IBM fires Power-powered Penguins at x86's weak spots

Big Blue has not made any huge proclamations to date, but it is not exactly a secret that the people in charge of IBM's Systems and Software Group want the Power7 processor and its follow-ons to grab a larger share of the systems racket. To that end, Big Blue is reviving a Linux-only variant of its Power Systems lineup with …

[Anonymous Coward]

Hmm...

I don't really buy it, propriety RISC Big Iron is a dying beast, the only reason that you would run Power based Linux under VM is if you have AIX to support as well, but these machines can't do that. It will therefore always be cheaper in terms of TCO to run VMware (or such like Hypervisor) with Linux and Windows sharing X86 hardware, probably Proliant. This way you can have a standard platform for the two largest OSes that an Enterprise will run.

[Kebabbert]

Re: Hmm...

"...propriety RISC Big Iron is a dying beast..."

I agree that x86 is catching up faster than anyone anticipated. For instance, the Intel Westmere-EX is only 14% slower than POWER7 in some benches:

http://www.anandtech.com/print/4285

But after Westmere, we have Sandybridge which is even faster than Westmere. And after Sandybridge there is Ivy Bridge which is still faster than Sandybridge. Thus, the Ivy Bridge should close the gap to POWER7, if not surpass. So, the Ivy Bridge Xeons should be close to POWER7 in performance. The Ivy Bridge is 40% faster than Sandybridge - using the same TDP. But Ivy Bridge uses less TDP, and still increases performance compared to Sandy Bridge.

If Intel's 130 watt Xeons were allowed to use much more TDP (in par with POWER7 at... 200(?)) then Xeons would be much faster than POWER7 and surpass POWER7 no doubt. Xeon is more efficient than POWER7 as we can see. If you have unlimited Watt budget you can theoretically reach any performance level you desire. The problem is to do that efficiently.

And where is the late POWER7+? It better be at least 50% faster than POWER7 to keep a healthy distance to x86. Or there will be no point in buying POWER7+ if Intel offers equal performance. Why spend 3x more money for 14% better performance? Why not buy three Intel servers instead of buying one POWER7 server?

.

Regarding this step by IBM: offering Linux on POWER. This only confirms the official statements from IBM, that AIX is going to be killed and replaced by Linux in the future:

http://www.zdnet.co.uk/news/application-development/2003/01/29/ibm-linux-will-replace-aix-2129537/

IBM is slowly preparing for exit. We see that from 2010, IBM even offers Mainframes which are Linux only. Soon Linux everywhere, on every IBM hardware. It all aligns with the official statements from IBM. Everything adds up. This heavy betting on Linux from IBM. Better polish your Linux skills.

[Anonymous Coward]

Re: Hmm...

You are still flogging this 10 year old article which references one comment made by one IBM executive who was not involved with AIX. In the nine years between when this article was published, IBM AIX has moved from third place in Unix servers to a dominant first place position with more than 2-1 the sales of either HP-UX - Itanium or Solaris - Sparc based systems. Why would IBM kill AIX now that they have the lion's share of a shrinking, yet still lucrative market? These are not "official statements" from IBM. They are the response from one IBM Software Group executive (nothing to do with AIX) to a question at one conference a decade ago which were immediately corrected by the AIX executive in the same article!

"Why spend 3x more money for 14% better performance?"

Because Unix systems are about much more than raw CPU power.

[Peter Gathercole]

Re: Hmm...

This 'new' ability to only run Linux is not new. If you have access to a Power 6 or Power 7 system, and look in ASMI or on the HMC (and I presume SDMC and IVM) at the entitlements section, there has been an entitlement for both AIX and IBM I for several years. Linux has been an officially endorsed OS by IBM on PowerPC for at least 7 years (they have had agreements with Slackware and SuSE), and there are official distributions of RedHat and Ubuntu from those companies.

This makes this a re-announcement of an existing policy, probably to remind some existing PowerPC shops that they can stick with Power rather than moving to another processor, even if they are switching OS. I very much doubt that the product announced will significantly differ from other systems that will still run AIX and IBM I.

This does not give any new reinforcement that policy that you bring up in every discussion about PowerPC or AIX. Both AIX and PowerPC will be here for some time still. That link lookes older each time I look at it.

