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

 


Forgot your password?
Close
typodupeerror
×
Software Operating Systems Unix

Heavy-Duty System Administration Utilities? 44

Posted by Cliff from the more-transactions-than-you-can-shake-a-stick-at dept.
leandrod asks: "I am in the process of helping a small software company define the infrastructure for their major client's new system. It is a big country, and it is a medium-sized client planning on going big. We are planning to standardize on Debian GNU/Linux. I am aware I can have IBM Tivoli Maestro for GNU/Linux for production scheduling, and BEA's Tuxedo TP monitor, but they are unsupported under Debian. I am also aware of one or two free TP monitors, but they are either incipient or stagnating. I couldn't find a production scheduler. I know I can do lots with the standard tools, but keep in mind I am targeting a transaction-processing bureau for a big operation with hundreds of thousands of terminals and millions of users, something like a poor man's Wal-Mart, or even Visa. Are there vendors out there willing to support Debian or just GNU/Linux in general? If not, are there free software projects that accomplish the same thing?"
This discussion has been archived. No new comments can be posted.

Heavy-Duty System Administration Utilities? More

Heavy-Duty System Administration Utilities?

Comments Filter:

  • Support (Score:2, Interesting)

    by !3ren ( 686818 )
    Perhaps this major client could fund development/maintenance on a missing piece of infrastructure.

  • Think long-term stability... (Score:5, Insightful)

    by duffbeer703 ( 177751 ) * on Friday February 13, 2004 @01:03AM (#8266738)
    Standardizing on Debian is a really lousy idea. At some point, your customer is likely to need or want commercial software (esp RDBMS), and they will find that Debian is simply not supported by any commercial software vendor. (To be honest, most non-Linux geeks have never heard of Debian)

    Go with Red Hat or Suse. You might find that going with a more stable (from a support POV) Unix OS like Solaris may be a good choice for certain systems as well. The support costs are real, but a Tivoli Management environment would cost a helluva lot more if the IBM salesfolk talk your client into it.

    If you in a signifigant transaction processing business, the money will come - spend the money now to start a Maestro or Tuxedo system so you don't need to waste valuable time (and lose business) later.

    I also hate to say this a longtime Debian fan... but the major commercial distros aren't going anywhere. RedHat and Suse have built brands and have major money & support flowing in from corps like IBM & HP. Can the same be said for Debian, whose stable release is starting to get a little crusty?

    Remember to ask yourself what you & your client needs and what is best for the business. Keep the tech-geek religious wars on Slashdot!
    • One interesting addendum to the above: SUSE Linux to receive international EAL2 rating [com.com]. Since SUSE is owned by Novell now, it also has the backing of a company that had, at least some time ago, the most popular network suite in corporations. Still in use a lot of places, actually.

    • Re:Think long-term stability... (Score:4, Funny)

      by the_other_one ( 178565 ) on Friday February 13, 2004 @01:19AM (#8266810) Homepage
      Debian is commercial software. For hundreds of thounsands of terminals, we at SCO, will be happy to provide invoices.

    • Agreed.

      Debian packages are too old, unless you want to jump into the "unstable" tree - and what business can accept unstable server software? Back-porting bugfixes just don't cut it in the business world. Debian lacks a powerful entity that governs development priorities - but it doesn't lack zealots who will die for their cause.

      For example... Debian hasn't got XFree86 4.3.0 in their stable tree (and they wont for years to come as they still have 4.1.0 stable) - whereas RedHat has had a stable XFree86 4.3

      • Re:Think long-term stability... (Score:5, Funny)

        by swillden ( 191260 ) * <shawn-ds@willden.org> on Friday February 13, 2004 @02:58AM (#8267326) Journal

        Debian packages are too old, unless you want to jump into the "unstable" tree - and what business can accept unstable server software?... For example... Debian hasn't got XFree86 4.3.0 in their stable tree (and they wont for years to come as they still have 4.1.0 stable)

        Right. Everyone knows that you absolutely cannot run a large-scale production server effectively without the new features in XFree86 4.3. It's obvious that 4.1 just doesn't cut it when you have millions of transactions to process... I mean, it can't even change the screen resolution on the fly!

        • Re:Think long-term stability... (Score:4, Informative)

          by Drakon ( 414580 ) * on Friday February 13, 2004 @04:49AM (#8267675) Journal
          I mean, it can't even change the screen resolution on the fly!


          ctrl+shift+(numpad Plus/minus)
        • You're missing the point. X is an obvious example, there are plenty of others.

          • Re:Think long-term stability... (Score:2, Interesting)

            by Anonymous Coward
            > You're missing the point. X is an obvious example, there are plenty of others.

            He's hardly missing the point, you are.

