[email protected] (Mike Schwab) writes:
> 10 or 20 Linux servers consolidated onto 1 x86-64 blade server.
> 300 Linux servers consolidated onto 1 zIFL.
>
> Now that looks reasonable.  A full speed z processor is still 15 to 30
> times faster than Virtual x86-64.

re:
http://www.garlic.com/~lynn/2012l.html#81 Blades versus z was Re: Turn Off 
Another Light - Univ. of Tennessee
http://www.garlic.com/~lynn/2012l.html#87 Blades versus z was Re: Turn Off 
Another Light - Univ. of Tennessee
http://www.garlic.com/~lynn/2012l.html#88 Blades versus z was Re: Turn Off 
Another Light - Univ. of Tennessee

there were two different scenarios ... the large proliferation of i86
servers with very low processor utilization being able to leverage
virtual machines to consolidate 10:1 ... even using the same hardware
(i.e. being able to consolidate ten 10% utilized systems into a single
system ... w/o changing hardware). In the middle of last decade Marines
were talking about 10:1 consolidation of datacenters (each datacenter
with approx. same number of servers ... using virtual machines to
eliminate 90% of the datacenters ... with their servers typically avg.
10% cpu utilization).

then there is the comment about the e5-2600 with 527BIPS being able to
consolidate at 20:1, prior generation servers. 1/20th of 527BIPS is
26BIPS ... 
http://en.wikipedia.org/wiki/Million_instructions_per_second

from above, Intel core2 dual-core as recently as 2006 was 27BIPS.
e5-2600 (@527BIPS) would be able to consolidate 20 such systems all
avg. nearly 100% processor utilization (potentially even w/o
virtualization ... if originally replicated running same application).

the 2006 core2 dual-core was 13BIPS/core ... compared to e6-2600
@527BIPS and 16cores is 33BIPS/core.  2.5 times per core performance
increase over approx. five years but also four times the cores/chip
gives factor of aggregate ten times increase per chip in five years
... and then two chips to give the 20:1. as per previous posts, e5-2600
@527BIPS is equivalent of 10.5 maxed out, 80-processor, 50BIP z196 or
sevens maxed out, 101-processor 75BIP zEC12 (z196 core is 50/80 or
624MIPS and zEC12 core is 75/101 or 743MIPS).

moore's law doubling every 18months hit single processor thruput
increase limits sometime ago ... this old thread quotes intels svp pat
gelsinger having to explain the realities to the CEO of microsoft and
need to adopt parallel programming in order to take avantage of further
thruput increases:
http://www.garlic.com/~lynn/2008f.html#42 Panic in Multicore Land

IBM had somewhat similar moment with 3081 ... which originally wasn't
going to have a single processor version ... but there was issue with
ACP/TPF not having (tightly-coupled) multiprocessor support (ACP/TPF had
loosely-coupled support from early days in 60s) and the major clone
vendor still offering faster single processor machine (potential that
all the ACP/TPF customers would migrate to clone vendor). Initially
there were all sorts of unnatural acts to try and make 3081 attractive
to ACP/TPF customers ... but finally a 3083 had to be offerred by
removing one of the processor from 3081 (there were some physical issues
since 3081 processor0 was at the top of the box ... and it would have
been more of a no brainer to remove processor1 in the middle of the box
... but that would have left the box dangerously top-heavy).

of course the 3081 also had lots of other issues ... one of the
quick&dirty efforts launched after failure of future system (FS internal
politics actively killing off all the 370 efforts ... lack of 370
products also allowed the clone processor vendors to gain market
foothold). misc. past posts mentioning Future System
http://www.garlic.com/~lynn/submain.html#futuresys

recent old email post discussing some of the clone & emulation competion
http://www.garlic.com/~lynn/2012l.html#email890707
in this post/thread
http://www.garlic.com/~lynn/2012l.html#72 zEC12, and previous generations, 
"why?" type question - GPU computing.

discusses 3081 technology being some warmed voer FS stuff, but compares
very poorly with clone vendors
http://www.jfsowa.com/computer/memo125.htm

-- 
virtualization experience starting Jan1968, online at home since Mar1970

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to