Now. I'm not going to argue with the fact that AIX (along with all proprietary UNIX systems) is on the downward side of the popularity curve, and I do not think that PowerPC development is in a good place at the moment. It's expensive to build new generations of any processor, and I think that IBM is really thinking hard about what to do with the PowerPC line, at least in high end servers. Sometimes I wonder whether IBM really wants to remain in the hardware business at all (products that have been sold include their printer division, their storage division, the desktop and laptop PC business, and most recently their ATM and PoS business).

This policy may extend to their server systems as well. Power7+ is late according to previous product roadmaps, and there is strangely very little pre-announcement information about Power8. IBM has also made statements that their previously loss-leading HPC work has to become more commercial (probably one of the reasons why IBM pulled out of Blue Waters), which means that future generations of IH HPC systems are at risk.

But one of the effects of there being a creditably competitor to Intel processors is that makes Intel aggressively pursue new processors. Once they are only competing with themselves (remember, AMD need chipfabs like IBM to create their products, because they cannot fabricate processors themselves), then the rate of product development will slow significantly, as Intel would want to get more return on their investment.

I am really not looking forward to a point where the only processor game in town is X86 derived, and that is looking like a possibility within a decade unless ARM moves upward.

[Kebabbert]

Re: Hmm...

"...This 'new' ability to only run Linux is not new...."

No, I agree with you on this. But the new thing is that IBM now offers Linux ONLY servers. Earlier, you could run Linux in different ways on POWER, for instance virtualize Linux. That is not new. The new thing is that there are Linux only POWER servers now. Why is that? That is new.

.

.

"...Now. I'm not going to argue with the fact that AIX (along with all proprietary UNIX systems) is on the downward side of the popularity curve, and I do not think that PowerPC development is in a good place at the moment.... Sometimes I wonder whether IBM really wants to remain in the hardware business at all (products that have been sold include their printer division, their storage division, the desktop and laptop PC business, and most recently their ATM and PoS business)....

I am really not looking forward to a point where the only processor game in town is X86 derived, and that is looking like a possibility within a decade unless ARM moves upward..."

It seems that you agree that the trend shows that x86 is fast catching up on other cpus. The future looks grim for other cpus, they will be eaten. Only niche cpus can survive. Intel does not venture into every niche (yet).

Myself dont like x86, it is too buggy and bloated. But analyze the trend and think a bit. The future is clear. POWER7 is going out, IBM is preparing for exit. POWER7+ is late, maybe IBM has problems with performance. IBM needs to beat x86 severely. Just little will not do. And about POWER8 we hear nothing. Is that canned? What happens? IBM should say something to calmd down worried users? Another thing is that pessimitstic people seems to be right and correct.

[Anonymous Coward]

Re: Hmm...

"Myself dont like x86, it is too buggy and bloated. But analyze the trend and think a bit. The future is clear. POWER7 is going out, IBM is preparing for exit. POWER7+ is late, maybe IBM has problems with performance. IBM needs to beat x86 severely. Just little will not do. And about POWER8 we hear nothing. Is that canned? What happens? IBM should say something to calmd down worried users? Another thing is that pessimitstic people seems to be right and correct."

Power 7+ is not late. Power 7 came out in 2010. Power 7+ will be out this year. The + always comes out in the third year of the cycle. Power 6 was released in 2007. Power 6+ was released in 2009. Power 8 will be out next year.

As far as calming down worried users.... As Power has never had this large a percentage of the Unix market share in the history of Power servers, it doesn't seem like there are too many people wringing their hands.

[Kebabbert]

Re: Hmm...

"....As far as calming down worried users.... As Power has never had this large a percentage of the Unix market share in the history of Power servers, it doesn't seem like there are too many people wringing their hands...."

Something similar said the Dinosaurs just before they were wiped out. Everything looked good on the surface, but a deeper analysis shows there are fundamental problems.

For the first time, x86 is seriously threatening POWER: performance and RAS. No one can just shove this fact under the carpet and pretend it is not true. IBM needs to do something radical. A cautious General prepares for the worst case. IBM should do that too. Maybe start to diversify and not rely to heavily on proprietary (soon inferior) IBM tech such as POWER and Mainframes. Maybe start to offer Linux on IBM prop tech. Maybe preparing for exit.

I heard that POWER7+ is late, but it is not late, you say? Can someone confirm if it is late or not?

