In the computer racket, it has always been much easier to get new hardware out the door than the software to take advantage of it fully. And so it is with IBM's Power7-based Power Systems machines and its AIX variant of Unix, which is slated for a 7.1 update in the autumn. AIX 7.1 is expected more or less concurrently with the …
The Most Interesting Question
..is what Matt Bryant has to say about these new machines.
RE: The Most Interesting Question
Not much to say, really. A good article, balanced between what is already known and a bit of conjecture, and not the usual IBM marketting fluff. I'm not sure why AIX development will have slowed down - has it, I thought it was ploughing along at a reasonable pace? Anyway, I agree with the poster that said arranging downtime between many departments can be a real pain, but my usual approach is to get the biggest bully possible from the boardroom to TELL the departments when they need to arrange downtime for. And the way to get the bullies onside is to translate the message into something painful they can understand, by highlight the financial costs of any outage caused by a failure if you don't patch. Works a treat!
Bit concerned the IBM mid-range 7x0 modular servers have to lose half their guts in order to link up - only eight sockets in 16U doesn't sound either dense or a cheap option, not compared to the Tukzilla blades or even Sun's old M-series kit (even an old M5000 gets eight sockets in 10U, so if they ever get an octo-core SPARC64 it would be interesting).
Bit surprised there is very little mention of Linux in the article, is it not on the IBM plan for the new P7 kit?
always a FUDster
The article was about AIX 7....why surprised about no mention of Linux?
Considering RedHat and Microsoft have ended support for Itanium its just a little of throwing stones from you glass house.
cheers from the UK
RE: always a FUDster
"The article was about AIX 7...." The article covered a lot of ground, including mentioning OS/400 and i, so why not Linux?
"....Considering RedHat and Microsoft have ended support for Itanium...." Always makes me laugh when the Power fanbois start regurgitating that chestnut considering that Windows has never run on Power in any form! A little defensive, aren't we? Could it be because, as was mentioned in the article, "....Power Systems sales continue to slump in parallel with System z mainframe sales...."?
Seperate kernels are a GOOD thing
You say that seperate kernels that require seperate patching are "a pain".
Try patching a Solaris Zoned machine, with maybe 10 zones, each used by a different Business unit, where the useage demographics mean you find it difficult to get agreement on outages. Then you will know pain!
Far easier when each you can obtain seperate outage for each instance.
Easier patching usually means more frequent patching.
Slowed down development
"The pace of change of AIX development has slowed, as it has for both Hewlett-Packard's AIX and now Oracle's Solaris."
Well, I'm actively watching Solaris development and I haven't noticed that it slowed down rather quite the opposite.
Then thanks to Solaris running on x86 platform as well in the long term it will be rather Solaris than AIX to be the last standing one.
AIX has not slowed
The Power platform is still growing and gaining gobs of marketshare from HP and Sun.
Considering Sun hardware was down 40% YTY in 1Q it does not look like Solaris is healthy or justifies investment. Oracle also just pulled HP's x86 support for Solaris and Larry has dumped all AMD boxes and will only be doing x86 appliance which I think they "sold" 25 in 1Q. Amazing what you can "sell for free" at an ELA renewal so you can go the the sales club in Hawaii.
Cheers from the UK
"The pace of change of AIX development has slowed, as it has for both Hewlett-Packard's AIX and now Oracle's Solaris."
This is utterly wrong. IBM shifts focus from AIX to Linux, so I understand if AIX development is slowing down. I dont know development about HP-UX nor OpenVMS (which is arguably the most stable OS out there).
But Solaris is developing rapidly. ZFS, DTrace, COMSTAR, CrossBow, etc. In fact, Solaris unique tech has won several awards and several OSes are porting or copying Solaris tech. For instance, DTrace is copied by AIX and it is called Probevue. I remember when IBM said that Niagara approach is bad, it is better with very few cores, but very high clocked. Now IBM has many slower clocked cores, just like Niagara. IBM has shifted from high clock speed to more cores - so Niagara approach is not bad anymore now when IBM is doing the same thing.
Also, an report sponsored by HP, showed that Solaris is the top Enterprise OS.
Maybe AIX is dying in favour of Linux, but Solaris is not. Solaris is ported to x86, which very soon will be the fastest CPU on earth. Next year we will see 32nm version of Nehalem-EX and it will be even faster. And Intel Sandybridge coming soon, will be much faster. And then we will see what AMD has to offer. x86 is cheap. And fast.
IBM POWER cpus has slow development, just like AIX. Solaris has utilized many threads for a long time now, just what AIX is trying to do now.
