[email protected] (Timothy Sipples) writes:
> Your choice of workloads for conducting such measurements matters far
> more than anything else. If you constrain your universe of workloads
> to only those that can "conveniently" run on the oldest system, you're
> already putting your thumb on the scale in favor of that oldest
> system. "MIPS" does that. Surely it should be obvious that one zEC12
> MIPS is on a completely different planet than one S360-M65 MIPS.

re:
http://www.garlic.com/~lynn/2014h.html#2 Demonstrating Moore's law
http://www.garlic.com/~lynn/2014h.html#4 Demonstrating Moore's law

different planet also shows up with the large cloud operators, their
scale of operation is such that they now assemble their own servers for
1/3rd the price of brand name vendors (hp, dell, ibm, etc) and the
server chip makers say they are now shipping more chips directly to
cloud operators than to brand name vendors.

some big shifts ... the cloud operators view systems as cost item while
the brand name vendors view systems as profit items. the other is that
the large cloud operators have reduced system costs so enormously
... that system cost/BIPS is no longer dominating factor (behind faster
& faster systems at lower cost) but energy/BIPS costs ... motivating
shift from focus on pure throughput to increasing focus on energy
efficiency.

they previously had to grapple with the largest systems that weren't
large enough to meet requirements and built megadatacenters typically
with hundreds of thousands of systems (with millions of processors ... a
typical megadatacenter has more processing power than the aggregate of
all mainframes in the world today), having to deal with all the details
of scaling to large millions of components.

with the enormous drop in system costs ... a megadatacenter can possibly
justify complete roll-over of all systems every chip generation or two
based on energy savings (also one of the reasons that the cloud
megadatacenters have been on the forefront of green computing). also
behind increasing references to energy efficiency in product news.  Big
megadatacenters are also testing non-i86 chips that were originally
developed for battery useage power efficiency.

some large part of the (cost, throughput, power useage, etc) innovation
over the past two decades is attributed to competing companies producing
chips

recent mention of Moore's Law

Computex 2014: Intel says Moore's Law will drive the next era in
computing
http://www.zdnet.com/computex-2014-intel-says-moores-law-will-drive-the-next-era-in-computing-7000030178/
Intel president: Expanding processor technology for smart computing
http://www.infotechlead.com/2014/06/06/intel-president-expanding-processor-technology-smart-computing-23211

other industry standard benchmarks are less useful for demonstrating
"moore's law" (throughput over the decades) ... because they are
periodically updated to reflect current conditions (more oriented
towards reflecting throughput across a broad range of systems).

one set of long-time industry standard benchmarks
http://www.tpc.org/information/about/history.asp
by Jim (I worked with Jim at IBM SJR during days of the original
relational/sql implementation ... and then he was palming bunch of stuff
on me when he was leaving for tandem)
http://www.tpc.org/information/who/gray.asp

IBM has extensive participation here, but only for their non-mainframe
platforms ... mainframe benchmarks haven't been available for decades.

-- 
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