Posts by Jesper Frimann

Well

In what way is Tukwila,POWER7, Nehalem-EX, Mangy-cours or SPARC64 VII not CMT processors ?

In what way does the same above processors not implement Multi-threading (if you count Mangy-Cours out) ?

I don't really think you understand why people say Niagara is cache starved. If used for what it was originality designed for, running many threads that all execute the same code, which uses data with a relative small memory footprint, then Niagara will really excel.

The whole concept of using the 'Niagara' CoolThread concept to for example running webservers is a very good idea. This is the niche that the Niagara style processors fit into.

BUT, running 8 different programs on each of the HW threads on a single Core, specially if this code uses data with a big footprint, is a terrible idea.

There is no magic here that allows Niagara to disregard the laws of physics. Although there are a lot of posters here who think so.

// jesper

RE:Bill

Jup, we sit and plan the downfall of Oracle....

Intel and IBM have been pretty consistent the last many years with beefy cores, with good single threaded throughput.

And yes the next logical evolution step of the TX processors is to get beefier cores OOE etc etc. Just like Xeon and POWER. But in doing so you are basically abandoning the whole coolthread concept.

And if you disagree, then perhaps it's cause you never really understood the CoolThread concept.

// Jesper

Coolthread

Lets see... T1, T2, T2+ and T3 all implement simple in-order SPARC processors that rely on fine grained multi-threading, and other techniques, to hide memory latencies. Another technique is actually having more L1 and L2 cache per thread per throughput unit, than for example POWER7.

From what I've been able to dig up on T4 then it's going to be a more complex core that implements OO and perhaps even SMT, that sounds a lot more like Nehalem-EX and POWER7, rather than the original CoolThread concept.

So you haven't even technically understood the whole coolthread concept, and actually also the beauty of the design. Cause for Niagara is a brilliant concept for what it was designed for. But again again the problem is that now it is peddled for workloads that it wasn't designed for.

And also the power usage of the T2+ and T3 are in the same range as for example Nehalem-EX.

// Jesper

Cache.

"Some processor architectures are impacted more than others with a small cache. ". Unless "Other processor architectures" only refers to AMD's x86, then your statement is wrong.

POWER implements SMT, with POWER7 implementing 4 way SMT.

Intel x86 implements 2 way SMT.

SPARC64 implements 2 way SMT.

Niagara style processors have up until now implemented fine grained multi-threading, which of cause will help greatly on throughput when hardware threads encounter a cache miss.

Now a T series machine running the workloads it was designed for, like for example a single application with a fairly small code footprint and possible many instances or the app using many threads, will not be as affected by it's small cache, as it will running for example a myriad of different applications with a lot of updates.

// jesper

So ...

Basically admitting that the CoolThread concept was wrong.

// Jesper

hmm

"Maybe Niagara sales are lower than Xeon, because of the same reason Itanium sales are lower? "

What HP have done is to drastic reduce the price of SuperDome's with a factor of 4-5 for the same performance. I tell you that is something that can be seen on sales numbers.

IBM have done much the same, as they are selling POWER7 systems that are 4-5 times faster than their POWER6 counterparts at the same price or less.

Oracle are basically selling servers at the same price, so for the M series it's ... well.. 20% more bang for the same buck. IMHO M series just isn't competitive any more with HP and IBM offerings.

T series is basically sold at the same price as before, so it's a factor of 2 better in price performance.

"Also, recently, the TPC-C world record of 30 million tmpc was done on Niagara running "Slow-laris"."

It's a cluster... CLUSTER. says nothing about solaris scalability or performance. Yes each server in the cluster does around 1M tpmc for a 64 core server.. That is about 25% of the nodes that IBM or HP could have used.. 3 1/2 years ago. The only real good story about the submission is that the software could scale to 27 nodes, which I find really terrific. But lets substitute those 27 nodes with POWER 780's or SuperDome2's.. and then you'll see throughput. Wroom Wroom.

