Slashdot Mirror
How Hard is it to Manage Different Unices?
vrmlguy asks: "Where I work has several Unix-based servers, all running the same vendor's OS. We are getting ready to buy another big server, and management wants to get bids from other vendors. However, our staff is only familar with our current vendor's OS. Yes, I know that any two flavors of Unix are more alike than not, and yes, I know about the Rosetta Stone for Unix that makes it easy to transfer skills. I want to know about the down-side: What's the difference in the cost of operations between a mono-culture and a shop running two or more vendors' OSs?"
You have a team of mechanics, and for the last 20 years all they have serviced, as well as driven themselves are Ford automobiles. Now, your boss tells them to jump right in and service Chevrolet autos too. How easy will this change be? Depends on the mechanics and how they've been trained I suppose.
What's the difference in the cost of operations between a mono-culture and a shop running two or more vendors' OSs?
$32,593.12
Now can we stop with these stupid, inane questions? I would rather read Jon Katz than these awful Ask Slashdot questions of the past 3 months or so.
Not a flippant post.
The quality of Unix sysadmins has declined so much over the past decade that what passes for a sysadmin right now is what I used to call "an operator".
We have 5 unix sysadmins (major transportation company). Not one of them could write a shell script if their life depended on it.
They insist on doing everything by hand and then complain there are no automated tools to them. Their definition of an automated tool really means "graphical front end to those grubby text commands".
They have no appreciation for the modularity of unix, and they look longingly at Windows servers.
Meanwhile, they're all getting paid twice what they're worth because apparently as dumb as the Unix sysadmins are, the NT ones are apparently on a different evolutionary scale where "rock" is considered the most intelligent life form.
So my point is that getting these sysadmins to switch won't happen. They'll piss, bitch and moan about the opportunity to learn something to enhance their skills, then complain the application is screwing up "their" servers.
If only ASPs would take off, my life would be much better, because sysadmin skills suck so bad, black holes pale in comparision to the event horizon of these so-called admins.
The same principle applies to natural and computer languages - the more you know, the better you understand the fundamentals.
:)
Sure, you might know how to do x,y,and z on your Solaris box, but once you understand how to do it also on RedHat and AIX, you'll understand much better how it works conceptually. Then when you get an HP box, it'll be pretty easy.
Of course, don't run killall on HP.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Generally, this is not difficult to do, as long as your admins understand the bases of UNIX. (Vendor-centric admins sometimes don't, as they get dependent on their vendor's tools.)
The problems can arise with:
1. vendor-centric admins who aren't willing to learn
2. different service contracts creating differing expectations of uptime between systems
3. added costs from maintaining multiple service contracts and training on multiple platforms
4. finger-pointing, if the systems interact
5. rewriting in-house tools which are needed on the new platform, but were not written generically before
6. 3rd party licensing costs may differ (if you are licensing the same product on both OSs)
7. dilution of expertise, since your admins will have to be more generalists (this is often overbalanced by the expansion of perspective in problem-solving that comes from broader experience)
Other than that, I can't think of anything off the top of my head which would make this hard. Generally, it is not a problem to do.
-- Two men say they're Jesus. One of them must be wrong. - Dire Straits
What's the difference in the cost of operations between a mono-culture and a shop running two or more vendors' OSs?
How much of a raise are you asking for?
-... ---
Statement of Bias: I "administer" several UNIX OS versions (Solaris, IRIX, Linux, occasional HP-UX), but in an isolated network with no outside connections (so very little emphasis on security).
Two factors come to mind:
No matter how close the systems are, you will still "loose" time to training (either formal or OJT) requirements for the new system. This may actually be a benefit for your staff (wider perspective, more to put on Resume).
Depending on how much focus is placed on security, you may end up doubling the time required to track vulnerabilities and install patches. Again, this may be an advantage as well since a single-os shop tends to have equal vulnerabilities on all systems. In a multi-os shop an attacker will have to work harder to get control of everything.
It is easist to manage servers from only one vender. Unix makes ti easy to transfer skills, but here is the contradiction: It is easier to manage servers from many different venders and versions, than to manage just one server that is different from the rest.
When you have all OSes the same it is easy because everything automaticly transfers. When all are different it is harder because you always have to remember the correct incarnation of each procedure, but because they are all different you get in the habbit of looking it up each time. When all are the same except for one machine you forget on that one machine that everything is different and you aply the wrong incarnation (ofte with disasterious results, see discussions of killall linux vs hpux on comp.risks) Because of this, the one different machine will get [invalid] complaints often due to these differences.
If you can't stick with one vender, then you should go with many so you are in the habbit of checking the differences. At the very least get some linux (debian, redhat, suse), and bsd (free, open, net) machines in house now, and use them for production. You need to make sure that your admins are used to subtile differences. The other alternative is to just stick with one vender, but not only do you pay more, but your admins become lower quality as they learn only one system. (think of it is a resume builder, you want different systems on your resume!)
It seems to me that the biggest cost is in sysadmin time. I figure it this way, at work I use a few UNIX systems. We have one machine running IRIX, a couple running BSD and one running Linux. Now, when I write a script one one of the BSD machines, it works on all of them, but it may or may not work on the Linux machine, and certainly won't work on the System V-esque IRIX machine.
