No subject

performance isn't the biggest issue.  Obviously there are many factors 
involved with throughput comparisons: quality of NIC and drivers, disk 
subsystem, client configuration, network congestion, etc.  If all of these 
factors conspire to make an SGI or Sun system only marginally superior, 
then there's less incentive to pursue the expensive and proprietary 
solution.  But if there's a demonstrable performance gain to using one of 
these, I'd like to know roughly what that gain is.

Further, we want to have as much fault-tolerance as we possibly can (our 
current SGI server has been less than stellar in terms of uptime), so we're 
contemplating two lesser servers to split the workload in place of one uber-
server.  If the performance gains of SGI or Sun are negligent, it seems 
more prudent to spend the money in multiple servers, as opposed to one big 
one.

Can anyone point me to some (relatively unbiased) performance comparisons / 
information between SGI, Sun, and i386 hardware and OSes?  As I said at the 
beginning, I've had no luck with google searches on those terms.  The 
closest I can find is this benchmark:
http://www.xinet.com/benchmarks.new/index.html
but the tests don't seem to be exactly comparative (different CPU 
quantities and different network switches).

Thanks,
Scott