Now if you could administrate 27 T3-4 as a single system image, then you'd have to use the same tools as used on BlueGene or SGI's Altix'es. Cause there would be 13.824 hw threads to keep track of. *BOGGLE*

// jesper

It's not just..

ZFS is not a shareable filesystem.

See under Limitations # 3.

http://www.solarisinternals.com/wiki/index.php/ZFS_for_Databases

// jesper

Perspective.

"Your point? So? It makes sense to protect your customers investments. I think Oracle should be applauded."

Sure, I am not complaining. I am simply sharing an observation with the readers here that might save them a lot of money. One that I am going to use in my role as an architect that does Oracle solutions.

Hence if you are buying a new Oracle License and are going to put it on a Itanium machine, then upgrade one of your existing machines to Tukwila rather than go buy a new one. I mean the savings can be easily be in the hundreds of thousands USD.

"Seriously? You don't understand? Only a marketing droid would have trouble understanding this..."

Well software licenses usage is meant to reflect how much capacity or clicks or whatever term you want to use, you are using the software for.

What doesn't make sense to me, as an oracle customer/partner, is that there haven't been a consistent mapping between T-processor throughput/capacity and the number of licenses you'd have to pay, for the five years that the T series machines have existed.

Sure if I were a 'marketing drone' I'd say Sure it's cause they wanna sell new hardware, but that is not my focus. I couldn't care less. But this constant changing of licenses, does take up to much of my time, discussing with our clients on how to protect their Oracle SW license investment, and minimize their cost. And sure that means that I am sold as consultant for quite some hours, but I should be using my time on making solutions.

// Jesper

Well you forgot that T3

Yes, basically they made it cheaper to use their own hardware.

Although there is a loophole in the Itanium per core licensing scheme. It clearly reads:

(For servers purchased prior to Dec 1st, 2010)

and

(For servers purchased on or after Dec 1st,2010)

So basically this price raise is only valid for new _machines_, bought after Dec 1st 2010.

Which basically means if you have a current Montvale/Montecito based machine that can be 'upgraded' to use Tukwila then the actual machine have been bought before 1st of December.

So the wording suggest that upgrades of processors goes free, as you are not buying a new machine.

Furthermore the licensing scheme doesn't make much sense, just look on how the 'per core' licenses for the T series processors seem to go up and down

T1 -> 0.25

T1(+) -> 0.50

T2 -> 0.75

T2(+) -> 0.50

T3 -> 0.25

// jesper

Well I think that Timothy Prickett Morgan got it wrong..

I don't think that Timothy Prickett Morgan, really nailed it.

If you compare the Old SUN roadmap:

http://regmedia.co.uk/2009/09/11/sun_sparc_roadmap.jpg

with the newer Oracle roadmap for SPARC:

http://regmedia.co.uk/2010/12/03/oracle_sparc_roadmap.jpg

T3/Rainbow Falls with x2 cores compared to Victoria Falls is clearly listed as T-Series x2 Throughput on the Oracle roadmap in 2010.

T4/Yosemity Falls with half the number of cores to Rainbow Falls (Possible x2 faster cores) and 1.5 times the frequency with gotta be the T-Series in 2011 with x3 the thread throughput.

Then on the Oracle map there is a M-Series with 64 sockets listed in 2012. This corresponds very well to the YellowStone Falls on the old SUN map.

And finally on the Oracle map there is a T-Series in 2013 with 1-8 sockets, which gotta be cascade falls, which is also listed in 2013.

The real surprise here, if my deductions are right, is that M is going CMT processors also.

But otherwise it seems that Oracle is just executing on the existing "SUN roadmap".

// Jesper

Try to understand this then...

"When T2 got a speed bump from 1.4GHz to 1.6GHz, it can be considered as a next gen cpu - and should be pitted against next gen cpu from IBM: POWER7".

Please don't use pseudo quotation marks on a statement that I have clearly not said, if you are not familiar with the use of quotation marks don't use them.

