TR Forums

Chrispy_ wrote:

slowriot wrote:

You won't find ESX or Hyper-V running on Power hardware. IBM has their PowerVM suite of virtualization technologies which tightly integrate with the Power processors. You can run several of IBM's own OSes (AIX, i aka OS/400, etc) and several flavors of Linux (RHEL, Ubuntu, SLES).

Does this mean they can't run Microsoft servers as VM's, or just that they're unsuitable from a value perspective?

They can't.

Our Power 7 720 box is controlled by a HMC (which is like the hypervisor console). We used to have 2 as/400 boxes, one for Production and one smaller one for Development. Now the 720 server is just a virtual host, so one box running virtual prod and dev OSes. They also no longer call them AS/400s but we still refer to it as one, the OS is still the same, just called IBM i/OS 7.1.

IBM DOES sell a "Pure" blade chassis system, that you can put a power pc mainframe blade in (I think it takes 2 blade slots), then in the other slots you can put Xeon E5 based IBM servers in, running VMware or whatever.

Chrispy_ wrote:

Ta. That makes a bit more sense.

Power8 is completely useless to me/my companies directly, but will impact the market in a way that might drive Intel/Microsoft prices down eventually.

Yes. Power is absolutely completely useless to your customer unless they are are going to buy a mainframe, build out an IBM HPC cluster, or build an embedded appliance. The last two are iffy since Intel and Nvidia have a large presence in the HPC space, and ARM is the dominant architecture in the embedded space.

Probably not, no. The bulk of the server market is x86, and x86 is the path of least resistance. MS has the advantage of vendor lockin, and they can just port their stack over to whatever ISA they want.

Flatland_Spider wrote:

... and ARM is the dominant architecture in the embedded space.

...except military/aviation, where POWER is still a large (possibly even majority?) percentage of the market. This may finally be changing though, as the ubiquity and service history of ARM in consumer/commercial applications gives the safety and security certification authorities more confidence in its reliability.

Flatland_Spider wrote:

Probably not, no. The bulk of the server market is x86, and x86 is the path of least resistance. MS has the advantage of vendor lockin, and they can just port their stack over to whatever ISA they want.

But other than the now-abandoned MIPS and Itanium ports, they have chosen not to do so. (I'm not counting the mobile stuff, as that is not relevant to servers.)

TwistedKestrel wrote:

Wow, that thing's a monster!

TDP of 250 watts?! Would like to see the chassis that this gets deployed in, hah!

Per the data sheets (registration required)] it is actually 190W. Chances are that IBM's system boards will be over engineered a bit to allow for 250W power consumption. IBM may opt to do dual die packages down the line like they did with POWER7/POWER7+.

Chrispy_ wrote:

slowriot wrote:

You won't find ESX or Hyper-V running on Power hardware. IBM has their PowerVM suite of virtualization technologies which tightly integrate with the Power processors. You can run several of IBM's own OSes (AIX, i aka OS/400, etc) and several flavors of Linux (RHEL, Ubuntu, SLES).

Does this mean they can't run Microsoft servers as VM's, or just that they're unsuitable from a value perspective?

Unless you can find an old version of Windows NT for PowerPC, the POWER8 won't run any commercially released version of Windows.* x86 uses an entirely different instruction set so it simply will not work. However, I believe there are tools that enable the management of both hypervisors from one central location.

IBM also has a blade center expansion chassis for their Z series mainframes. These do have centralized management for VM's across System Z, PowerPC and x86 platforms. In addition, a Linux partition on a System Z host can run x86 and POWER applications using an additional blade as a coprocessor. You still need an x86 and/or POWER chip in that mainframe for x86 and/or POWER applications to work.

*Microsoft does have a port of the NT kernel for PowerPC. It is used for the Xbox 360 which has a triple PowerPC core chip. At one time early in the XBox 360 development, they also had an incomplete port of Windows running on Apple's PowerMac G5. These were internal development boxes used to prototype code before the real Xbox 360 development hardware arrived.

sjl wrote:

Interesting. 24 cores/192 threads and 1 TB of RAM at the high end, at least to start - compared with up to 256 POWER7 cores/1024 threads and 16 TB of RAM in a p795. Wonder if IBM plans on scaling up, and hence this is just the "early" high end option - I'd be very surprised if not.

IBM has already indicated a 2 hop topology for a 16 socket system. The real question is if IBM will scale to even higher socket counts and/or need additional glue logic to do so.

sjl wrote:

No great surprise, though, to see IBM open up - it'd be damn hard for any single company to match Intel in CPU R&D these days. I do wish them luck: the thought of a CPU monoculture (or even just two options - ARM and x86) scares the hell out of me.

