[astroalex]

I found the preamble at the beginning of the announcement charmingly dated:

> The Minor Planet Electronic Circulars contain information on unusual minor planets, routine data on comets and natural satellites, and occasional editorial announcements. They are published on behalf of Division F of the International Astronomical Union by the Minor Planet Center, Smithsonian Astrophysical Observatory, Cambridge, MA 02138, U.S.A. > > Prepared using the Tamkin Foundation Computer Network

Looking up the Tamkin Foundation Computer Network: https://www.minorplanetcenter.net/iau/Ack/TamkinFoundation.h...

> The OpenVMS cluster consists of nine single-CPU workstations and one four-CPU server. All the machines are running the extremely robust and secure OpenVMS operating system. The twelve Alpha-based machines are arranged as an OpenVMS Cluster, allowing all machines to share disk storage, execution and batch queues and other resources, as well as simplifying system management.

Assuming "Alpha-based machines" is referring to the DEC Alpha, these computers are ~30 years old. https://en.wikipedia.org/wiki/DEC_Alpha

[404mm]

Maybe not as old. I deployed a few racks of HP Alpha DS25s in 2007-2008 before they were replaced with Itanium based Blades (running OpenVMS 8.4). I do not miss working with OpenVMS one bit. It was rock stable (basically an on/off appliance) but the user experience left me wanting (coming from Linux).

I can see how they may be still stuck on Alphas because unless they can somehow simply recompile for x86-64 OpenVMS, it’s a complete rewrite from scratch.

[bastardoperator]

Could they not get more juice out of a single, modern server? I get porting over to a new system and migrating is a huge time suck and a good enough reason not to do it if everything is working, just seems excessive for 14 cores.

[Macha]

> Could they not get more juice out of a single, modern server

They could probably get more performance out of one core on a modern phone, never mind a single modern server. But you see some really old systems in a lot of equipment, not because the porting costs are expensive, but the certification of proving the new system works the same is more than the operational cost of the legacy equipment.

[api]

I’ve heard of consultants who will virtualize systems like this in place using qemu emulation of CPUs like Alpha and Sparc and run it on a single server or in the cloud.

[dragonwriter]

Sure, but the capital and one-time cost of acquiring and shifting to the modern server would be non-zero, and it would entail some risk. (While OpenVMS is maintained and runs on newer systems, that doesn't mean the software that matters on the existing cluster would run without modification.)

It probably would save operating costs, and probably over a reasonably short window, if it was done successfully, though.

[rubitxxx10]

> Could they not get more juice out of a single, modern server?

Maybe the software they use won’t easily run on a modern server.

You could ask them, but you might have to hook up your modem and try to call them. Maybe they have a BBS you could leave your question on.

[ccgreg]

In 2020 I toured the machine room and those boxes were powered off.