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

 



Forgot your password?
typodupeerror
×
Linux Software

Replacing Novell with Linux? 24

RatzMilk asks: "The firm I work for has several clients who are still running Novell 4.11 which Novell is stopping support for in early 2001. As such they will require expensive upgrades to the latest version to be able to continue their support with Novell. I have suggested to one of the clients that they could replace their Novell with Linux which would have the added advantage of being able to utilize their existing hardware. The Novell upgrade would also require a hardware upgrade has well. They have agreed but their only concern was stability as the Novell boxen has been very reliable and Linux is an unknown to them. My company's policy is to provide the best possible support to our clients. My question to Slashdotters is, do you think I have made the right recommendation? Should I just do the switch with one client and see how it goes, or go the hole hog and recommend it to all?" Would Linux be the ideal solution in this case? Would BSD be any better or worse in this situation? And is holding on to the outdated hardware a good idea? As it is, systems can't go on forever, and depending on age of this system, maybe a full upgrade is a good idea.
This discussion has been archived. No new comments can be posted.

Replacing Novell with Linux?

Comments Filter:
  • by Anonymous Coward
    The consensus around my chapter of the ACM is that Linux makes a sufficient desktop OS, but for server boxen, you want a flavour of BSD. As for upgrading the h/w now, why? Wring every last minute of uptime from the old hardware and then take the same amount of money you'd spend on a Pentium4 now and wait to get an Itanium for the same price. Or hell, an SMP G4/G5. But there's no use in decommissioning perfectly good hardware.
  • amen I've got a P75 sitting beside me right now that I still use as a server for various things, I'm soon going to punch in a P166MMX chip to speed things up a bit but I don't REALLY need it. P.S. Anyone interested in a partialy fried P90 chip that can only run underclocked to 75mhz?
  • As a rule the first thing when talking server class systems is do you want a desktop class or workstation class.

    The usual arugument is as folllows:
    You can get a Sun workstation for about $50k
    You can get a Dell runing NT in the 10k.

    The Nt on a dell averages about 40hours of downtime a year.
    The Sun, while less bang for the buck, averages about 2 hours of downtime a year.

    Linux/BSD on a PC falls into a similar boat, you it's difficult to beat the quality of a mainstream OS on industrual hardware. Though the downtime of a Linux/BSD box is primarly based on poor hardware support, drivers written by hackers in their spare time. BSD is by far more stable and networthy though.

    I'd rather have either one then SCO

    But the fine choice always is, what OS will run the applications you want, how much do you have to budget to the net information system. Which OS can be supported.

    There is no shortage of Linux geeks, BSD geeks are harder to find. SunOS/SCI/HP/SCO/DEC a touch harder.

    But it all depends on your application, and how much keeping the system online is worth to you.

  • If the hardware works... leave it. If the hardware has lasted as long as you seem to suggest... it's liable to last years longer. Replacement power supplies are cheap, so are CPU fans. Otherwise get a HD mirror or better and leave the rest alone. Server hardware should be upgraded when it gets too slow or breaks, not when the next OS upgrade comes out...

    Where I work I have a couple each of Novell 4.11 and Linux boxen... don't have problems with either. I'm upgrading the Novell servers to 5.1 *only* because Novell offers great deals to non-profits... it's a good excuse to get some practical experience for a CNE. :-)

    Bottom line, Novell has amazingly good software but if it weren't for the great deal I probably wouldn't be upgrading the Novell boxen. Linux/BSD offer impressive remote admin abilities, internet cache, mail server, DNS etc etc on the cheap... a 30 license Novell upgrade will cost ~$3,000 plus the Groupwise update plus ZEN (if you like) plus Bordermanager plus new hardware.

    -sid

  • Hardware won't last for ever. And usually once it eventually fails, sourcing replacement parts is costly.

    A former customer of mine was very reluctant to upgrade the hardware of one server. The disk hosting the root filesystem was not large enough for the system to be upgraded to the current version, or even patched to the maintenance release.

    Eventually the other drive failed. I had to spend 6 hours with a filesystem debugger fixing the filesystem enough to allow a backup to be taken. Then when I came to reboot the box the machine failed to boot altogether. As this was on a Saturday, and off the mainland, getting an engineer to look at it was impossible. And the existing drives were EDSI, the OS the Microchannel, and we couldn't move the drives to another machine, as we could locate anything suitable nearby.

    It cost us a lot of money to replace the machine, and a week's work implementing the change. All because the hardware was now obsolete, and we were not able to get a valid backup before it failed. You'll probably have the same problem.