            New-featured software is NOT what a production business server needs. Such servers need stable, well-tested software that is patched for security and serious functional problems - but NOT constantly enhanced with new features that contain their own set of new bugs. The idea of business server software is to gradually approach a bugfree state - and you don't get that by constantly addin

          • Re:Think long-term stability... (Score:4, Informative)

            by swillden ( 191260 ) * <shawn-ds@willden.org> on Friday February 13, 2004 @06:39PM (#8274971) Journal

            X is a *bad* example, easily attacked, but your argument is fundamentally wrong. The goal with production servers is stability and security, and the Debian stable approach is the hands-down best way to achieve that (though simply running stable isn't enough, it's a good start).

            The only valid reason to upgrade production software is if you must have the features available in the new version, or if you can no longer get support for the old version. Outside of that, you keep what works. Debian stable is very well-supported and it's never more than two or three years behind the cutting edge. It's really ideal, with the caveat mentioned up a few posts that if you need to run closed source commercial software, you will get much better support on Red Hat and SuSE.

    • Can the same be said for Debian, whose stable release is starting to get a little crusty?

      There is a lot of popular support for Debian. Also, there are people out there who apprciate the crustiness as a tradeoff for the higher chance of it working well. Further, Debian has one of the more flexible installation programs out there (/cdrom and /target can be NFS mounts during an install, for example, which is _extremely_ useful).
    • At some point, your customer is likely to need or want commercial software (esp RDBMS), and they will find that Debian is simply not supported by any commercial software vendor.

      What about MySQL [mysql.com], or do you not consider them to be "comercial" because their support offerings are optional and you can use their product for free if you choose to? MySQL offers comercial support to any and all who use their product, regardless of the platform you've installed it on (even Windows!), but of course you'll have to ge
      • How about database clustering?

        Or remote replication?

        Or security and access roles compatible with HIPPA regulations?

        Or reliable data recovery compatible with large-scale backup systems like Legato or TSM or Veritas?

        MySQL cannot do these things. A RDBMS like DB2 or Oracle can deliver an incredible amount of value for certain applications, particularly things like high-volume transaction processing.

        You need to use the right tool for the right job. MySQL is the right choice for alot of database application
        • How long has it been since you looked at MySQL's features?

          How about database clustering?

          I'm not sure about that one - but it seems like either regular replication or 2-way replication oughtta get those needs taken care of. Or, read This article [mysql.com]

          Or remote replication?

          Replication in MySQL [mysql.com] is easy - I'm running a couple of replicas right now for backup purposes

          Or security and access roles compatible with HIPPA regulations?

          MySQL's access control can grant and deny access down to the column level.
          • Sorry but the question poster is talking about a VERY high volume OLTP operation, no one in their right mind would use MySQL in such an environment. There are only two choices for such an environment, Oracle and DB2, anything else is just being stupid. IBM supports DB2 on a pretty wide range of linux distro's but Debian isn't one of them. For a large enough client they might sell a support contract for a non-listed environment (since some of those systems listed are fairly oddball I assume that's how they g
            • I agree that in a critical environment MySQL probably wouldn't be the first choice. However, it's not because of any of the reasons given by the person I replied to. :) Then again, I can't think of any other good reasons. Have you got one (other than "it's free, and therefore must not be as good as something expensive")? ;) The write speed isn't as good as some other DBs (MySQL's better at reading than most DBs), but I'm pretty confident that it'd be acceptable - esp since it'd presumably be behind some

  • Sun's N1 (Score:3, Informative)

    by stonebeat.org ( 562495 ) on Friday February 13, 2004 @01:14AM (#8266786) Homepage
    Check out Sun Microsystem's N1 Grid [sun.com].
    I hear Sun offering suport for Linux as well, these days

    • Re:Sun's N1 (Score:3, Informative)

      by afabbro ( 33948 )
      N1 does not support Linux right now.

      In fact, N1 does not support normal Solaris boxes right now.

      It only supports Sun's blades. All of these other things will be supported in the bright, shining future, of course.

  • WTF? (Score:5, Insightful)

    by smoon ( 16873 ) on Friday February 13, 2004 @04:36AM (#8267638) Homepage
    Part 1: Rant (stay tuned for part 2)
    Why debian? Don't get me wrong -- debian is great for a lot of things, but ... sheesh. Are you making a political statement or trying to support an enterprise need?

    Focus on the tools you need _first_ and the OS they run on second. Getting a great OS with no tools is a lousy place to be, especially after a few months when the client has refused to pay some bills because things aren't working and you have to explain at a meeting between their CEO/CFO/VP of whatever and your CEO/VP/whatever exactly why not, and that to fix it they need to invest $X more in some other platform along with $Y to migrate.

