http://207.35.137.66/raxco/Webinar_OpenVMS_i2ServerPerformance.pdf
I can only keep the previous document on line for a couple of days so
download it while you can.
Neil Rieck
Waterloo, Ontario, Canada
http://www3.sympatico.ca/n.rieck/OpenVMS.html
And a "doosie" is, what ??
It can also be found here:
Tim.
I see no major difference from what was presented at the
OpenVMS Technical Update Days.
Check out the maximum amount of memory.
BL860c i2
192 GB (w/8 GB DIMMs)
384 GB (w/16 GB DIMMs)
When I started working on 32-bit VAXs (way back when) I had no idea
"I" would ever see a future system with 4 GB let alone 384 GB. I guess
I should have paid more attention to Moore's Law :-)
NSR
Ignorance of the law is no excuse! :)
I have various sized memory chips and cards hung upon my office wall to
remind me of this industry's humble beginnings.
--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)ORG
All your spirit rack abuses, come to haunt you back by day.
All your Byzantine excuses, given time, given you away.
I remember when the draft of the Alpha Spec came out (Yes, I still have
my copy!) The local VAX Guru (Hi, Lori, but I doubt your reading this)
and I used to laugh about it. We talked about needing an entire room
to fit the needed number of SIMMS in and needing cryogenic cooling to
accomodate the heat it would generate.
How times have changed.....
bill
--
Bill Gunshannon | de-moc-ra-cy (di mok' ra see) n. Three wolves
bill...@cs.scranton.edu | and a sheep voting on what's for dinner.
University of Scranton |
Scranton, Pennsylvania | #include <std.disclaimer.h>
I'll go you one better. Although people have stopped asking about it,
there is an RL01 disk pack on a shelf in the student computer lab next
door to my office. All 5MB of it!!
Here are some pictures:
http://207.35.137.66/remote-link/core-memory-16x14-a.jpg
http://207.35.137.66/remote-link/core-memory-4x8-bits-a.jpg
NSR
My reaction: old news.
They talk as if QPI was brand new. I've had QPI based system under my
desk for 1.5 years already. So IA64 is late catching up to the rest of
the industry.
There is a graph showing increased performance per socket. If you double
the cores per socket, I guess you should expect increased performance.
But it doesn't mean an application that relies on one core will see
improvements. However, later on, there is a slide that shows some
improvement per core. But is this due to IA64 finallty getting QPI memory ?
Interesting that they use database access as a "CPU intensive" example.
Always thought it would be IO intensive.
I see the document more as a sales pitch for the new blade based systems
than for VMS.
BTW, does this mean that all IA64 systems from HP are now blade based
with some cabinets having 1 blade and others having manyt blades ? As I
recall, the superdomes are now blade based too.
The ability to bridge QPI between blades is where HP has made some
innovation in the industry. It is only a matter of time before this gets
implemented for the 8086s as well.
For Galaxy/Wildfire class systems, VMS had special utilities to deal
with NUMA nature of memory. For instance, as I recall, INSTALL has
special feature that would allow one to install an
application/open/shared in memory that was local to the processor that
would run the application. (to reduce NUMA memory access across
processors). This document does not seem to mention whether VMS has
regained such abilities now that IA64 also has NUMA memory system.
Or is the performance penalty for NUMA memory accesses via QPI low
enough that it isn't worth bothering ?
Check out the maximum amount of memory indeed.
Do you know how much memory is available in a modern x86-64 blade (eg
the ones from HP).
How does 1TB in 64 DIMM slots sound? BL680c G7 (4 sockets, up to 32
CPUs, Xeon 7500 series). Or 512GB (32 slots), with 48 CPUs, in the
BL685c G7 using AMD Opteron 6100.
There is still some plausible hardware differentiation between IA64
and x86-64 in Superdome-class systems, where x86-64 (and Proliant in
particular) does not yet have the ultimate hardware expandability of
CPUs, memory, and IO that IA64 has in today's Superdomes.
