Very slow booting and running and braindead OS's 4533

Trust me, they do. Not in some ideal world where all your HW is the same and you can run the same OS on all your boxes and they all run the same set of aplications and there are no complex interactions between machines. But in the real world, well.

Let's say you have a few 10s of blade chbuttis bought over a year and a half: are they all the same? Like hell they are. Some have a single disk, some have two, some have a RAID controller which half works, but only in an old version of the OS, some have IDE, some have SCSI, some have one CPU, some have two. And they all come with different firmware, which you have to upgrade, except for some of the chbuttis you can't. And then there are the switches in the chbuttis, which somehow are never quite all the same, and have their own horrors of firmware. On this you have to run a bunch of applications, most of of which have their own awkward requirements for OS versions. And, oh yes, they have to talk to each other, so you're about to find out a whole bunch of stuff you probably didn't want to know about network bisection bandwidth, firewalls and worrying about who can see what data and how badly you're failing to comply with various regulations.

And those are the *easy* cases. The bad ones are the random crap that the blades were meant to replace, but somehow didn't because there are node-locked license, applications which only run on really left-field versions of the system, etc etc. And over there in the corner there's a bunch of bigger machines: all from the same vendor but varying from 6 year old 2 CPU box to a 48-core machine and with 3 or 4 different versions of their OS on. Most of these can't be taken down to upgrade them, and if they could the licensing for the application means they can't. Oh, except there are some awkward biggish x86 boxes as well, running a different OS of course.

And then there are the development machines, but we can ignore them as they're not critical (except, when some of the important ones break, 200 developers can't do anything, so may be they are critical after all).

Now of course, you can write installation & rebuild systems which will handle all this, but don't tell me it's easy: it's not. It's very hard.

--tim