    Part 2: Some ideas
    The TP monitor (e.g. CICS) is frequently done now in a database, so use begin trans, commit trans or rollback trans, and you've got transactions. At least until your database or number of users gets too big. Postgres is a good open-source database that has commercial support options and supports transactions. There are several others, sapdb I think is one. Not sure if mysql supports transactions or not. This is an area where a commercial app (DB2, Sybase, Oracle) may be a worthwile investment, especially if you get into clustering or HA hardware setups.

    Many people use the J2EE framework. In open source that pretty much means jboss. Runs great on linux and you get to deploy lots of apache servers and use buzzwords like 'entity bean' and 'xml'.

    What in the h*ll do you need to do schedule-wise that can't be done in anacron and some simple shell-scripting? There is a reason there aren't really any open source schedulers: cron and anacron are ubiquitous and do what they do extrememly well.
    • Many people use the J2EE framework. In open source that pretty much means jboss. Runs great on linux and you get to deploy lots of apache servers and use buzzwords like 'entity bean' and 'xml'.

      Yes, but, everyone, please go get some real training before going all-out on J2EE. J2EE is really great, but the developers need some perspective of its complexity before it can be used effectively.

    • Re:WTF? (Score:2, Informative)

      by jbplou ( 732414 )
      MySQL supports transactons on table types of InnoDB and BDB.
    • Not sure whether mysql supports transactions or not. MySQL supports transactons on table types of InnoDB and BDB.

      I know someone else already set this straight in this thread but I suppose you can't stress this simple and important fact more than enough! Moderate this as redundant, I can afford to loose some Karma over a database that has constantly kept bread on my table this and the entire last year. Why looks like next month I'm going to get another project migrating from Informix to MySQL :-)

  • Sad reality is multiplatformic (Score:3, Insightful)

    by korpiq ( 8532 ) <-,&korpiq,iki,fi> on Friday February 13, 2004 @05:27AM (#8267762) Homepage
    IT is infrastructure for companies. It is not (arguably ever) their essential business. Their essential business is a combination of management, sales, accounting, production and assorted support procedures for the aforementioned as well as for their clients. Each stream of the gigantic process that is the company has specific tool requirements. I can not see a sustainable model wherein an IT department (outsourced or not) could keep up to its clients' necessities from any monocultural standpoint. I'd rather prepare to separate each task to a platform that best supports it, then define these separations and make my staff follow the definitions, of course redefining where need arises.

    Now as for small companies that cannot afford a multitude of servers, I'd offer a single solution that best fulfills most of their needs and that I would like to keep up and polished for them myself... A debian.

  • UserLinux - "Debian Enterprise" (Score:5, Informative)

    by Humble Legend ( 254888 ) on Friday February 13, 2004 @06:56AM (#8268037) Homepage
    Bruce Perens is leading the UserLinux project with the specific goal of creating a global "co-operative" if you like, of Debian GNU/Linux service provider companies (and consultants), with a completely Debian (ie. 100% Free Software) core. Additionally, the GNOME desktop has been standardized upon specifically to simplify custom/ proprietary development on top of UserLinux. See the homepage [userlinux.com]. The only thing might be the release-ready (version 1.0) timing; depends on your deployment timetable of course.

  • Possible Solution (Score:2, Informative)

    by jmorey ( 38458 )
    Talk to the guys at PerformanceIT.com. I haven't kept track of their recent developments but it seems to me they could satisfy most if not all of your requirements.

    Note: I'm not fully unbiased. I have a friend that works there.

  • TP monitoring? (Score:3, Funny)

    by Anonymous Coward on Friday February 13, 2004 @08:29AM (#8268275)
    TP monitoring? I usually just keep an extra roll on the back of the toilet.
  • Hi,

    Since Debian is not supported and Redhat Linux probably is, why have you decided to standardize on Debian? It beats me!

    Regards,
    Seun Osewa
    • Maybe because there's no legitimate reason to support Redhat Enterprise Linux over any other distribution appart from kickbacks to some marketing fuck?

      Linux is Linux, and if some company fucks their distro up so much that a vendor hast to bend over backwards to support it, that distro should NOT be supported.

      There is NO reason why what works on one distro wouldn't work on another. No LEGITIMATE reason.

  • My .02 (Score:5, Interesting)

    by abrotman ( 323016 ) on Friday February 13, 2004 @10:20AM (#8268848)
    Debian is a great choice. Everyone complains about the old software, but the advantage for enterprise is the long release cycles. This is exactly what RH is shooting for with RH AS/ES. At my company we have a mix of RH and Debian. The Debian servers don't crash for no apparent reason(seems to be a kernel oops related to swapping, dumb on a www server with way more ram than it uses).

