Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A Raspberry Pi (VAX) Cluster

designspark.com

Post reply on HN
Print view View original

1–10 of 20 posts

Re: A Raspberry Pi (VAX) Cluster

#3

Post by gouranga »

Ahh the memories.

At my first proper job, we had a 8 node VAXCluster running Oracle 7, VB3 clients on Windows 3.1 on Compaq and DEC machines, Motif clients on Solaris 2.4 on SparcStations :)

The VAXCluster had 100% uptime for 4 years when they killed it and replaced it with Alpha machines.

Re: A Raspberry Pi (VAX) Cluster

#4

Post by excuse-me »

gouranga wrote: post #3

Ahh the memories. At my first proper job, we had a 8 node VAXCluster running Oracle 7, VB3 clients on Windows 3.1 on Compaq and DEC machines, Motif clients on Solaris 2.4 on SparcStations :) The VAXCluster had 100% uptime for 4 years when they killed it and replaced it with Alpha machines.

I remember DEC sending out a fix for the "day 10000 bug".

Some system counter would wrap around if it was up for 10000days (27years)! Compare it with the Windows95 49day bug that nobody ever reached.

Re: A Raspberry Pi (VAX) Cluster

#6

Post by Hoff »

excuse-me wrote: post #4
gouranga wrote: post #3

Ahh the memories. At my first proper job, we had a 8 node VAXCluster running Oracle 7, VB3 clients on Windows 3.1 on Compaq and DEC machines, Motif clients on Solaris 2.4 on SparcStations :) The VAXCluster had 100% uptime for 4 years when they killed it and replaced it with Alpha machines.

I remember DEC sending out a fix for the "day 10000 bug". Some system counter would wrap around if it was up for 10000days (27years)! Compare it with the Windows95 49day bug that nobody ever reached.

Some DEC code for OpenVMS had been converting variables containing times using the Unix epoch date to and from the OpenVMS-format dates* using the LIB$ LIBRTL calls, and the LIBRTL calls had a documented format limit of 9,999 days.

When 19-May-1997 rolled around, you needed to use a different conversion sequence for your Unix epoch dates, or to have applied the then-available LIBRTL 10K Delta Time patch that extended the permitted day field.

More to the OP's point, the longest continuous OpenVMS server uptime I'm aware of was around seventeen years.

The downside of that being seventeen years of unapplied patches.

*VMS uses 17-Nov-1958 as its base date; that date was chosen to match the date commonly used by the Smithsonian Astrophysical Observatory

Re: A Raspberry Pi (VAX) Cluster

#9

Post by excuse-me »

Hoff wrote: post #6
excuse-me wrote: post #4

Earlier quoted context omitted.

I remember DEC sending out a fix for the "day 10000 bug". Some system counter would wrap around if it was up for 10000days (27years)! Compare it with the Windows95 49day bug that nobody ever reached.

Some DEC code for OpenVMS had been converting variables containing times using the Unix epoch date to and from the OpenVMS-format dates* using the LIB$ LIBRTL calls, and the LIBRTL calls had a documented format limit of 9,999 days. When 19-May-1997 rolled around, you needed to use a different conversion sequence for your Unix epoch dates, or to have applied the then-available LIBRTL 10K Delta Time patch that extended…

Actually "1858" - the base for Modified Julian Days.

Although ironically MJD was invented in around 1958 to cope with a computer limitation.

vms is also famous for having one of the system time settings specified in "micro-fortnights"

Re: A Raspberry Pi (VAX) Cluster

#10

Post by tytr »

excuse-me wrote: post #4
gouranga wrote: post #3

Ahh the memories. At my first proper job, we had a 8 node VAXCluster running Oracle 7, VB3 clients on Windows 3.1 on Compaq and DEC machines, Motif clients on Solaris 2.4 on SparcStations :) The VAXCluster had 100% uptime for 4 years when they killed it and replaced it with Alpha machines.

I remember DEC sending out a fix for the "day 10000 bug". Some system counter would wrap around if it was up for 10000days (27years)! Compare it with the Windows95 49day bug that nobody ever reached.

Yet people criticised the VAX. I guess every computer has its flaws (nothing is perfect!), but was the criticism of the VAX justified? (Assuming uptime and reliability means anything, and compared to what followed it. Windows NT? C'mon.)

It is just one more bit of evidence that tells me that many of the supposed "experts" on matters of computing are anything but. There's FUD and then there's just people who just have no idea what they're saying. (Of course the CEO of DEC got caught making the dumbest comment ever. No one is immune.) Maybe we should just focus on results and not what people say?

Post reply on HN