With regards to T2 versus T2+, some call it a speed bump, others actually lists it as a separate processor. But if I am sooooo wrong why then does the SPARC WIKI list's the T2 and the T2+ as a separate processors. Try to have a look:

http://en.wikipedia.org/wiki/SPARC

Now is that Wiki written by SUN/Oracle hostile bad guys who want to bring down the world ?

Not likely.

POWER6+ was also just a speed bump, but it's still an, to use a Intel term, *tick* *tock* generation. Just as Westmere-EP is to Nehalem-EP.

"There are numerous other weird claims from him."

Perhaps that is cause you don't understand them. Yes I am known for thinking out of the box, and coming up with creative solutions to traditional problems.

"I think it is quite fun that his statements are so extraordinary remarkable that people question the truthness in his statements! That is clearly a sign of how strange his statements are."

Since when did you start to become plural ? My claims are normally always backed up by 'facts'.

Not that I'm not wrong sometimes cause I am.

"When he claims that POWER6 is faster cpu than Niagara cpu because POWER6 has a faster core? Yes, it is true. He said that! No matter how many cores Niagara has - if the core is slow, then the entire cpu must be slow. According to Jesper."

Try to read what I wrote:

http://forums.theregister.co.uk/post/672443

I list the performance difference between T2+ and POWER6 at equal Thread count, equal Core count and Equal Chip count. You don't get more honest than that ?

And my conclusion was:

"So basically Niagara has a slim lead on POWER6 only when comparing chips to chips, and at ithe best result is a 60% lead and that is only against lower clocked POWER6 on SPECint_2006rate. On specfp_2006rate (16 way) each 8 core, 32 threaded T2+ chip is only 20% faster than 2 core 4 threaded POWER6+ chip."

If you had bothered looking after a post where I get enough of your rant and lash out at your arguments then you should have quoted this one from (http://forums.theregister.co.uk/post/673128) :

"The whole point is that it doesn't suck, power6 is damn fast. The project I am responsible for has no less that 16 power 570'es and a few p595's. And damn they are fast, and actually quite forgiving due to their high Ghz. Again if there is a CPU core that sucks then it is the T2. Now both Itanium and POWER has managed to keep One of the reasons why Itanium still sells fairly well is that it's single threaded performance actually is pretty good. And it will, IMHO, get better with Tukwila. And you still can't get it into your thick head that the one machine with the myriad of CPU cores is the T5440. It has bloody 32 cores with no less than 256 threads. Man that is half the threads of a maxed out M9000."

Again notice that I use CPU core, not CPU, not CHIP.. And I still stand by my point.

Ok it is now official...

You are mad.. simply mad.

Have it ever occurred to you that the reason you cannot find any link to me saying that the POWER6 CHIP will always do more throughput than a Nehalem-EP or than an T2+ Niagara is cause I have constantly written CPU core or core. Admit it you have misunderstood me and get over it. Look... I have constantly tried to list the facts.. Good example here:

http://forums.theregister.co.uk/post/672443

I list the performance difference between T2+ and POWER6 at equal Thread count, equal Core count and Equal Chip count. You don't get more honest than that ?

So perhaps you should duck out of the way of the Orbital Mind Control Laser next time it passes.

// Jesper

Yeah Yeah Yeah..

"Who the heck cares about scaling?"

Well I do.. Scaling is a very important. Which is why it is something that should be looked at when comparing systems and processors. Specially if you are comparing Itanium/POWER/SPARC64 versus x86. Cause Itanium/POWER/SPARC64 will do from 4-256 cores with the same chip. Intel x86-EP will only natively go to 2 chips and 8-12 Cores. Hence when you pull out your Nehalem-EP numbers.. then they are only valid for workloads that does not require more scalability than the Nehalem-EP chip can deliver. For larger workloads you have to look at Nehalem-EX. It's actually pretty simple.

If I need 300 specINT_rate2006, then it's kind of stupid to be looking at a Nehalem-EP x5570 processor cause it'll only take you to about 266. Where as the nehalem-EX X7560 will take you to.. well.. 1400 or so..

