Subj : Re: Weird Sun Pricing
To   : poindexter FORTRAN
From : tenser
Date : Sun Feb 11 2024 13:15:12

On 08 Feb 2024 at 08:14a, poindexter FORTRAN pondered and said...
 
 pF> -=> tenser wrote to poindexter FORTRAN <=-
 pF> 
 pF>  te> Yup.  The Ultra 5 had PCI and, I think, IDE disks.  That
 pF>  te> was the signal that Sun had lost, for me.  I remember a
 pF>  te> sysadmin friend showing me an Ultra 5 or 4 next to a Pentium
 pF>  te> machine from IBM: "yup, the Sun is twice as fast, but the
 pF>  te> PC costs a quarter what the Ultra does."  That was it: x86
 pF>  te> was going to eat SPARC's lunch, and it did.
 pF> 
 pF> I lived through those times, too. I was a *BSD/Solaris admin in the mid
 pF> to late 90s, Linux wasn't quite ready for primetime.
 pF> 
 pF> In 2000, I managed a web site that ran a stack of cheap Intel linux
 pF> boxes running Apache with an Oracle/Sun back end. Sure, it didn't run as
 pF> well as a Solaris front-end, but you could buy a bunch of white-box
 pF> servers for the price of a Sun.

Yup.  Google ran into that exact same thing, and realized
that for the price of a big machine from Sun, DEC or SGI,
they could buy an enormous number of cheap x86 machines
and scale horizontally.  At one point, we were even buying
manufacturer reject RAM for like 25c on the dollar (those
days are long gone, though).  We realized that if you plan
for unreliability at the software layer, you can save a
huge amount on hardware, and we had the sort of scale where
the economics on that actually made sense.

--- Mystic BBS v1.12 A48 (Linux/64)
 * Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

.