In a blade-based system, that "ultimate expandability" distinction
doesn't exist (almost by definition, blades are small replicated
things rather than a single big box). In a blade-based system, the
hardware balance tilts in favour of the mainstream (x86-64) product
whose technology is more recent and therefore more advanced in terms
of density:
x86-64: http://h18000.www1.hp.com/products/quickspecs/DS_00207/DS_00207.PDF
IA64: http://h18000.www1.hp.com/products/quickspecs/DS_00184/DS_00184.PDF
Of course, the Proliant one isn't much use (yet) to customers whose
needs still require HP-UX or VMS. Those customers, even in a blade
environment where there is no visible hardware reason to prefer IA64,
still have to buy IA64.
Isn't technology wonderful.
"It is only a matter of time before (NUMA/QPI) gets implemented for
the 8086s as well."
Not really. NUMA/QPI is already there in the rack Proliants:
http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA3-0643ENW.pdf
(or search for 4AA3-0643ENW, or "HP Proliant DL980 G7 Server with HP
PREMA Architecture"). That's the one with 8 sockets (64 cores) and 2
TB of memory - all twice what's on the equivalent x86-64 blade.
I don't immediately know if the blade Proliants have an inter-blade
link in the same way as the IA64 blades, but QPI is QPI, so if it
exists for the IA64 blades it must presumably also be technically
feasible for the x86-64 blades (maybe).
That said, whether anybody wants or needs inter-blade links on x86-64
when a single x86-64 blade can already have 1TB and 32 Intel (48 AMD)
CPUs on the blade, and when the rack alternative already offers 2TB
and 64 CPUs, is a different question.
Incidentally, the node controller and inter-node crossbar fabric in
the DL980 G7 is said to be "leveraged from" the Superdome 2, and
appears to be based on quads of processors; within each quad each CPU
has its own local memory, and above that is then a global interconnect
(hmmm, if only I could remember where I heard that before). Something
else where there's now little differentiation between IA64 and x86-64
hardware then?
I have no doubt that QPI is the biggest game-changer in 20 years and
we all should remember that QPI came from CSI technology developed for
Alpha by DEC/Compaq engineers so kudos to them.
Comparing QPI in x86-64 to Itanium is above my pay grade so I need to
rely upon the opinion of others. One reason why I say this is that QPI
seems to have been given a lobotomy in various x86 products. Intel
refers to QPI in the Core-i7- 900 series but then uses the name DMI in
the Core-i7-800 series and lower. Some have claimed that DMI is QPI
with portions of the supporting chipset disabled but I have never
found any proof of this.
With regards to overall Itanium2 speed, some have claimed that the
real speedup came from supporting DDR3 and I suspect this may be true.
One wonders about the throughput of a modern Alpha chip because surely
Alpha would have supported DDR3 before 2010-2011. But, as we all know,
Alpha is dead and unless HP is doing trial ports to x86-64, Itanium is
all we've got.
My employer still runs a lot of Alphas. These Alphas were purchased
from Compaq and cost (relative to similar sized VAX installations)
approximately ten times less. My employer is VERY conservative and we
weren't allowed to move to Alpha until they were in the market place
for ~ 8 years (this isn't a hard rule; the bean counters are trying to
squeezing every penny out of the current investment -AND- they could
only stand us whining for 8 years). OpenVMS on Itanium wasn't released
until Feb-2005 so our group probably would need to wait until 2013 if
history is a guide. Now if Itanium systems were 10 times less
expensive than Alpha (which they currently are not), then we could
make a good case to the bean counters to jump sooner.
Neil Rieck
Kitchener / Waterloo / Cambridge,
Ontario, Canada.
http://www3.sympatico.ca/n.rieck/OpenVMS.html
That was what was told live (on a TUD day I think) from someone
whin insight in the architecture and who had runed some
benchmarks. Most of the better perf was from the faster
memeory subsystem, as he said.