So that whole scalability comes at a price, and that is the per core throughput. Again the two above processors in 2 socket configs:

http://www.spec.org/cpu2006/results/res2010q3/cpu2006-20100621-11923.html

and

http://www.spec.org/cpu2006/results/res2010q1/cpu2006-20100315-09857.html

That is 266 and 33.25 per core specINT_rate2006 for the Nehalem-EP and 385 and 24,06 per core for the Nehalem-EX. So the 45% increase in chip throughput comes at a price of a 38% drop in per core throughput.

And you talked about TPC-C

Then

Nehalem-EP (Submit date 04/08/10 ) 631766 tpmc and 78971 tpmc/core:

http://www.tpc.org/results/individual_results/HP/HP_DL370_G6_OEL_TPCC_ES.pdf

And

Nehalem-EX (Submit date Aug 27, 2010) 1807347 tpmc and 56489 tpmc/core

http://www.tpc.org/results/individual_results/HP/HP_ProLiant_DL580G7_2.26GHz_es_100830_Energy_v2.pdf

And finally

POWER6 (Submit date May 21, 2007) 1616162 tpmc and 101010 tpmc/core

http://www.tpc.org/results/individual_results/IBM/IBM_570_20070522_ES.pdf

Again Nehalem-ex is 40% slower per core than -EP. And the picture is clear POWER6 is faster per core, (on this benchmark) than Nehalem.

And also notice that the POWER6 submission is from 2007, not the fastest POWER6 made, and the Nehalem are using SDD drives (which IMHO is a pretty big factor).

Now a current POWER7 system like this one:

http://www.tpc.org/results/individual_results/IBM/IBM_780_TPCC_20100719_es.pdf

does 1200011 tpmc and 150001 per core. Which is almost x3 per core of Nehalem-EX and almost twice that of Nehalem-EP per core performance.

And the sheer fact that you can come with such a statement, only shows that you still have much to learn.

"No, it is you that dont get it. Let me ask you, if you needed the highest performance in the world, which company would you have to go to, IBM or Oracle? Oracle! How the heck can IBM and IBMers claim they are still fastest in the world, because "IBM cores gave 4.7 times more tpmC per core"?? That is an outright lie, that technically ignorant executives might believe. FUD and lies, again. This is a lie, Jesper. Dont you see the lie?"

I wouldn't go to Oracle that's for sure. It's not the fastest machine.. it's a cluster .. CLUSTER. And if you read the pricing information on their TPC-C benchmark you'll see that you don't even buy the software... you lease it.. For the exact amount of years that the benchmark has to do TCO on. You have to pay the listed amount of money every 3 years... And there is NO upgrade protection.

So who is trying to Bull who ? On the POWER 595 benchmark you at least buy the software and then only has to pay (which can be expensive enough btw.) software maintenance and support.

So buying 2 years of extra support for DB2 on the (going from 3 year -5 year TCO) would cost you ... 203,827 USD-Discount. On the Oracle solution using the pricing scheme that they use on the benchmark it would be.... 7,872,000 USD for 3 years.

Now who is b*llsh*tting who ?

You are so drunk on the Oracle/SUN Koolaid that you don't care to read the fine print.

"In short, when I show links or benchmarks, you immediately dismiss them as FUD, lies, and amateurs. When you show links or benchmarks, I accept them. What does that tell you, about Jesper Frimann?"

That I'm much much better at coming up with links than you are ?

It's not my fault that IBM doesn't release any benchmarks on Mainframes. Go complain to them. I just know the numbers that we use. Would a 8 core Nehalem-EX chip faster than a 4 core Mainframe chip on benchmark like specINT_rate2006. Jup sure no argument from me.

But 8 times faster, is so far off the target that it can only be described as FUD, or well.. stupidity.

Also here the type of workload plays a big part. Mainframe Cores have never been known for their ability to crunch numbers, Moving data on the other hand they are pretty good at, and that is also one of the secrets why they can run at such high utilization. Which is also a factor that you have to look at in RL.

