🚀 go-pugleaf

In article <405e4m$3vm@Jupiter.mcs.net>, Karl Denninger <karl@Mcs.Net> wrote:
>Simple.  You build FMP cluster configurations such that no single CPU is
>overloaded, and that both handles the problem and gives you enormous
>redundancy.
>
What do you mean by FMP?

>Then you cluster them with high-performance network architectures.
>
The question comes down to: Can you get the reliabilty from PC based
hardware?  From my experience on three providers around here who are
running on PC based hardware, the answer seems to be leaning towards
no.  Keep in mind all three of those providers do outstanding jobs,
but it doesn't stop their networks form exhibiting frequent problems.
It seems such high-pressure systems as newsservers, and mailservers
are the most prone to troubles and go down with enough frequency that
it's no longer a surprise.  News service it seems is the most
difficult to do properly on those machines.  how many services have
had disk failures and SCSI card problems, Telerama has had some(in
PGH), as well as MCS in the past(or were those just disk failures?)  

>It depends on what you're doing and how you structure your backbone.  We are
>installing DS-3s now (the first is in service as of today) to provide for
>more performance, and more capacity, across our backbone.  This is a
>pro-active step we're taking to help our customers work more efficiently; it
>is not a question of our being out of capacity.

are the links thru sprint and net99 still there?