.

You all like when I show credible links from different IBM executives, and analyze the trend, and talk about the dark looking future of AIX and POWER, right? Again and again, right? How do you feel? Warm and cosy? How do you think we felt when you started spreading the FUD about HP and Sun and Oracle many years ago? What comes around, goes around. If you IBM supporters hadn't FUDed so aggressively earlier, no one would be interested in rubbing this in your skin. Suit yourself.

Admit it: x86 is catching up on POWER, and IBM is betting more and more on Linux. IBM starts to offer, for the first time, cheap POWER Linux only servers? Bad sign. What does that mean? That POWER has a rosy future? Hardly. IBM only does high margin business. I have talked about it for years: IBM needs to cut the price of POWER, as cheap x86 catches up in terms of performance. This is happening. Now. Today. My predictions are starting to come true. The next step is that when POWER prices are low enough, IBM will kill POWER and switch to x86. Probably, IBM will continue with cheap PowerPC cpus, but expensive POWER cpus will be killed.

[Anonymous Coward]

Re: Hmm...

No one is saying that x86 is not catching up with Power. I am sure IBM is well aware. It is obviously much closer than it was 5-10 years ago, but there is a big gap between x86 catching up to Power and assuming that IBM will then drop Power. Unix servers started getting close to and replacing mainframes 30 years ago, but IBM did not drop mainframe. There will still be people that want a managed Unix operating system which is engineered to task with the physical hardware. Will that number of people be the same number as in 1996? Obviously not, but IBM can still profitability serve that niche market as they have done when the mainframes and minicomputer eras passed. Unix will be a niche, high-end solution, but I don't think every bank, government, hospital, etc wants to put all of their mission critical applications on x86 and semi-tamed Linux with a la carte-ing and integrating hypervisor, systems management, hardware. Unix (which increasingly means AIX) is Mac in a Windows PC world. Mac is probably never going to touch Windows PC market share, but there is still a market for people that want the integrated system despite being able to perform the same tasks on Wintel.

"You all like when I show credible links from different IBM executives, and analyze the trend, and talk about the dark looking future of AIX and POWER, right? Again and again, right? How do you feel? Warm and cosy? How do you think we felt when you started spreading the FUD about HP and Sun and Oracle many years ago?"

It was not a credible link that is at all indicative of the future direction of AIX or Power. It is 10 years old and the time that has passed is evidence in and of itself that the statement is not correct. It was one off hand remark to a question at a conference from an IBM executive that wasn't at all involved with the product.

I don't think you can compare IBM to Sun or HP's Unix systems. Sun was in deep financial trouble and no one needed FUD to create uncertainty... their financial statements created the uncertainty. The FUD against HP has been coming from Oracle, not IBM.

[Jesper Frimann]

Re: Hmm...

Well IMHO and as TPM also described in this article http://www.theregister.co.uk/2011/10/14/ibm_power7_plus_redux/, POWER7+ is 'late'.

When I write 'late' I don't mean late as in what you Keb would like it to be, problems.. bugs etc.

see for example Charlie D's article here: http://semiaccurate.com/2012/03/21/ibm-power-7-spotted-and-it-is-a-monster/ , but 'late' as in they didn't need to release a new Processor.

If sales are strong and the competition isn't really releasing anything that will affect you, why release a new product ?

Again the Q4 launch of the 'C' generation of machines was perhaps not a big one to the untrained eye. I mean 200MHz more on the processors on the POWER 770 ? What's the big deal?

But there were more under the covers, the whole Internal IO system got a much needed upgrade, doubling the amount of bandwidth and number of IO transactions that the 'internal' IO system is capable off. Furthermore the amount of memory supported doubled.

And there were also some other smaller things.

And with regards to the 'new' Linux only servers. Well you obvious haven't been following the UNIX marked for a very long time have you ?

I mean this is not the first time that IBM have tried this approach, they did it with the OpenPower servers back in 2005. So again nothing new.

// Jesper

[Peter Gathercole]

@Kebabbert Re: Hmm...

I was not clear about entitlement in my earlier post. There were Linux only Power 5 systems back in 2005 or so. What I was trying to say was that they were the same systems with the AIX and the IBM i entitlements turned off. They were also significantly cheaper, and also made it easier to use non-IBM branded disks.

