I was running stock market matching systems with the smallest possible fault-tolerant multi-machine cluster; a pair of AlphaServers with a shared SCSI bus for the quorum disk. One such cluster in Dublin had an uptime of 1114 days that was only shutdown to move the whole datacenter. Most of the clusters I ran worldwide had uptimes over two years, usually brought down by datacenter outages.
At one point I ran a cluster that had machines in datacenters in New York City, New Jersey and Boston. To the end users it was just one big virtual machine. They didn't know, or need to care, if I had to bring one machine, or one datacenter, of the group down for an outage, patch or hardware upgrade. It could be done transparently.
When a machine was replaced with a newer, faster, model, all I had to do was
shutdown the old machine
pull the system disk
put the system disk in the new machine
boot {some caveats apply}
At home I have a cluster that goes from one DS10L AlphaServer upto five alphas, and two vaxen, depending on what project I'm working on at the time. The main cluster foundation node still delivers the mail, serves the webpages, clears the credit cards, and does a myriad of other things regardless of the nodes I bring in, or take out.
Yeah, I'm happy. I spend my time developing, not managing.
Slashdot Mirror
I'm tempted to buy a VAXstation from Ebay.
Do yourself a favor and buy an Alpha instead. Better speed and easier to find peripherials.
You can still get free licenses for hobbyist usage at http://www.openvmshobbyist.org/.
I was running stock market matching systems with the smallest possible fault-tolerant multi-machine cluster; a pair of AlphaServers with a shared SCSI bus for the quorum disk. One such cluster in Dublin had an uptime of 1114 days that was only shutdown to move the whole datacenter. Most of the clusters I ran worldwide had uptimes over two years, usually brought down by datacenter outages.
At one point I ran a cluster that had machines in datacenters in New York City, New Jersey and Boston. To the end users it was just one big virtual machine. They didn't know, or need to care, if I had to bring one machine, or one datacenter, of the group down for an outage, patch or hardware upgrade. It could be done transparently.
When a machine was replaced with a newer, faster, model, all I had to do was
At home I have a cluster that goes from one DS10L AlphaServer upto five alphas, and two vaxen, depending on what project I'm working on at the time. The main cluster foundation node still delivers the mail, serves the webpages, clears the credit cards, and does a myriad of other things regardless of the nodes I bring in, or take out.
Yeah, I'm happy. I spend my time developing, not managing.