"That is hilarious. And proves the z196 is an abomination that should have been killed off and never left the laboratory."

You really really don't get it. Your fanaticism is scary.

"You make it difficult for me. Sometimes you say one thing, the other time you say the opposite.."

That is most likely cause you don't understand what I say.

Now as for response time. I've only pointed out differences in that on the benchmark and said that you can trade response time for throughput.

On the TPC-C benchmark you are quite right there is a big difference in response times, between the POWER 595 and the T5440 clustered benchmark.

On that particular benchmark then one of the key factors is surely the use of SDD on the T5440, which wasn't available back when the POWER 595 benchmark was made. It's a pretty big difference.

But the T5440 clustered benchmark do seem to have idle processing power that could have been used to increase the throughput at the cost of response time. How much is hard to say, why well according to a friend of mine who's a certified Oracle RAC dude, then it's cause they are pushing the limits for scalability on the benchmark, hence they aren't hitting the optimal per core throughput.

// Jesper

Damn you are scary...

"Just as you denied that Intel Nehalem was faster ...."

Yak Yak Yak... You are like a record stuck in the same loop. You never understood a word I said. That is clear.. and you just keep on repeating the same and the same again and again....I've never denied that Nehalem-EP was a faster CHIP than POWER6, but I've denied that it was a faster core.

Furthermore the scaling of Nehalem-EP was.. well.. EP like :)=

"That is maybe why you claimed that POWER6 is super fast - when in fact, it sucked badly."

Again in the myriads of benchmarks you have managed to find one multi-tiered Oracle controlled benchmark where you can claim a victory.

Now on Industry standard benchmarks like SAP2Tier, specINT_2006rate, specFP_2006rate, specJBB2005... POWER 570 wins with a good margin.. And even on your clustered, then each T5440 is 2.5 times slower than a stand alone POWER 570.

Wake up you level of fanaticism is scary.... Oracle is Great.. !!! All hail SUN I mean .. Oracle !!

"If you needed the highest performance in the world, you had no choice but to use Oracle,"

Again here you don't get it... real IT people who do sizing work and architect solutions know that if IBM submitted a 80 node POWER 750 node DB2 Purescale clustered submission, that this wouldn't make the POWER 750 the fastest machine in the World. It's just a node in a cluster.

" You forget that I have a math degree, I would never think like you do."

Jup, and you wrote your thesis in Bistromathic, which seems to be the only math you can use to prove your points.

"In short, IBM = Master of FUD. Jesper, you have been fooled by IBM. You DO believe the IBM Mainframes have fast cpus. You DO believe the POWER6 is fast."

Keb. I bear no illusions what so ever about how fast mainframe CPU's are. Both on a per core level and also on a per CHIP level. As a part of my Architect job at a Major CSI player I have access to NDA IBM sizing information. I've sized mainframe systems using that data and seen it hold water.

So, the CHIP isn't great IMHO when it comes to throughput, but they are nowhere as bad as you claim. And I couldn't give a damn how many amateur wannabe hackers you link to, it doesn't make it more real. Cause it's on the internet doesn't make it real.. hope you don't believe this guy is a great hacker just cause he's on the Internet:

http://www.youtube.com/user/NextGenHacker101#p/a/u/2/SXmv8quf_xM

Cause watching him hack, reminds me of someone...

// Jesper

Yeah right.. again again

".....evenue went down 13%"

Hey, I simply said you were right, Oracle does not have high margins on Hardware they are loosing money IMHO. Get over it. Don't go there if you are not prepared to face the findings.

"....Me, I dont lie and FUD like you do, I accept that IBM has higher TPC-C. I dont deny that, like you do.".

What sensible people like me are trying to hammer into the heads of people like you is that a cluster submission on the TPC-C benchmark is not a system.. it's a CLUSTER. Nobody ever denied that Oracle had the best clustered TPC-C benchmark submission. And that it was also the submission that had the biggest number of transactions...