My views about proprietary UNIX being on the downward curve has not changed. I have felt this way for most of the last decade. I still see Power having a place for many years to come.

Intel becoming predominant is much more about them having volume and critical mass in the processor market than speed or technology. PowerPC is still a relatively well architected processor, but for many companies developing products, it makes sense for them to use what is fast becoming a commodity product (Intel) rather than something that they have to put significant design effort into. A high-end PowerPC SoC would be interesting, but I don't think IBM would be interested in creating one of these for the server market.

[Anonymous Coward]

Re: Hmm...

IBM has never been interested in making "hardware" (empty boxes). They have always been a systems company (hardware integrated with software). They have dropped out of PCs and POS because there is not much they can do to make them substantially better than any other provider as they all run Windows and have virtually the same specs. Not much value to add in making a better looking case. Printers are just a dying, or at least rapidly declining, business. They are interested in systems, but they don't want to be a volume box pusher.

[Davidoff]

The PowerPC 6XX processors did run Windows NT 3.51 for a brief somewhat shining moment

The moment hasn't been *that* brief. PPC support was in NT 3.5, NT 3.51 and also NT 4, and IBM did also sell PowerPC-based PCs for a few years.

Of course, the lack of native Win/PPC software and the absence of a x86 emulation layer (which with DEC's FX!32 was available for the Alpha AXP variant of NT) resulted in rather disappointing sales.

[Beachrider]

There are hardware differences between Power and x86...

If IBM can establish a price advantage, it is not impossible to market LINUX/PowerVM configurations against LINUX/VMWare configuration. The Power architecture has allowed for more-cache-per-socket since 2010 (Power 7) while establishing much-higher GHz on these high-cache engines. The ability to work 10GbE and 8GbFC is also stronger than Intel, because of the faster I/O buses. Intel can approach these speeds with special chippery, but that is when they get much more expensive.

It is all about the Benjamins...

[Anonymous Coward]

Re: There are hardware differences between Power and x86...

It would have to be a pretty good price advantage to cancel out the disadvantage of having Linux and Windows VMs on separate hypervisors, requiring separate virtualisation teams or multi-skilling your virtualisation team. There isn't any realistic problem with the speed of IO on top end proliant servers, a DL7XX makes a very good data mover, to the extent that Symantec NetBackup people are targeting new versions of their product at Linux or Windows X86_64, depending if you're a Windows or UNIX shop. They still produce proprietary RISC UNIX versions, but are seeing less and less interest in it.

[Stevie]

Bah!

I've been sitting in some meetings in which the option to boot Linux has been raised.

For the life of me I cannot see the logic in buying a Power computer and booting a Linux O/S on it. You get none of the AIX features that make the Power frames so attractive in the first place, and you pay lip service only to saving bux by running an inexpensive open O/S on Gold-Plated Bog-Roll Holder fitted hardware that cost an arm, a leg and several internal organs to acquire.

At the risk of opening a can of double-entendre worms re: contract bidding, I have to ask: "where's the payoff?"

[sprag]

Re: Bah!

Well, if the price/performance is better than x86 then it might be a good fit for high-cpu workloads (video processing, for example). AIX used to suck severely when it came to building open source tools (back in the 5.x days. ffmpeg was nearly impossible to build) so if you're using the linux stack and various open source tools there might be an advantage.

[Anonymous Coward]

@All x86 Fans

IBM Power has always been a leader in single-threaded performance (maybe even in multi-threaded). Also, IBM hardware is usually rock-solid, highly reliable and extremely well-designed. The "corporate knowledge" of more than 100 years in computing technology shows...

So if they can actually price this hardware competitively, I am sure many Linux users would love to use Power CPUs. I only doubt that they can even get close to intel-based system prices.

Also, reliability is not as important for modern architectures, which consist out of dozens of nodes and have fail-over built into the software architecture. Google is a good example of that.

[Anonymous Coward]

Re: @All x86 Fans

I agree. I am not sure that this system will take off for the reasons mentioned, but it is not like CPU's reached perfection with the 30 year old x86 CISC architecture. Intel has squeezed an amazing amount out of it, but every RISC option should not be rejected out of hand because Microsoft does not want to create a port.

