).



    Now here's a company that runs on Unix, that has an IMAP server that can scale HUGELY on one (or many) boxes. That can give Secretary Joe the ability to do the admin on his group's 100 users and do that for 200 groups so that the system admin can do more important things than deal with adding a mailbox for this month's temp receptionist.



    QMail? Postifix? Who? Go talk to the CEO's, the stockholders. Given Dan's support group a call at 4AM when your TLS mail isn't working right [securityfocus.com] or general stability of the organization, this isn't a choice for those who don't really want to spend all their money running their computers.



    Recall that when you're trying to run mail for 500+ people, there just aren't a lot of options out there. Notes and Exchange tack on the IMAP letters on their product and claim it supports standards.

    For those in the Real World, take a look around at how many actual standards based tools there are with solid commercial support.



    So Sendmail's MTA, IMAP server and Webmail client run on the Mainframe!? Bitchin', now I have something to counter those MSCE's who claim that we must run Exchange to survive.

  • sendmail doesn't give users any "access" to their mail ... do you mean that it runs a pop/imap server as well? I imagine it well could

  • "GNU/Linux"??? (Score:1, Offtopic)

    by tswinzig ( 210999 )
    Timothy, you pushover!

  • ARGH!! (Score:2)

    by cabbey ( 8697 )
    and as tommorow dawns and the hundreds of /. reading beemers trickle in and scan slashdot while waiting for Lotus Notes to load up their mail file there will be a cacophony of anguished screams as the poor abused geeks are left stammering in their seats trying their best to explain why the corporation still insists on loading them onto windows based domino servers (excepting those at the IBM Rochester facility who are lucky enough to be housed on AS/400 based domino servers instead).

    (Don't get me wrong folks, domino is a great database and colaboration tool, but that's just NOT what is needed for an email solution.)

Slashdot Top Deals

The one day you'd sell your soul for something, souls are a glut.

Close