winscape 2295

Dont think less capable. The P4 Xeon SMP boxes I'm putting in (3ghz 4gb memory dual cpu) could out run all the mainframes of your clbutt. the local disk (mail server-database server-web server etc) is running on striped and mirrored disks ( think of a bunch of RP06's with interleaved structures and those are mirrored to another set).

The drives are hot-swap... if one fails plug in another one on the fly and the system will run and automatically rebuild the data from the error correction code stored on the remaining drives. Degraded operation is a bit slower -- but it takes less than 1-2 a day to rebuild one of these drives... while the machine still serves data for a couple of hundred thousand users.

Well, consider mail services provided by a cluster of a half dozen machines. They sit behind a network load balancer box that pbuttes the mail SMTP packets to them from the actual internet router or switch in front of them.

These machines appear to be one big honking box on the net with one ip address and seemingly unlimited bandwidth.

Actually, there's a half dozen boxes and you get your mail connection in either a round robin fashion or some other mode.

This means if any one box goes down the only thing that happens is it can't take any more mail in and the stuff on the disk that hasn't been flushed from the queue to back end mailbox storage will get flushed on the next boot after the machine is rebooted or repaired.

One day I had to run up to the colocation site after one of my boxes blew the motherboard.

I pulled it out and slapped another box in it's place and moved the disks from the mail server to the backup.

Bang -- reboot, file system check, mail flush. The downtime was just a delay in the mail from the mail server to the user mailbox. Nothing lost.

The back end file storage is either a network attached storage device in some kind of cluster (kind of like an SMP box mix with disks dual or more attached.) If there's any failure the disks fail over to the backup server-servers and the data continues to be stored.

Nah... we just keep all the stuff spinning on terabytes of redundant spinning platters. Tape is for full disaster recovery.

We migrated an entire data center by restoring a set of backups on the new hardware and than rsync'ing the changed and updated data from site1 to site2 over a VPN on the net.

Then... kill site1 and site2 replaces it.

Bill (who just realized how much of this i take for granted these days and that I first saw most of this capability on Vax clusters in '86... )

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