But as it is a clustered submission.. then every sane semi skilled IT person should know that it's easily beatable by simply clustering together a few largish machines. Like IBM did with a 3 POWER 780'ies.

"...If you need six IBM POWER servers to match one Sun T5440 - how can that be FUD? Just read the white papers you linked to! I never make up things....."

Man ... you are dense and your argumentation is so flawed that it stinks.

Lets try to use your argumentation..

Yeah, the M9000 is twice as fast as the POWER 795.. and that is a fact just read the benchmarks:

POWER 795 specInt_2006rate -> 1440

M9000 specInt_2006rate -> 2590

http://www.spec.org/cpu2006/results/res2009q4/cpu2006-20091012-08891.html

So you need 2 POWER 795 machines to match one M9000, and that is a fact.

Only problem with this argument in the above example is that the POWER 795 used 4 chips out of 64 on the submission and the M9000 uses 64 chips out of 64 chips.

The real POWER 795 submission with 64 chips gives specInt_2006rate -> 11200

http://www.spec.org/cpu2006/results/res2010q3/cpu2006-20100817-12974.html

Kebbabert people are not laughing with you when you do math like that.. they are laughing at you.

And the whole mainframe CPU versus x86. When discussing this with you it's like discussing sex techniques with a virgin. You have absolutely no clue.

// Jesper

Talking about FUD

"Oracle have high margins..."

Yes they have very high margins.. on Software. On the hardware I don't think they are high enough. I don't think that Oracle is actually making money on their server sales. And besides that direct hardware revenues it only accounts for 4% of Oracles revenue.

http://www.oracle.com/us/corporate/investor-relations/financials/q3fy10-detailed-financials-080347.pdf

On page 3:

Oracle made 458 Million USD on direct Hardware revenue + Support. And lets give them 10% of the services revenue that it 10% of 2,797 for a total of... 738 Million USD.

But then comes the expenses related to making that revenue.. and that is when the bad story begins. Again if we look at the above link there are direct expenses and % of other expenses (Lets be real nice and use 4% it's in reality much higher and 10% on services)

Expenses:

Hardware systems products 206 MUSD

Hardware systems support 116 MUSD

Sales and marketing 133 MUSD

Research and development 88 MUSD

General and administrative 25 MUSD

Amortization of intangible assets 55 MUSD

Services (10%) 243 MUSD

-------------------------------------------------------------------------------------

Expenses 866 MUSD

Revenue 738 MUSD

-------------------------------------------------------------------------------------

Loss 128 MUSD

===========

And then comes the restructuring cost and acquisition costs relating to the buy of SUN.

Which for Oracle as a total was a total of 517 MUSD.

Lets just say that 25% of that was related to the SUN's hw business that would then put the Oracle loss on the old SUN HW business at 257 MUSD.

And IRL the actual number is much bigger, cause it's a much more labour intensive job selling hardware than software or rather in Oracle's case software maintenance which is a stunning 56% of all Oracle revenue, so the 4% I used is most likely much higher, at least 8% which would put the loss of over the 500MUSD mark.

Now such a big loss is not something that Larry will put up with in the long run.

This reminds me when I was a Freelance consultant 13 years ago, and advised my client against making Digital Servers the strategic platform for the next 5 years, cause the company was in trouble. I wasn't popular but I was right.

This is most likely the story that IBM sales is telling consulting companies like Accenture,McKinsey, KPMG, CapG etc etc. And that will hurt Oracle, if these 'independent' advices start to put the thumb down on Oracle HW.

"lso talking about "questionable value proposition" - you needed 6 (six) IBM Power P570 servers to match ONE Sun T5440 server in SIEBEL v8 benchmarks. "

Again you keep repeating that story again and again.. it's 2/3 filled up POWER 570 and a 1/8 th filled up POWER6 p570 versus one T5440. Hence your claims are ... false.. and stink of FUD.