  • If what you are looking to do is to replace NetWare with Linux running NetWare emulation, such as marsnwe, you might want to take into account that it "looks like" a NetWare 3.x server to the clients. Since you mention you have NetWare 4.1, this would actually be a "downgrade" for you. Now, migrating to Linux is a different thing. If you only use NetWare for storage and printing, then that shouldn't be too painful to replace it with Linux. You'd still have to use Samba to allow any Windows clients to access it as an NT server. Depending on the size of your organization, switching your clients over could be a long, drawn-out process. There may be an easier way to do this, with or without Samba, marsnwe, etc., but not that I'm aware of.
  • Firstly, if it ain't broke, don't fix it.
    Secondly, if you can afford it, going to NetWare 5.1 is probably your best bet.

    But in the meantime, here's some stuff to try.

    You can run Novell clients and servers on Linux. Caldera OpenLinux 2.2 and up includes a Netware client which is fully NDS-aware, and you can integrate a Linux server into the NDS network so that, say, it uses Novell user authentication to get at resources shared with Samba.

    There is also Netware for Linux. This runs a real live Netware 4.x server on Linux, but doesn't support NLMs, Netware filesystems etc. A free demo is available from www.caldera.com but you do need to buy licenses. Also, it only runs on kernel 2.0.35. However, you can probably pick up a copy of OpenLinux 1.3 very cheaply now!

    There are some open-source Netware servers such as MarsNWE but they aren't NDS aware.

    So add a Linux box now. It's cheap and you will learn useful stuff from getting it integrated with Netware.

    But it's no replacement. Yet, maybe, but...
  • if propery cooled motherboards cpus and memery
    will last allot longer than most peaple think.
    my take on it...
    replace the fans and the harddrives and install linux.
    Its probably time to give them a good gunk removal also.
    Some of our clients are still running 386 novell setups.
    -mostly as a secondary system because the cost of converting the data over to a modern format is $$$$-
  • As long as the machine's not chock full of nonstandard bits, like that PS/2, and is regularly backed up (like you should be doing if the data's important) what's the harm in allowing the customer to use their system a bit longer? Granted, if it's a 386, it might not be worth keeping, but 486s and low end Pentiums are still more than adequate for a lot of jobs.

    The original post never said exactly how badly the system would need to be upgraded. For all we know, it could be a PPro machine with only 64MB of RAM. If you have good hardware, it should have no problem running for 5 or more years. Besides, isn't one of Linux's selling points that it can extend the usefull life of 'obsolete' hardware?

    Not everything needs the newest, fastest, whiz-bang space heater from Intel to work properly. Back in HS, we had a Xenix system that would comfortably run 16+ users (vi + compiler + rogue) on a 386 w/ just 16MB of RAM. Unless this machine's doing heavy database work, if it's got PCI it should be more than enough


    On a related note, I'm in the middle of setting up a Linux network at school that's going to be made up of a bunch of surplussed P133s + P166s (I even managed to get a P200MMX for the server). I've got a pair of 386 running linux, and somehow I'm the dept's Linux expert... Well, here's to hoping that I get all the lab machines w/ a dual boot by next semester =)
  • Novell is dropping support. So what?

    When Novell drops support, you'll then have about the same level of support that Linux has. USENET, third-parties and else. If support is the issue, you're going in the wrong direction by going Linux. (Moderators: that's insightful not flamebait.)

    Novell 4.x has been on the market for many years. It is about as rock solid as an operating system can get. Do you need really need Novell's support?

    Unless you need some new Novell features, I would not upgrade. I've got 3.x servers still running. I know of a Novell server that has been up for nearly four years (no hardware problems, knock on wood) and no one knows the admin password. We've been through two Novell admins since it was installed and since nothing had to be done to it, it just sits there spinning. (This is not a suggested practice.)

    If I did anything at all, it would be to upgrade the hardware only. I would not change operating systems nor would I change versions. Get the box on new hardware while it's still supported. If it runs for three months, it'll run three years. If, in a year or so, you decide Linux (or NT) would be a better choice, you've got good hardware to put it on.

    InitZero

  • How is:

    The consensus around my chapter of the ACM is that Linux makes a sufficient desktop OS, but for server boxen, you want a flavour of BSD. As for upgrading the h/w now, why? Wring every last minute of uptime from the old hardware and then take the same amount of money you'd spend on a Pentium4 now and wait to get an Itanium for the same price. Or hell, an SMP G4/G5. But there's no use in decommissioning perfectly good hardware.