> One wonders about the throughput of a modern Alpha chip because surely
> Alpha would have supported DDR3 before 2010-2011.
Yes, and if the same amount of money/resourses had been pured
over Alpha as has been invested in IA64... :-)
Well, we'll never know...
Jan-Erik.
> Yes, and if the same amount of money/resourses had been pured
> over Alpha as has been invested in IA64... :-)
>
> Well, we'll never know...
When I next jumop to a different universe, I'll have to choose the one
where Bob GQ Palmer get bitten by a dog and dies of rabbies, and Digital
has bought Compaq and HP, and the Alpha chip is industry standard with
AMD making clones... And in that universe, I'll be rich because my VMS
skill will have some value :-)
> [...] Especially one German CPU board (I forget the vendor's
> name) where the reqisters are implemented with cores that have a
> diameter a little smaller than a pencil.
The PCBs aren't marked accordingly? Siemens, Telefunken, Nixdorf, ...
> Here are some pictures:
> http://207.35.137.66/remote-link/core-memory-16x14-a.jpg
> http://207.35.137.66/remote-link/core-memory-4x8-bits-a.jpg
Lots of transistors (TO-18 metal case), resistors and so on.
Michael
--
Real names enhance the probability of getting real answers.
My e-mail account at DECUS Munich is no longer valid.
> Me too. But most visitors to my office become speechless when they see
> core memory. Especially one German CPU board (I forget the vendor's
> name) where the reqisters are implemented with cores that have a
> diameter a little smaller than a pencil.
When the system engineers talked about "core memory" in our 2050,
I first thought it was out of habbit. When experience convinced them
DEC no longer had anyone who knew how to tweak core to eliminate
faults, we replaced it with a later technology and the core memory
was seen sitting in the corner of the computer room for a couple
of weeks.
I'll tkae the universe where Ken Olsen realizes he is selling
software (VMS), and ports it to a RISC chip before everyone
undergoes the pain of moving to UNIX.
Then he can put a great office suite on the thing and actually
advertise it to folks who learn that computers don't crash.
Ah, I see you subscribe to Hugh Everett's Many-worlds Interpretation
of Quantum Mechanics
NSR
I should post the graphic. It is etched into the PCP and looks like a
bird (so typical for a German company :-) where one each wing looks
like the letter "F".
NSR
What?
The impact seems extremely small to me.
> My employer still runs a lot of Alphas. These Alphas were purchased
> from Compaq and cost (relative to similar sized VAX installations)
> approximately ten times less. My employer is VERY conservative and we
> weren't allowed to move to Alpha until they were in the market place
> for ~ 8 years (this isn't a hard rule; the bean counters are trying to
> squeezing every penny out of the current investment -AND- they could
> only stand us whining for 8 years). OpenVMS on Itanium wasn't released
> until Feb-2005 so our group probably would need to wait until 2013 if
> history is a guide. Now if Itanium systems were 10 times less
> expensive than Alpha (which they currently are not), then we could
> make a good case to the bean counters to jump sooner.
Not that unusual.
Arne
I think it would take a lot more than a dog taking out Palmer to change
the direction so dramatically.
Arne
> My employer still runs a lot of Alphas. These Alphas were purchased
> from Compaq and cost (relative to similar sized VAX installations)
> approximately ten times less. My employer is VERY conservative and we
> weren't allowed to move to Alpha until they were in the market place
> for ~ 8 years
Smart move: buying systems only when their processor is being EOL'ed.
JF has been in a different world for many years! ;-)
Okay does anyone recognize this logo from a German CPU circa 1960?
http://207.35.137.66/remote-link/cpu-board-logo.jpg
NSR
Volkswagen made CPUs as well as cars ? :-)
(I don't recognise the logo, but that's what it looked like initially to
me :-))
Simon.