Now, if your sysadmins employ a lot of scripts, figure you'll have to spend twice the time maintaining them if you have two different platforms that are not fully compatible. You can minimize this if you stick to POSIXly correct scripting, but you'll never completely eliminate it.
The same goes for custom programming. For instance, if you're running everything on BSD, and you want to take on a Sun machine running solaris, there may be some issues with the occasional socket call that Sun implments differently from the rest of the world.
So, the more custom scripting/custom apps you have, the more time your sysadmins will have to spend maintaining/porting/testing the stuff.
---
Play Six Pack Man. I
Money.
You obviously already know that managing support contracts from multiple vendors is going to suck. I would also recommend taking a long hard look at ongoing support charges.
For example, we have both HP/UX and Sun platforms where I work. We have both servers and workstations. For the workstation support contracts on similarly sized machines, there was a world of difference in cost.
The annual fee for an HP C240 workstation was somewhere between $2500 and $3000. The same annual charge for a Sun Ultra of equal speed, was between $1000 and $1500. Multiply that by the number of workstaions you have to maintain, and it can add up very quickly.
The up front cost typically isn't where they get you. It's on the back end. I would research the back end on all of the platforms you are considering very carefully before making any final decisions.
Hope that helps a little.
For example, on Solaris (without Veritas Volume Manager), you have to "carve out" your disk filesystem by filesystem, and work with devices in /dev/dsk/cAtBdCsD format. On AIX, the concept is totally different with Logical Volume Manager, wherein filesystems can be created on the fly. But HP-UX uses both in an odd fashion, forsaking slices and using a "castrated" form of LVM. This is just one example, as you will find other things in HP-UX such as the useradd command being identical to Linux and Solaris, and the SAM tool being very close to AIX's SMIT utility.
In the end, as you will find, there is no uber-Unix that will carry over to all of the other flavors. IMHO, HP-UX is as close as you will get. But, my personal preference of all Unices is AIX due to its ease of use (an IBM tool easy to use? I know it sounds like an oxymoron) and robust capabilities, combined with Linux integration in the most recent versions. Flame as you will, I'm interested in hearing anybody else's insight.
--Chag
It actually says "Requires IE 5 or better". So, I installed Mozilla.
"What kind of unixes do you run?"
"Oh, we have both kinds. RedHat and Debian.
Yeah, basic management is similar. HOWEVER: maintaining multi flavors of unix is quite expensive in terms of admin time and effort.
Think patches. Now you have to track multiple vendor advisories, handle patch management, what does the patch break, depend on, etc.
Next comes config changes. No longer can you write a simple script that makes the change on all boxes, you have to support multiple scripts, or write scripts to handle the ideosyncracies.
Third comes binary compatability. What I generally do is build a local set of binaries for all the specialized stuff. They can either be blasted across all systems or mounted via NFS. With multiple versions / flavors of the OS, work gets doubled, tripled, etc. What used to be a 2 hour task turns into a day long task.
Can't forget about security. What you do for one system you have to do differently for another. Different tools, binaries, rc scripts, etc.
The bottom line is that if you needed 4 people to support your current single OS environment, you may need 5 or 6 or even more when you go multi-platform.
I ran a shop where we supported Win98, NT4, Win2000, Solaris 7 & 8, RedHat Linux, FreeBSD, Macos9, MacOSX, and a smattering of other OS's for 500 users. This get's non-trivial Very fast.
1) You can install the same shell on just about all UNIX's. Most people where I am prefer tcsh as it has some nice features.
2) You can standardize on scripts, either use csh (blah) or sh. We prefer sh as it is found on just about EVERY unix (Sun, HP, AIX, BSD's, Linux).
3) Avoid vender extensions to the basic shell. HP has done some aweful things there in its bourne shell and they are not compatible with Sun and in some cases Linux either. I.E. Always use `cat foo` and not $(cat foo) in sh scripts. There are other things like that.
There are problems in supporting more than one UNIX, but there are also workarounds if you do it right.
Only 'flamers' flame!
Knowing that (as an example) AIX has a pretty self tuning kernel, that Solaris has a modular kernel, and that UnixWare needs a recompile (relink) for any minor changes forces the admin to think about the operating system instead of just drooling on the keyboard.
The biggest differences are still SysV vs BSD. Understanding those is vital in a mixed OS environment. Beyond that, there are usually differences in disk layout (and filesystems), but they just add to the rich diversity that is my favourite OS.
At my work, we are big users of Solaris, but because we develop software for multiple platforms, I've also had exposure to AIX, UnixWare, Sequent Dynix/PTS, HP-UX and DRS/NX. These days we've dropped Dynix/PTS (EOL anyway I think), HP-UX (too expensive for our customers), and DRS/NX (dead?) but we're looking to port to OpenUnix 8 and Red Hat Linux, so things are still pretty mixed. I just think it's a shame that I don't get to work with HP-UX and that Unixware is dying (yes - I like it!).
We also port to NT/2000, which is good to compare - it's a nightmare to work with when used to UNIX.