I've got a stack of 8 VS3100/m38s that're gathering dust right now,
because I've been too lazy^H^H^H^Hbusy to put NetBSD on them since the
DMA SCSI stuff came about. Since you've given me an excuse to, I'll get
them booting this weekend and we can chat. Feel free to email me off
list.

Andy Sporner wrote:

> In the interest of keeping the air clear, please email
> anything about this to me or let's create an architecture
> neutral list where this can be discussed (IE: BSD-CLUSTERS)
> so that nerves don't get pinched.... :-)
>
> Andy
>
> PS.
>
> > The project notes are a bit skimpy on details.
>
> I am going to create a website for this to track details
> I already have a volunteer and could use more.
>
> > Is your "Phase I" to implement a load-balancer like the Cisco
> > LocalDirector,or the Linux-based LinuxDirector (name ripoff)?
>
> It's already done, just hasn't been vax-ported because I have
> a flaky drive that keeps causing the machine to panic when
> I get to a bad part on the disk.  NetBSD 1.4F
>
> >
> > The bit about "reverse proxy server" sounds like a Squid
> > web-accelerator type thing, but is it really?
>
> No.  currently it is set up connection forwarder so that it
> masquerades as a webserver listening on the actual port one
> would expect but is actually making real connections behind
> the scenes to the real server and acting as an intermediary--
> sort of the kind of thing that rinetd does, but with a little
> more smarts.  I really would like to leverage a kernel level
> scheduler instead of this...  It is a real performance impact.
> But not as bad as you would expect though.  It tops out about
> about 300 hits/sec on an architecture that's a dirty  word
> here ;-)
>
> >
> > I only have a single VAX (VS3100/M38), so clustering that
> > requires like
> > architectures won't work there.  Now I do have plenty of i386
> > boxes though.
> > :-)
>
> Welcome to my problem... :-|