    ....flamebait?
  • I thought that was odd too.
  • We knew the data was important, the customer didn't. We inherited the customer from a rival who moved out of this sector, and didn't realise how bad they were until this site visit, not long after we took over the contract. Other sites inherited were equally shambolic. The site wanted the most recent data off the machine, they only backed it up once a week due to it having a slow 1/4" streamer. This machine failed on a Friday, and they wanted all that week's data.

    Our standard policy during new system roll-outs was to ensure that all customers knew how to make backups, and took them daily. We also ensured that they performed full backups, and not incrementals, which took bloody ages to restore (having had to do them myself).

    The account manager for this site got quite a kicking once he received my visit report. No site audit prior to exchanging contracts, no competant permament on site staff (administrators flew in from the mainland when neeeded).

    An I remember those Xenix boxes. Installed a few of them. Even had to guide one buffoon through a reinstall as during fsck he had given for the scratch-file name the device name of the root device. 6 hours on the phone......

  • Also, keep in mind that Novel is really secure. It's secure cause no one knows a blasted thing about it.

    Hackers avoid it for this reason.

    I would not setup a Novel system today, but if one already exists and it's serving the job.

    But without knowing your application needs, it's difficult to make a judgement. For simple low load 20 employee file - print sharing - e-mail - centeralized backup can be accomplished by anything of the above.

  • by maggard ( 5579 ) <michael@michaelmaggard.com> on Friday September 08, 2000 @07:33AM (#795776) Homepage Journal
    It's folks like you that scare so many customers. You want the customer to go to Linux, or BSD, but can't provide a reason aside from "it's hip". So now you go fishing for reasons?

    First, Novell is a company & Netware is a product. Novell makes a bunch of products and although Netware is their flagship being accurate tends to reassure customers.

    Second the client already has Netware. Clearly it works well for them. Clearly it's integrated into their operations. They seem to have support for it, the users are used to it, the files and rights are likely well established and well understood. There are doubtless ancillary systems like backups, print servers, email, etc. that all rely on Netware that would need to be reconfigured or replaced should Netware go away.

    Next support folks for Netware are easy to find and have well established credentials. Support for Linux or other comperable OS's is pretty much a dice-roll these days. Application support is the same - everything understands how to work in a Netware environment - few applications yet know how to optimize their file calls or locking in a Linux/etc. environment. Have a problem with Whizbang2000? Call their support line and say Netware or NT; it's a click in their decision tree. Say something unusual like Linux or BSD and the phonetech will go into a crash & burn, likely after first trying to blame your Linux/etc for the problem.

    Look, I'm not knocking Linux or BSD varients or anything else. I'm just pointing out that there's a hell of a lot of issues involved in changing from one server OS to another. Lots (most) of these issues have very little to do with the quality or qualities of the OS itself but rather with external situations like legacy compatibility, support mechanisms and the basic horrors of reengineering a typically complex network environment.

    To encourage a customer to change their server OS, particularly when it's one that they seem so happy with, is not to be done lightly. To encourage them to move to an OS that you & your company don't know well enough to even recommend a particular choice is sheer folly.

    I can see suggesting Linux or some other non-traditionial OS (damn we need a word that encompasses Linux & the BSDs but leaves out the more obscure varients) for a client with a specific need for something not found in their current Server-OS-of-choice. I can even see encouraging them to use Linux/BSD/etc. in new deployments where there isn't much legacy material to worry about. I'd definately encourage them in cases where they've found thier current Server OS unsuitable or they've outgrown it (all assuming I was well familier with the specific product I was recommending.)

    In your situation? It's a good thing your company is researching the alternatives. It's a better thing your boss will likely take your results, read them with interest, and not mention this to this client. Be cheered though that your studies won't have gone to waste as it wouldn't be suprising if soon some other client either requests Linux/etc. or has a situation where you folks can legitimately recommend Linux/etc.

  • not if Novell is not supporting this version. I ran across this problem with Oracle.. they stopped support of the version 2 days before we had a disk failure/database corruption and they wouldn't even stay on the phone with us.. the company ended up hiring a consultant to come in and fix the problem... from which they never fully recovered...
  • Amen brother!

    I get this type of question from clients more and more these days..."Should we upgrade our system to Linux? Everybody says it's much better."

    That's cool if you have an NT server for file/print, but it's a whole different thing when your business is built on stuff that runs on the current server OS.