[Anonymous Coward]

Re: @All x86 Fans re "reliability is not as important "

"reliability is not as important for modern architectures, which consist out of dozens of nodes and have fail-over built into the software architecture. Google is a good example of that."

Any sense the rest of your reply may have made is devalued by your Google reference.

Google are not a good example of delivering reliable software, as (e.g.) any Virgin/Gmail user will know.

That's all for now.

[Anonymous Coward]

Re: @All x86 Fans

"Also, reliability is not as important for modern architectures, which consist out of dozens of nodes and have fail-over built into the software architecture. Google is a good example of that."

I agree that with a massive Hadoop style node based processing system (e.g. Google's architecture) you can get very high levels of reliability though massive parallelism. The problem is that every meat producer and retailer doesn't have 200 MIT and Stanford engineers around to build a crazy scale Hadoop architecture and keep the whole contraption up and running. A few very solid scale-up boxes makes sense for those companies. Everyone marvels at Google and fb's roll your own infrastructure because they have achieved nearly mainframe levels of up time with cheap x86 servers, but you have to be crazy smart to build and maintain your own supercomputer architecture. Their architecture is very inexpensive in capital costs and very expensive in operational (staffing) costs. I often wonder if they would not have been better off financially buying big iron (like Apple) and letting the enterprise computer companies build the enterprise computers while they focus on their web services. As the recent Gmail outage demonstrated, it is pretty hard for the smartest people in the world to keep it running.

[Don Mitchell]

IBM once had 70% hardware market share

What is this new devilry? Something from Microsoft?

No. It is IBM, a demon of the ancient world.

This foe is beyond any of you...run!

[-tim]

Small Iron?

I'll buy Power. If its reasonable. It turns out that the 8080 and its long line of friends love to run code on the stack and most RISCs don't want to even if they could. I want a moderately priced 1RU shallow appliance with 2 3.5" disks that I can slap a secure OS on.

[JimTO]

Linux-only mainframes

You make a statement at the end of this article "IBM has also dabbled off and on with Linux-only mainframes, the last of which were announced back in December 2009." According to IBM about 20% of the installed mainframe capacity is running Linux. The most recent mainframes for Linux are the zEnterprise 196 (announced in July 2010) and the zEnterprise 114 (announced in July 2011). Any mainframe can be Linux only (i.e. all Integrated Facility for Linux processors).

[tpm]

Re: Linux-only mainframes

Yes, but these "official" Linux only mainframes had much lower prices, just like these "official" Power-Linux boxes do.

[Matt Bryant]

Re: Re: Linux-only mainframes

"Yes, but these "official" Linux only mainframes had much lower prices, just like these "official" Power-Linux boxes do."

Which is why I now feel extra sorry for IBM mainframe customers. Up until now, they were just being gouged to subsidise the the IBM Linux mainframes and the AIX servers, from now they're going to be severely rooked to pay for what is going to have to be massively subsidised Linux-Power servers. Why? Simples! Three examples why are the hp DL160 Gen8 (same number of cores, much more RAM, more disks, and all in 1U), the hp DL360p Gen8 (same number of cores, much, much more RAM, and waaay more disks, all still in 1U), and even the old hp DL385 G7 (2 x 12-cores, again oodles more RAM, more disks, all in 2U). And, IIRC, the Gen8 version fo the DL385 is going to have the 20-core Opterons this year. The DL160 starts at just $1769 list, before any discounts. IBM trying to put Power up against Linux on x64 is an instant fail. Hard luck, mainframers!

[tpm]

Re: Re: Re: Linux-only mainframes

Just like the AS/400 shops paid for IBM to attack Sun and HP in the Unix racket.

[Anonymous Coward]

Re: Linux-only mainframes

True, you could also add how any company funds any new product. Oracle's database customers paid for PeopleSoft and Siebel and Sun. HP's ink and legacy OS customers paid for Autonomy, Voodoo, Palm. General Mills' Cheerios customers paid for Lucky Charms. There is nothing insidious about it. Companies pay for new products with money they have made selling previous products.

[Jesper Frimann]

Re: Linux-only mainframes

"IBM trying to put Power up against Linux on x64 is an instant fail. Hard luck, mainframers!"

There is no doubt that with current volumes, then the traditional UNIX hardware cannot compete with x86 when it comes to TCA for the Hardware.

