The mid-sized ISP I used to work at primarily used an OpenVMS (Alpha) cluster to do most of the core functionality like centralized RADIUS authentication, SMTP, POP3, DNS, web hosting, and even shell accounts for tens of thousands of users (though only about 3 users only ever telnetted in). For the first few years, the TCP/IP stack on those servers was a commercial package sold by a 3rd party vendor (like Trumpet Winsock) and I hated that SSH was commonplace by then but we still had to telnet into these VMS servers because they had no SSH daemon.
I migrated a lot of functionality to OpenBSD servers (and eventually a new ops team replaced those with Linux) but RADIUS always stayed on that VMS cluster since we ran a custom RADIUS daemon developed by an employee which was heavily tied into SYSUAF and VMS’s clustering functionality. Only the owner and one remote employee on the other side of the world actually knew enough about VMS to fix things when they broke, so everyone in tech support and ops mostly limped along only knowing a few commands to reset user passwords and stuff.
The owner was always saying how great OpenVMS was because of its native clustering, but honestly we never really used it that much and the TCP/IP failover that we did use later could have easily been replicated with OpenBSD’s carp or similar. VMS’s versioned filesystem was neat in concept but terrible in practice. For each time having an automatic backup of a screwed up file saved us, there were a dozen times when we ran out of disk space because the old versions were eating up all the space. It was like ZFS with none of the useful administration tools or speed.
I haven’t worked there in 7 years but I see they still have one OpenVMS machine online with telnet running. All of the old cluster hostnames resolve to that one machine which makes me kind of sad. I wonder if it’s still online just for the owner to use and everything of importance has long been migrated off of it.
Every time I hear about OpenVMS all I can think about is the Deathrow Cluster which looks like it’s currently offline.
The mid-sized ISP I used to work at primarily used an OpenVMS (Alpha) cluster to do most of the core functionality like centralized RADIUS authentication, SMTP, POP3, DNS, web hosting, and even shell accounts for tens of thousands of users (though only about 3 users only ever telnetted in). For the first few years, the TCP/IP stack on those servers was a commercial package sold by a 3rd party vendor (like Trumpet Winsock) and I hated that SSH was commonplace by then but we still had to telnet into these VMS servers because they had no SSH daemon.
I migrated a lot of functionality to OpenBSD servers (and eventually a new ops team replaced those with Linux) but RADIUS always stayed on that VMS cluster since we ran a custom RADIUS daemon developed by an employee which was heavily tied into SYSUAF and VMS’s clustering functionality. Only the owner and one remote employee on the other side of the world actually knew enough about VMS to fix things when they broke, so everyone in tech support and ops mostly limped along only knowing a few commands to reset user passwords and stuff.
The owner was always saying how great OpenVMS was because of its native clustering, but honestly we never really used it that much and the TCP/IP failover that we did use later could have easily been replicated with OpenBSD’s carp or similar. VMS’s versioned filesystem was neat in concept but terrible in practice. For each time having an automatic backup of a screwed up file saved us, there were a dozen times when we ran out of disk space because the old versions were eating up all the space. It was like ZFS with none of the useful administration tools or speed.
I haven’t worked there in 7 years but I see they still have one OpenVMS machine online with telnet running. All of the old cluster hostnames resolve to that one machine which makes me kind of sad. I wonder if it’s still online just for the owner to use and everything of importance has long been migrated off of it.