    Almost every single time, this "upgrade" has been recommended by the person on the staff who "knows computers" or (more frequently these days) by the previous consultant who I'm cleaning up after. Not once has any of these people conducted a systems audit to see how this would impact the business. All they know is "Novell is dead" or "Microsoft is evil" or "Linux is free." (I agree with 2/3 of these, but it's foolish to base a migration on any of these ideas)

    It's kinda like saying "I'm gonna put a turbine engine in my Camaro, because they're much more powerful, efficient, durable, and I can run it on cooking oil." You can't just drop the new engine in there, Gomer. All of the things that depend on the old piston engine will either not work or freak out. It's a lot better idea to design your car around the engine you plan to use.

    So yeah, on new installations or simple networks moving to Linux may be a good thing. If you're planning on taking away NDS, converting applications and re-training all of the end users I just hope they don't hire me after they fire you.


    -------------------------------
  • I'd like to know where you get 40hr./yr. of downtime for NT... Where are you getting those numbers from? If you're an NT administrator and you have 40 hours of downtime a year, you're doing something wrong. I've got 16 NT servers (On various flavors of Compaq Proliants) and I'm lucky if I've got 4 hours a year for all 16.
  • I'd rather have either one then SCO

    Hey! I've had SCO Boxen (ODT2.0) that have been running for 7 years. Only downtime was

    H/W Upgrade

    Power Supply on one box failed

  • The original question was really two: the first the OS and the second the hardware it's on.

    Frankly most businesses with any technical acumen insist on a three year replacement cycle on things like x86-based servers. Why?

    1. They're paid off.
    2. They're no longer state-of-the-art. Getting replacement parts for them is becoming difficult for equipment long out of stock.
    3. They're becoming 'worn'. Bearings have been spinning for three years. Power supplies have taken three years of small hits. The motherboard & cables are becoming brittle. Metal contacts are slowly getting gunky or working out of position.
    4. The computing environment has moved on. Three years ago a server could be a PII 233 with 128 MB RAM and a 16 GB drivespace with a 10bT NIC. Now the clients are PIII 700's with 128 MB RAM & 24 MB HD' pulling down files over a 100bT wire.
    5. Drivers are no longer being developed or tested for the OS's. Simply to maintain security patches one must update the OS occasionially & the more outdated the drivers are the greater the likelyhood of an incompatibility popping up.
    6. After 3 years folks will no longer be facile with this generation of hardware. Something that we now take for granted was novel and required special consideration then - something we may well forget to take into account now when a problem arises.
    7. Lastly - the chances are getting higher and higher that the manuals have been lost, the floppies holding critical device-drivers have 'rotted', documentation no longer matches the actual configuration, and the reasons for many of the originial implementation decisions are lost.

    True lots of hardware chugs on happily year after year but it's also true that every day that pases after the originial burn-in period the device becomes more likely to fail. While failure of a home computer isn't a big deal (pop in a new power supply, replace the failed fan, whatever) in a business the results can be significent.

    In the case of a file server failure many of the files that were in use during the failure may be truncated or damaged. Changes to them since the last backup may well be lost. Finding and replacing these damaged files can be a mess, popping up problems months or even years later. Employees are inconvienced when the services handled by that device are no longer availiable and the inevitable confusion results. The cost of even a a half hour outtage in a small business can quickly exceed the cost of a new server, in a midsize business it's a few minutes and in any large business any interruption of service is a serious issue.

    The IS department is likewise disrupted with diagnosing the problem, determing the solution then picking up the pieces. Should the support be outside then there's the additionial travel time and a greater likelyhood of unfamiliarity with the device and it's functions leading to yet even more confusion and a longer outtage.

    In short, spending the money on keeping hardware current is money well spent. Few businesses have any problems with understanding this need and will respect a 3-year lifespan on hardware.

  • I agree, it's not necesarily the change thats the big deal, its the starting from scratch. Many installations have been there for such a long time that if you had to format and reinstall from scrach the SAME OS, you could never get it back the same again.
  • Let me give you some background first so you know where I'm coming from.

    I've been using Novell Netware since it ran on an XT (Netware/86- required 640k of RAM abt 1985). I've been a CNE since '93. I've been using Windows on my desktop since about '88 (2.0ish - 286 w/16MB ram). I've been using Linux since about 1992 (not sure which kernel - 0.something). I've used Solaris in an ISP environment since late 94. I've been using FreeBSD since about 95 (2.2.x). I've also set up/administered systems with NT, SCO, DYNIX/PTX (Sequent) and so on. I've also done my stint with low level programming and TCP/IP development, etc. etc. etc.