And rather than just quoting the now gone BMSEER you should try to look at the facts.

http://www.oracle.com/apps_benchmark/doc/sun-siebel-8-14000-pspp-on-solaris-benchmark-white-paper.pdf

and

http://www.oracle.com/apps_benchmark/doc/IBM_Siebel8_7000_PSPP_On_AIX_POWER6%20Final.pdf

Now the reason why the T5440 is on top here is the response times, which is 4 times faster for the majority of uses on POWER6/POWER5 solution. Which someone who actually worked with RL IT systems, should know.

"is really dog slow. You need 5-10 of the z196 cpus to match one modern x86 cpu."

Now that is just pure FUD, you have no what so ever facts to match that statement. And IMHO it is clearly untrue. And shows that you don't know what you are talking about.

"Earlier, when Oracle had the TPC-C world record, IBM said that because IBM has faster cores, the TPC-C world record still belonged to IBM. If you looked at no 1 on the TPC-C list, you saw Oracle. But still IBM said that they had the record, because they "used faster cores"."

BZZZZZ.. Wrong again.. Oracle had the 'clustered' TPC-C record, not the single Machine. Now they don't have the clustered record, it is also a POWER score.

"Or, the POWER6. You need four POWER6 cpus to...."

You do know that POWER7 have been shipping for a long time don't you ?

Or is the only thing you can do whine FUD, come on grow up learn to research your arguments and do some logical reasoning. It's pathetic.

What records ?

Care to enlighten us ?

And show some comparisons to other platforms ?

// Jesper

Which will mean that ...

If it's not decided if Venus will go in a server yet? Then there is at least half a year of testing and and and .. which will mean that the current Venus will be coming out at the same time as POWER7+.

And seeing how a 8 CPU POWER 780 is faster than a a 64 CPU M9000 on every single benchmark where these two machines have made submissions. Then Venus has to be many many many times faster than SPARC64 VII, just to catch up.

Not going to happen.. Sorry.

// Jesper

Ok, Venus can be found in which Oracle server ?

Just out of curiosity...

// Jesper

It's not that simple.

"There are VERY few work loads where you MUST have a Mainframe."

I don't think there is such a thing as "must". But Again having been involved in many 'porting off legacy platform X' projects, I would say that things aren't as simple as you try to make them look.

Now a mainframe running CICS with PL/1/Cobol exits which might even have a little assembler thrown in for fun, which have been running for 20 years, is not just something you replace like this *SNAP*.

I've seen projects where people have tried to port off "mainframes" (of various types), and where they ended up with 5% of the expected throughput at 5-10 times the response time. Why ?

1) Cause there is a big difference in running interpreted java code and natively complied code.

2) 'Mainframes' might be slowish on certain tasks, but it's not always that part that counts.

3) Don't underestimate 10+ years of tuning and customisation.

4) Often hardware vendors trying to replace a competitor, kind of use benchmark values for their own solution stack and worst case for the opposition.

5) Software stacks tend to bloat up and become slower, as software companies cut cost.

6) Usually the solution that you have to migrate to is some scale out solution that isn't nearly as efficient as a centralized solution.

So be very careful of migrating of 'mainframes', and by mainframes I am not only talking about IBM zSeries, but there are also a lot of other platforms in that category.

To some Wintel guys UNIX machines fall in the same category :)=

// Jesper

Thank you to the SUN marketing depardement.

And we will ever only need one computer per continent right ?

// Jesper

Not a practical problem.

Yes I guess that the T3 then finally puts Linux on SPARC in the grave, as a serious platform.

And 128 threads is more than enough for the sweet spot of Linux. I mean that is a 8 Socket Nehalem-EX box with Hyperthreading enabled, a 16 Socket Itanium and a 4 Socket POWER7 box. Sure with todays virtualization, what this actually mean is that you have a max. virtual machine size of 64 cores on x86, 64 cores on Itanium and 32 cores on POWER7. Which is kind of enough.

And in 2015 we'll all have flying cars and live forever.... yeah... right..

// Jesper