--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world
Two more clues:
1) The logo appears to be the letter "w" with wings representing the
letter "E" or (less likely) "F". So the company name be represented
by:
WE
EWE
WF
FWF
2) I wonder if the logo represents the company that manufactured the
circuit board rather than the computer manufacturer.
NSR
Look a bit down on this page :
http://www.looks.gd/design/the-evolution-of-corporate-identity-design
(Found entering "logotyp german computer manufacturers",
selecting "Images" and visualy searching through the
result. On the second page I think.)
"Audi represents the 1932 merger of four independent
motor-vehicle manufacturers: Audi, DKW, Horch and Wanderer."
The logotyp is from Wanderer.
http://en.wikipedia.org/wiki/Wanderer_%28car%29
> "Audi represents the 1932 merger of four independent
> motor-vehicle manufacturers: Audi, DKW, Horch and Wanderer."
Note that a man named Horch had a company of the same name. When
copyright reasons forced him to change it, he translated it to Latin:
Audi.
Which in English is "hear" or "listen"... :-)
> On 2010-12-12 13:40, Phillip Helbig---undress to reply wrote:
> Which in English is "hear" or "listen"... :-)
Right. There is the somewhat archaic English word "hark", which is
related. (Around this time of year, the herald angels remind people of
the word's existence.)
Gee, I always thought "Hark" was the name of the herald angel.
You know, "Rudolph, the red nosed raindeer", "Frosty, the snowman",
"Hark, the herald angel".
> Gee, I always thought "Hark" was the name of the herald angel.
Hark is a name, a Frisian name. Just yesterday, I visited someone by
that name.
I'll take the one where Apple moved to Alpha instead of PowerPC and this
gave the DEC folks enough stability to second source the chip and allow
it to be a solid also-ran like SPARC until the internet boom of the late
90's made it the standard 64 bit chip for business servers... over the
Sparc...
"Boss... the plain, the plain."
"Yes Tatoo. Folks welcome to Fantasy Island."
-------------------------------------------------------------------------
Nice game to play. We need to do a game board and some hexadecimal dice
for the hostmaster. Some kind of board game based on CPU's and
Monopolies... an alternative history.
The one where the DEC AT&T merger actually went through in '84.
My wife is pissed she through out the Bell Labs internal newsletter where
some of it leaked one day in the past.
(They also were going to have DEC take over operations of the
datacenters and all Unix Sysadmins.)
Me -- I was just fixing Vaxes back then.
Bill
--
Digital had it then. Don't you wish you could buy it now!
pechter-at-pechter.dyndns.org
Ton.
Hey, thanks for pointing me in the correct direction. This board was
made by "Wanderer-Werke" which was eventually acquired by Nixdorf
Just for a laugh do a google search for "Wanderer Werke" in Google,
then repeat the search on the image page
NSR
I've often wondered why an omniscient and omnipotent God needed a staff
of angels!!!
But how many CEOs are omniscient and/or omniscient?
The "Dilbert" comic strip is supposedly drawn from daily corporate life.
If this is true, then clearly, there are very few in upper management
who truly have a clue.
> In article <ie2m2v$3nd$1...@online.de>,
> hel...@astro.multiCLOTHESvax.de (Phillip Helbig---undress to reply) writes:
> Ost Frisian?
The person I know is from North Frisia, but perhaps the name also exists
in East Frisia. (East Frisia is in the west of Germany, since West
Frisia is in the Netherlands. Most people who speak the Frisian
language speak West Frisian, a few speak North Frisian while East
Frisian has essentially died out. This is really a separate language
and has nothing to do with Low German, except that (like Swedish and
English as well) it is a Germanic language.)
And a smarter move would be betting the business on untried
technology?
But they all think they are and still keep toadies around.
>
> The "Dilbert" comic strip is supposedly drawn from daily corporate life.
> If this is true, then clearly, there are very few in upper management
> who truly have a clue.