    I've been consulting and doing independent system/LAN administration for at least 15 years, pushing the bleeding edge well before my clients.

    Today, the best commercial product out there for File/Print is Netware. Period. Yes, it's expensive, but compare the bottom line with NT and you'll find that NetWare is actually much cheaper. I have printouts of servers which have uptimes of over a Year. I get calls over down servers usually only over a hardware failure, as that is about the only time it dies. In short it works and works well.

    On the other side of the commercial arena is NT. I've supported NT. If I was into wanting to being able to bill my clients lots of hours and causing them lots of frustration, I would recommend NT to them. But in reality, I have more work than I can handle, so I don't. I'd rather see a customer put a peer-to-peer network in place than use a NT server. Note I'm talking about File and Print here. About the only place I'd use a NT server is in those applications where it is required for a specific need. And then, I would recommend that the client run that app and only that app on it.

    Now, back to the question at hand.

    Would I pull out a *working* Novell Server and replace it with ANY other OS just for the sake of being "supported" by Novell?

    Nope.

    Would I potentially consider using an alternative system in place of a Novell Server if the server needed to be upgraded?

    Probably.

    I'm going to digress a minute before I come back and explain that:

    My opinion on Linux vs BSD (and FreeBSD in particular, as I haven't dove into the others yet) is that BSD makes a much better SERVER than Linux does, and that Linux makes a much better CLIENT than BSD does. For server applications, I would never consider Linux at this point. I just don't trust it as much as I do FreeBSD. I might be brainwashed by the BSD camp, so please don't flame me for this - but I really feel that the FreeBSD development methodology is better at producing consistent, stable, mission-critical code than the Linux method. I also feel that the Linux method is much better at getting those apps you NEED on the desktop ported.

    That said, I would STRONGLY consider putting in a FreeBSD box running Samba to do file/print. I just haven't found the right client - well that's a little bit of a lie, the ISP I do sysadmin for has a corporate server running FreeBSD and Samba. To make a long story short - this client is more technical savy than most, and as such is better equipped to deal with the FreeBSD box.

    In any case, it sounds like the original poster's client was worried about Support. About the only way you're going to get commercial support, is to upgrade Novell. (Or move to NT, but again I didn't mention that)

    If they really aren't worried about support, I'd say leave the Novell server in place until such time as the server is not doing what it needs to do anymore (i.e. not fast enough, server old enough you're worried about relability, etc.) then look at the upgrade options. Then, if the client is technical savvy enough, consider a switch to a Open Source OS.

    One piece of information which is lacking is the size of the firm this is for. In a larger firm, the value of Netware's NDS becomes much more critical. If you've got a hundred desktops managing them with Netware and NDS and ZenWorks is almost trivial cost-wise compared to any other option. In this larger environment the cost of the server licensing is dwarfed by the PC support costs.

    One final comment - make sure you look at Novell's Netware for Small business. A 5 user upgrade is around $1000, or $200/seat, but it includes not only the Netware but also Netscape Enterprise server, A web cache, NetObjects Fusion (web editor), McAfee Netshield & Virusscan, Tobit Faxware and Groupwise. The added value of Groupwise Alone might pay for the upgrade costs.

  • All this stuff is easy compared to porting custom apps. If you've written custom NLMs that run on your server for a specific business purpose. I could be near imposible to port them to linux. And to go the other way, Linux to NetWare. With custom apps that would probably be many times more difficult.
  • Well there's a big market: not. Custom NLM's never really took off and most of what did get written either migrated years ago to easier platforms or is in hard core Netware shops where it will likely stay for awhile.

    As for Linux/BSD/etc. to Netware I expect most folks would just skip the NLM's entirely and code Java servlets for the Netware server (if they had to do this migration for some reason.) While a servlet wouldn't be as fast as an NLM it'd be easier to work with and has the advantage of being that much more portable in the future. Honestly though I've not heard of anyone porting anything custom from Unix to Netware in years but I suppose someone might do so for a specific need.

    Frankly the market for custom Netware-based applications seems to be pretty minute. If one is going to develop something propriatary for in-house then NT or Linux/BSD/etc. would be the far more obvious choices. With Netware's increased connectivity one could host the app. on a separate box and still validate against the NW/NDS box. Netware is a great OS for what it does but NLM's are not a joy to write or debug. Besides if something is important enough to develop internally then it's likely important enough to get it's own box to run on.

E = MC ** 2 +- 3db

Working...