Both Intel and Solaris is developing fast. And, Solaris runs on Intel, which gives good single threaded performance. On the other hand, if you need many light threads, you run Solaris on Niagara. Whatever work load you have, Solaris can give you what you need because it runs on several CPU architectures.
AIX needs to be ported to x86, otherwise there will be no reason to pay several times more, for just 10% higher performance, or equal performance. I predict those OS that are not ported to x86, will die.
Instead of generic truths, please get yourself access to a Power Box. It will change your expectations of what a "fast" CPU is. Or maybe look at the current www.spec.org results, but that is not as good as running a program yourself.
With the Power architecture, IBM demonstrates that they are still the "top dog" in computers. How long they can sustain this, I don't know, but for the time being they are the undisputed king of the "fast CPU kingdom".
I agree that IBM has the fastest CPUs right now. Just look at the benchmarks. No, I do not reject them just like Jesper Frimann does, when SPARC wins "no, the benchmarks are cherrypicked by Sun. No Sun has crafted these spec_int2006 benchmarks. etc etc". Hilarious. When SPARC wins, all benchmarks are immediately rejected. When POWER wins the same benchmarks, they are good and to be trusted. Really credible guy Jesper is, especially when he claims "POWER6 is faster than Nehalem despite you need four 5GHz POWER6 to match two 2.93GHz Nehalem". I wouldnt trust much of that.
Anyway, POWER7 is fastest for now, as we all know. But the question is, how much more are you prepared to pay for that extra performance? Intel Nehalem-EX 45nm achieves 70% of the performance today, for a fraction of the price. Next year, Nehalem-EX will come in 32nm version which will be even faster. And soon Intel will release the Sandybridge architecture which is much faster than Nehalem - "the biggest step ever". In just two years, Intel CPUs will be on par with POWER, or faster - for a fraction of the price. Intels development pace is extremely fast, just like Solaris.
And IBMs development is slowing. POWER is too expensive, and the release cycles are too long. I fail to see how any other CPU architecture than x86 can survive in the long run. It will be fastest, and cheapest. And buy several cheap, fast servers and you have good redundancy. AIX must be ported to x86, or it will surely die in the long run. But as IBM is slowly shifting focus to Linux, AIX may never be ported to x86. I dont know how easy it would be to port AIX to x86? Is AIX plattform independent coded so it is easy to port the CPU depedent parts? In some aspects AIX seems antique, so maybe not? I read that on an IBM forum, they say it takes 90 minutes to boot IBM P595. I mean, AIX clearly had problems earlier to scale to many CPUs and cores. So much for IBMers here talking about "POWER scales better than SPARC". Solaris has scaled well for ages. What is the point of having many CPUs if the OS can not use them?
get your head out of the sand sun lover
10%? look at the sap.com/benchmark results you will see Power7 is 2.3X Nehalem EX in performance. If you look at TPC-C you will see Power7 is less per transaction than Nehalem EX.
As far as SPARC its the only hardware platform in the market that is going up in price thanks to Larry cash cow the hardware ellison.
"But Solaris is developing rapidly. ZFS, DTrace," wow I think those came out about 5 years ago and in reality are not that exciting....the recent purge of Sun employees included about 22% of the Solaris development team, so don't expect much and OpenSolaris is dead, gone, history.
You forget that a the 32nm follow on to Nehalem-EP, Westmere-EP is already here. And it just keeps up with Nehalem-EP in the per core throughput.
And if you compare Nehalem-EP with Nehalem-EX it's not like it's a huge improvement.
Lets take for example the latest SAP-2Tier benchmark
Nehalem-EX best 4 socket result:
10450 Users@4 Chips and 32 Cores@2.26GHz
Nehalem-EP best 2 socket result
3800 Users@2 Chips and 8 Cores@2.93GHz
Westmere-EP best 2 socket result
5100 Users@2 Chips and 12 Cores@3.33GHz
Nehalem-EP -> 475 Users per core
Nehalem-EX -> 327 Users per core
Westmere-EP-> 425 Users per core
So basically The core performance of Nehalem-EX is 69% of that of Nehalem-EP
Now that isn't good.
And the Westmere-EP core performance is 89% of that of Nehalem-EP.
And that isn't good either, with 14% more GHz.
Nehalem-EP -> 1900 Users per Chip
Nehalem-EX -> 2612 Users per Chip
Westmere-EP-> 2550 Users per Chip
So a Nehalem-EX chip is 1,38 times faster than Nehalem-EP chip.
That is an impressive 38% extra Throughput on x2 the number of cores.
And a Westmer-EP chip is 1.3 times faster than a Nehalem-EP chip
That is an impressive 30% times faster with 50% more cores and 14% more Frequency.
