VAX 6460 being slow, IO bottlenecks and SMP woes ...

From: Chuck McManis <cmcmanis_at_mcmanis.com>
Date: Wed Mar 20 21:51:15 2002

At 10:15 PM 3/20/02 -0500, Gunther Schadow wrote:
I'm certainly not complaining, just wondering and trying to
>understand what exactly it is that made people pay millions
>of dollars for a harware that was only a few years later
>outpaced by the poor man's desktop ...

It wasn't. It does thing that today's desktops can't do. When is the last
time you got 99.999% uptime out of a PC? Even though the theoretical
bandwidth of the PCI bus was 120+ MB/sec systems rarely achieved > 1 - 2
MB/sec sustained throughput.

>... ... now, I don't
>know that yet, I actually don't believe it. I still suspect
>that once the bottlenecks are uncovered, I can find
>workarounds to coerce more of the idle hardware into working.

Always true. Unlike the PC there are knobs for making things faster on the VAX.

>For instance, calling GNU make with the -j option and
>calling gcc with the -pipe option appears to create a
>lot more CPU load:
>
>cpu us% ni% sy% id% csw sys trap intr ipi ttyin
>ttyout
> 1 30.3 0.0 6.4
> 63.2 89k 288k 300k 218k 13 3k 190k
> 2 22.4 0.0 1.4
> 76.2 27k 84k 74k 0 24 481 22k
> 3 20.5 0.0 1.4
> 78.1 28k 80k 70k 0 13 526 17k
> 4 23.1 0.0 1.5
> 75.4 28k 86k 74k 0 25 531 23k
> 5 21.2 0.0 1.5
> 77.3 29k 88k 72k 0 25 550 19k
> 6 19.5 0.0 1.5
> 79.1 29k 86k 73k 0 26 600 25k
>
>that looks better, at least the others are not
>97% idle now.

Doing a parallel make would be even faster.
--Chuck
Received on Wed Mar 20 2002 - 21:51:15 GMT