Dilbert is a cartoon. While satire may reflect someone's view of reality
I would not take it all that seriously.
Of course not. And thus a good reason not to bet the business
on Alpha until the late 1990s.
Unfortunately, in this case, the gap between "untried" and "EOL"
wasn't large.
More detail than I expected. :-)
When i lived in the Nord-Rhein the Ost Frisiche were the brunt of many
a joke, particularly regarding a certain common facial feature. (try
to picture the late Jimmy Durante). Kind of like what they used to do
to West Virginians over here. :-)
> Hey, thanks for pointing me in the correct direction. This board was
> made by "Wanderer-Werke" which was eventually acquired by Nixdorf
>
> Just for a laugh do a google search for "Wanderer Werke" in Google,
> then repeat the search on the image page
Just two links:
(1) "Wanderer" bicycles, sold by Manufachtum, e.g.
<http://www.manufactum.de/Produkt/172606/1445631/Wanderer-Sporttourer-S600-Herren.html>
(2) photo of the company's logo (hood ornament),
<http://commons.wikimedia.org/wiki/File:Wanderer-Werke_A.G._hood_ornament_W126.JPG>
or
<http://commons.wikimedia.org/wiki/File:Wanderer-Werke_A.G._hood_ornament_W126.png>
Michael
--
Real names enhance the probability of getting real answers.
My e-mail account at DECUS Munich is no longer valid.
> When i lived in the Nord-Rhein the Ost Frisiche were the brunt of many
> a joke, particularly regarding a certain common facial feature. (try
> to picture the late Jimmy Durante). Kind of like what they used to do
> to West Virginians over here. :-)
Usually, they are the same jokes in all countries, just the group who is
the butt of the jokes varies.
Yeah, there is always someone on the receiving end but these jokes were
rather explicit. Most stereotypes are based on some piece of fact even
if the logical conclusion is faulty. I never had the chance to visit
Osfriesland so I can't say if the basis was accurate. It wasn't particularly
nasty or anything. I have heard worse.
> Gee, I always thought "Hark" was the name of the herald angel.
>
> You know, "Rudolph, the red nosed raindeer", "Frosty, the snowman",
> "Hark, the herald angel".
Don't forget Olive. You know, Olive, the other raindeer. :-)
Alan "The Other AEF" Frisbie
Almost right.
West Frisia is a piece of the Netherlands about 30 miles north
of Amsterdam, west of lake IJsselmeer. Some people there
speak a dialect that is between Dutch and Frisian.
Frisian is the second official language in the
Netherlands. Frisia (Friesland) is east of lake IJsselmeer,
also in the Netherlands. Ost-Friesland is in Germany,
North Frisia is on the border between Germany and
Denmark. All Frisia's are at the North sea coast, and
are relatively devoid of anything except cattle and grass.
Hark means 'rake', a common tool in the grassy lands.
When I visited Wyoming it didn't feel that different, except that
the Wyoming farmers hadn't yet finished getting rid of
the bumps in the grass.
> Usually, they are the same jokes in all countries, just the group who is
> the butt of the jokes varies.
Experienced Newfie pilot to the new co-pilot:
Flying at night is really easy. See the red light on the left wing, and
the green light on the right wing ? You just need to fly between the two !
> Almost right.
>
> West Frisia is a piece of the Netherlands about 30 miles north
> of Amsterdam, west of lake IJsselmeer. Some people there
> speak a dialect that is between Dutch and Frisian.
> Frisian is the second official language in the
> Netherlands. Frisia (Friesland) is east of lake IJsselmeer,
> also in the Netherlands. Ost-Friesland is in Germany,
> North Frisia is on the border between Germany and
> Denmark. All Frisia's are at the North sea coast, and
> are relatively devoid of anything except cattle and grass.
I think the difference is that in German Westfriesland refers to
everything in the Netherlands, while within the the Netherlands there is
the distinction you mention.