But when you start to put virtualization layer, takes utilization into account etc etc. then things even out. On our local 'cloud like' offering, it is actually cheaper to get a Virtual AIX server than a Virtual Windows server of equal capacity.

And it's cheap shit 2 socket x86 versus 8 socket 4TB RAM POWER servers.

// Jesper

[John Savard]

Undermining

Whatever they do, they must not undermine... sales of machines which have a higher price for comparable performance. Whether this is aimed at AS/400 or System i customers, or System z customers, basically it's a tax on people who are stuck with legacy software from IBM.

They need to stop trying to maximize profits at their customers' expense. A miffed customer is not a happy customer. Instead, work on making the price comparable across all the systems - and then blow x86 out of the water by pointing out that these equally-priced systems can also run IBM's flagship mainframe software.

Because x86 can run Windows, and so pitting a Linux-only system against that is selling less for an equal amount of money. Offering System i or System z software would be something to balance against Windows.

That way, IBM would be competing. This way, it's offering semi-competitive Linux systems... and making money off of existing legacy customers who are unlikely to be joined by new customers. That is not a strategy for growth.

[Anonymous Coward]

Re: Undermining

"Whatever they do, they must not undermine... sales of machines which have a higher price for comparable performance. Whether this is aimed at AS/400 or System i customers, or System z customers, basically it's a tax on people who are stuck with legacy software from IBM [etc]."

Indeed.

This was the challenge DEC had with Alpha and in particular with Alpha and VMS. There were Alpha chips and systems aggressively priced (not necessarily with DEC chips or badges) because they were designed to not run VMS. There were Alpha chips and systems traditionally priced (ie apparently expensive) because they could run VMS.

Anyone remember how that worked out for them? IBM have rather bigger reserves and better diversity.

[Jesper Frimann]

Re: Undermining

@AC

.....

"Anyone remember how that worked out for them? IBM have rather bigger reserves and better diversity."

Well the main difference I remember, between DEC and the other *NIX vendors, was that the DEC sales people had tailor made suits with gold cufflinks, just like the IBM mainframe sales people.

Now the HP/SUN and IBM *NIX sales people didn't.

// Jesper

[Anonymous Coward]

More Expensive, Binary Incompatible, Vendor Lock-In

Tell me why this is better than x86?

x86 servers were originally about cost. They were much cheaper than proprietary minicomputers of the era.

The Linux OS originally was about cost. Linux was a "free" or at a least cheap alternative to UNIX, and later Windows.

Today, x86 is enterprise class, and Linux is enterprise class. But Linux/x86 is still cheap because as a customer, I can compete server vendors against each other (HP, IBM, Dell), or even go whitebox. I can compete x86 hypervisors against each other (VMWare, KVM, Xen), or use a self-supported open source hypervisor. I can compete Linux distros against each other (RHEL, SLES, OEL) or roll my own.

IBM Power servers (no choice) plus IBM PowerVM (no choice) plus fewer Power Linux distros (less choice) equals higher cost.

All of this leads to this platform being absolute overkill for open source web and cloud software like the LAMP stack, Tomcat, OpenStack, or CloudStack.

Also, Linux on Power is binary incompatible with Linux on x86. So that cuts the ISV application ecosystem significantly.

Finally, there are ISV licensing costs on Power7 vs. x86. Don't even think about running Oracle software such as their database or WebLogic on these machines.

[Anonymous Coward]

Re: More Expensive, Binary Incompatible, Vendor Lock-In

If you have a pure Open Source Stack, Linux on Power will feel almost exactly as Linux on x86. But it will be very, very fast and efficient. I welcome diversity, as do others.

[Jesper Frimann]

Re: More Expensive, Binary Incompatible, Vendor Lock-In

@AC

"Finally, there are ISV licensing costs on Power7 vs. x86. Don't even think about running Oracle software such as their database or WebLogic on these machines."

Well it depends, it's not a clear cut case, as you try to portrait it.

If you use per user license, then your argument is not valid.

If you are running on only 1 core.. then the price is the same.

And then there's the whole virtualization angle, where for example running Oracle software on VMware is a nogo, but on POWER it's more or less the default.

But if you are your normal average shop, that runs a few Oracle servers on bare metal where you pay for processor licenses, sure.. then it's normally more expensive to use POWER rather than x86.

// jesper