Nehalem-EP -> 162 Users per GHz/core
Nehalem-EX -> 145 Users per GHz/core
Westmere-EP-> 128 Users per GHz/core
So basically Westmere only manages 79% of the users per Ghz/Core. That's not impressive, that's... well.. a Westmere-EX doesn't really look that impressive in this light.
And POWER is slowing down in development ? Keep back on the mushrooms Keb.
If you compare POWER6 to POWER7 then the Per core users have increased with 23%.
The Per chip number of users have increased with 393%, and the per user per GHz/core has increased with 60%.
So basically you are just putting out FUD without knowing what you are talking about.
And sorry for my harsh words but it has been a really bad day.
...And the Power solution costs how many times more than the x64 solution, and has to come with how much more expensive an IBM support contract, and how much IBM Global Screwups consultancy to actually even get working, let alone get close to the brochure figures?
So you think that ZFS and DTrace is not exiting and old? Well, IBM thinks that DTrace is exiting, they are copying it now and calling it "IBM Probevue". And Dtrace has won awards in Wall Street Journal. That finance news paper awarded DTrace, that means DTrace is very important. Which also IBM thinks, but not you. And ZFS and Dtrace is still fast developed. For instance, ZFS is just getting Deduplication which is very hot, but you didnt knew that. Or you knew, but didnt consider it as hot because it is Sun tech. But I bet you will praise Probevue as the best thing since sliced bread.
Westmere-EP is a nice CPU, but it is much cheaper than a POWER7 machine. I also cited Intel who claims that Sandybridge CPU "will be the biggest step ever". The next generation CPU is already soon here. Where is POWER8? Not visible. You can reject my claims as FUD as you normally do, but I say again: Intel development pace is much faster than IBM. Quite soon x86 will be the fastest and cheapest CPU. Yes, I know this is FUD to you. But there are people who dont consider my claims as FUD.
"And POWER is slowing down in development ? Keep back on the mushrooms Keb"...Seriously, could you just think a little bit? IBM has an officially outspoken strategy to migrate from AIX to Linux. AIX will die. Why would IBM continue to develop a shit load expensive platform as POWER when AIX will soon die? Linux on x86 will soon be faster than POWER and much cheaper. IBM is cutting POWER and AIX development because Linux on x86 is the future - says IBM.
"So basically you are just putting out FUD without knowing what you are talking about".... Jesus. Everyone knows that IBM is migrating from AIX to Linux. It follows that POWER will also be migrated to x86. Just think a bit. It is not FUD. I know that IBMers FUD a lot, which we have seen in the Itanium and SPARC articles. But I do not FUD, I can always back up my claims with white papers, benchmarks, etc. On the other hand, you FUD a lot. For instance, you claimed earlier that 5GHz POWER6 is faster than 2.93GHz Nehalem on TPC, even though you need four POWER6 CPUs to match two Nehalem. How can POWER6 be faster then? Now this claim is obviously wrong, but I bet there are many other claims that is not obviously false. I advice everyone to be cautious to Jesper's claims.
"And sorry for my harsh words but it has been a really bad day."... No problem? Heh? You have never earlier apologized for this standard behaviour of yours. Always accusing me for FUD, despite I am always able to back up my claims with white papers, benchmarks, etc. Which you reject. "They are cherry picked by Sun" - funny, especially when IBM first chose that benchmark and then Sun published their superior result.
Have a break, have a kit kat....
take a moment guys, this slanging match is starting to wear thin even if it does give us some laughs at times.
Personally, I don't see AIX dying, POWER will continue to grow until there is no longer any business requirement for it, Nehalem is proving to be quite-a-lot-of-what-was-claimed-it-would-be, Linux is showing its strength on the commodity x86 platform and lastly, when did anyone ever say anything other than the "next blah-blah-blah will be the BIGGEST STEP EVER"?
I'm confident the aix division is deep in the red. Why not ditch it and use linux instead. linux scales to 64TB and 2048cores with the proper kernel patches. Should only be to add some hardware drivers for the various power architectures and compilers. Linux runs well through powerVM so it should not be that much work. And you get the rest of the ecosystem for free.
RE: why aix
"......Why not ditch it and use linux instead......" Because then Global Screwups wouldn't be able to fold customers over a barrel and rodger them senseless at great profit.
- Comment Renewable energy 'simply WON'T WORK': Top Google engineers
- Game Theory Dragon Age Inquisition: Our chief weapons are...
- 'How a censorious and moralistic blogger ruined my evening'
- Leaked screenshots show next Windows kernel to be a perfect 10
- Amazon warming up 'cheapo web video' cannon to SINK Netflix