    Using debian for commercial stuff isn't as easy as it should be. Many companies don't support debian and seemingly have no desire to. One of those is Oracle. Oracle can be installed on Debian, and there are tons of docs out there to do it. In the end, if you really want debian, stick to your guns until you run out of bullets.

    Another option, tell these companies you're gonna buy X dollars of thier stuff, but only if they make it run on debian on your hardware. If you're willing to spend a few million on Oracle, I'm sure oracle will make it work. The same goes for IBM. I know IBM does tons of software sales, and says they support RH/Suse, but theres no reason they can't make it run on debian.

    I really dont understand why companies dont support debian(please no RPM vs. DEB), in spite of the long release cycles. IMO, it makes it the perfect candidate. Once something is released as stable, it will generally stay that release for at least a year, sometimes two. Oracle could begin to certify Debian/sarge now, and when its released they would have a deployment platform for quite a while. Hell .. They could even setup thier own apt repository with Oracle specific software/patches.
    • I would agree that long cycles is probably better for enterprise customers. Having worked with HP-UX systems before, I know first hand that Debian is downright snappy compared to HP "Official" packages. I my memory is correct, the version of Perl that installs with HP-UX 11i is Perl 4.036.
    • Considering that only RH AS/ES and United Linux are listed as supported Oracle Linux platforms I REALLY don't think it would be a good idea to try to run a production system on Debian. Yes they will still support the DB on other flavors of Linux but if you run something other than the supported platforms then Oracle has an automatic out for any problem that they can reasonably point at the OS for.

  • scheduling tools (Score:5, Informative)

    by cavehobbit ( 652751 ) on Friday February 13, 2004 @12:34PM (#8270270)
    Scheduling is a lot more complex that I have seen anyone give credit for so far.
    Add in : multiple platforms, vendors in different countries with different time zones, holidays, cultures and calendars, (try matching our Julian or Gregorian calendar to a Lunar calendar), with varying schedules, such as cyclical/hourly, daily, weekly, monthly, quarterly, annual. Matching them to accounting needs like a fiscal year that does not match the calendar year, production schedules that cross week and month boundaries, and online systems that are up or down different days of the week and things can get very crazy very fast.

    I can point you to 2 companies that seem to offer the most complete solutions for multiple platform shops, other than IBM's Tivoli, which works fairly well from what I have heard:

    First:
    Cybermation in Canada makes a product called ESP:
    http://www.cybermation.com/solutions/jobsche duling

    They have a cute little site at:
    http://www.replaceyourjobscheduler.com/sitele t.htm l .

    I was at a very large company when they swapped over from the CA7 tool to this one on their MVS systems. I was impressed with the product and the company's support. That was 8 years ago or so, so I cannot vouch for the product or the company now, but I have heard only good things about them currently.

    Second:
    BMC markets a product called Control-M, with all kinds of modules including an Enterprise Manager: http://www.bmc.com/products/productlist/0,2831,190 52_19429,00.html

    I currently use this product in an MVS/Unix/WinNT/Oracle/SAP environment. It does work. It has it's issues and shortfalls, and we have some problems with support, but we have managed to complete our schedule across all platforms every day, with only a very few exceptions in the past 3 years I have worked with it. We run in excess of 10,000 batch/background processes per day across many platforms.

    In all 3 cases, Tivoli, Cybermation, BMC, licensing can be a bit pricey. But if you research the products closely, and only license what you really need as opposed to what you think you need, you can get by.

    I also strongly suggest you hire an experienced scheduler to help out. This is a very undervalued and complcated specialty. Like programming, many can muddle through but few are truely good at it.

    Tom

  • Talk to IBM sales (Score:3, Insightful)

    by metamatic ( 202216 ) on Friday February 13, 2004 @03:20PM (#8272365) Homepage Journal
    As far as IBM is concerned, money talks. If you are planning a 100,000 seat Debian deployment and want IBM tools, contact IBM sales and tell them. If enough people ask, maybe Debian will become a supported platform.

    Plenty of us inside IBM would like to see some free Linux distributions supported, but the company makes its decisions based on commercial pressures, not ideology, and right now not enough people want to run their enterprise on Debian, Gentoo or any other free (beer) Linux.

  • We use IBM & Debian (Score:4, Informative)

    by Howard Beale ( 92386 ) on Friday February 13, 2004 @06:12PM (#8274685)
    And even though Debian is not an officially supported distribution, IBM *has* supported us. Mind you, we're not a real big shop (xSeries 330 with a 1 TB EXP300 array), but they've done a fantastic job and won our support.

Slashdot Top Deals

Get hold of portable property. -- Charles Dickens, "Great Expectations"

Close