IBM 610 workstation computer 3460

I've got boxes with over 480 days of uptime on old RedHat linux. The apps have been upgraded. Software maintenance rarely requires a reboot, sometimes services (apps) need a restart following maintenance or a problem.

The services have a load balancer between the end users (Internet) and the server. The load balancer moves the connections to the working machine in the event a service gets turned down.

With a bunch of web servers, mail servers, database servers all load balanced it's kind of a Virtual system made up kind of like the Vax clusters of the late '80's.

The point is as long as the service continues to be provided to the end user does SMP matter?

It's like the power company taking a feeder line out of service for transformer maintenance. Sometimes it's completely transparent. Sometimes there's an occasional glitch or flicker.

Kind of like a timeout on a web server fixed by a browser refresh.

Well with 5+ years on most of my hardware w-o replacement or maintenance it's different than it used to be.

Here's one production work box for example that's over 5 years old. It's standard rack-mount Intel stuff from a large vendor of desktop and server hardware. One who is really fond of the Intel Inside logo... 8-)

# uptime 10:25pm up 359 days, 5:29, 3 users, load average: 0.02, 0.01, 0.00

Hell, my FreeBSD box that's not on a ups does a good six months at a clip -- the downtime was for a power failure. The box isn't on a UPS. My Linux firewall is along with my network switches, cable modem etc.

# uptime 10:27PM up 63 days, 9:36, 1 user, load averages: 0.01, 0.01, 0.00

These home and work uptimes are pretty non-impressive in the Unix world. The BSD box is about 6 or 7 years old -- an AMD K6-2-450 with 512 of memory and two 5 1-4 full height antique 9gb 50pin narrow scsi drives.

It was buttembled from the collection of second hand crap I purchase (only brand name quality second hand crap). It's been rock steady for 6 years.

I've had to replace power supplies at times, and video cards on boxes where the video wasn't on the motherboard and sometimes SCSI and IDE drives. The motherboard and memory has been solid on my home cheap box since day one. A pile of dumpster dive rescue boxes with genuine Intel motherboards have been put in service at home for various test uses. They run solid as the new 64 bit stuff we just got for the office.

In the field when a problem happens:

If the drives are in a RAID configuration with hot swap trays it's pull out, push in and continue to function while rebuilding.

The non-raid stuff just goes out of rotation for repair and goes back up when ready.

I can fully format, load and install a web-mail-nfs server remotely from my office or home (the servers are 45 minutes away).

The machines are ethernet bootable (either Intel x86 with PXE and Linux kickstart or Sun Sparc w-jumpstart). I can format, install and turn up a spare box in less than 15 minutes. The auto-install stuff we've got handles all the mixes of different hardware and disk size automatically... The files load over HTTP, FTP or NFS and boom -- instant server.

This isn't rocket science these days and it's easy to do with UNIX type stuff. The RedHat and Solaris installers are all designed for this these days... there's no linking and all the disk sizes and config stuff can be automated based on certain hardware configs (and we key them to the MAC ethernet hardware address). The config stuff is all flat file table driven and can easily be updated when the hardware is upgraded, changed or replaced.

Here's kind of the way it all works:

If it's MAC address 00:04:AC:38:89:73 it's an Intel PIII with 120gb disk capacity and should have disk parbreastioning x and be a web server.

The one 08:01:23:45:DE:AD is the mail server with 4 18gb SCSI's in Raid 5 configuration and 08:23:22:34:55:66 is a database server with mirrored 73gb scsi drives.

The first two are a Linux os from source A and the latter is a later Linux os rev from source B.

So... with a couple of ready on-line boxes one does the work of two while a replacement for the dead box can be provisioned in minutes.

All together the stuff I run works like this in at least 4 countries. All systems administration operations done here...at this point in time.

Local folks do do the hands on parts replacement when we're working outside the local area... but we do all the software, operations and sysadmin and coordinate the "remote hands" physical stuff.

Sometimes my Field Service skills, rusty as they may be, come in handy.

I'm sure this is pretty common in the Internet space.

Bill -- -- digital had it THEN. Don't you wish you could still buy it now! pechter-at-ureachtechnologies.com