The PCIe based coherent bus interface is rumored to the be same as nVidia's NVLink that was recently announced. This is a big change as not only is IBM changing the bus topology from what was introduced in the POWER4 but they're also sharing it for usage in another vendor's products. This would also enable HSA-like functionality between the POWER8 and an nVidia Pascal based chip.

ronch wrote:

Yeah, it's a monster, alright. IBM wants to be the ARM of servers, as the article mentions. It's interesting to note that while ARM is planning to make a go for it in the server space, IBM's already well-positioned in that space and may actually stand a better chance of succeeding. Intel's got their hands full right now after beating AMD to a pulp. Say, am I the only guy who thinks AMD should also try its hand in the Power space? I know they're practically penniless these days but hey, they may be able to pull it off. Plus, they have a long history with IBM, trading tech as well as their very employees, who seem to be like electrons zipping in and out between the likes of AMD, Samsung, Nvidia, and IBM (and probably Intel too).They should really focus on ARM and something real to sell in the server space if they're really bent on ditching x86. They've helped Intel enough already, and have lost enough blood too doing it.

IBM and AMD have worked together in the past in the server market. IBM has (had?) a Hypertransport license. The only shipping product that used it was the CPC 945 north bridge that brought DDR2 to PowerPC 970 systems. Athlon 64/Opteron south bridge chips could have interfaced with this northbridge though I'm unaware of any examples. At one point, the rumors of the PowerPC 970's successor was to integrated both a memory controller and several Hypertransport links. After Apple decided to switch to Intel's x86 chips, this project was killed as it lost its larger customer.

After this, rumors did pop up that IBM and AMD were going to use a common socket for future chips. This would have been Socket G3 from AMD and used G3MX memory buffers. AMD cancelled plans to use that socket and transitioned to Socket G34. IBM did go on to release the POWER7 which did use similar memory buffer technology though it is unlikely that it was the same as AMD's Socket G3 proposal.

The problem for future joint ventures is that AMD has effectively existed the high end server market for the time being. Warsaw was simply a low power (if you call 99W low power) refresh of the Abu Dhabi chips released back in 2012. AMD's next major server products are single socket ARM based SoC's which is a different market than what POWER8 is targeting. It would make sense for AMD to adopt IBM's CAPI protocol for cache coherency over PCIe and to enable HSA-like functionality with other devices. Regardless, AMD's next multisocket x86 chip likely isn't going to arrive until 2015 at the earliest and in a new form factor.

the wrote:

The problem for future joint ventures is that AMD has effectively existed the high end server market for the time being. Warsaw was simply a low power (if you call 99W low power) refresh of the Abu Dhabi chips released back in 2012. AMD's next major server products are single socket ARM based SoC's which is a different market than what POWER8 is targeting.

Hmm... it may be a different market, but there's still overlap IMO. Where POWER8 would handle logically distinct servers via virtualization, an ARM based system could just assign each logical server its own physical CPU (or CPUs). If you take a step back, both seem to be targeting cloud-style flexible server deployments, where you've got a pool of resources you can carve up however you see fit.

just brew it! wrote:

the wrote:

The problem for future joint ventures is that AMD has effectively existed the high end server market for the time being. Warsaw was simply a low power (if you call 99W low power) refresh of the Abu Dhabi chips released back in 2012. AMD's next major server products are single socket ARM based SoC's which is a different market than what POWER8 is targeting.

Hmm... it may be a different market, but there's still overlap IMO. Where POWER8 would handle logically distinct servers via virtualization, an ARM based system could just assign each logical server its own physical CPU (or CPUs). If you take a step back, both seem to be targeting cloud-style flexible server deployments, where you've got a pool of resources you can carve up however you see fit.

For physicalizaiton there is indeed a bit of an overlap but I see raw pricing being more the determining factor in favor of AMD's ARM solutions. A POWER8 box certainly can run such workloads but it'd be serious overkill if it were to exclusively run them, even in numerous quantity via VM's. Rather I see such efforts on POWER8 in conjunction with virtualized application and DB servers on the same device for a common environment. The idea of consolidating a few small servers on a device with a few medium heavy VM's seems to be the more ideal workload in terms of cost/performance.

One niche case scenario I'd like to see but totally impossible would be having a POWER8 available as a fail over host in the event other smaller servers die. Obviously you can't fail an ARM guest to a POWER8 host but it'd be nice in an emergency scenario.

really agree with the remarks on the isa and its position right now. I don't think Intel is going to be too keen on losing its hold on the enterprise, and when battling Intel, it's more than just their technology.

As for the comments about the consumer space, I can't see that happening. We're more likely to be served there by high-performance ARM designs like Denver than we are by anything from this, right?

fhohj wrote:

really agree with the remarks on the isa and its position right now. I don't think Intel is going to be too keen on losing its hold on the enterprise, and when battling Intel, it's more than just their technology.

Yes, but IBM is a pretty big gorilla too. Just because they've been focusing on the services sector lately doesn't mean they still don't have the hardware chops or the marketing savvy.

fhohj wrote:

As for the comments about the consumer space, I can't see that happening. We're more likely to be served there by high-performance ARM designs like Denver than we are by anything from this, right?

Yup.

Unless you count the possibility that POWER8 might be powering the cloud servers that your ARM-based mobile devices rely on a couple of years down the road.

just brew it! wrote:

Yes, but IBM is a pretty big gorilla too. Just because they've been focusing on the services sector lately doesn't mean they still don't have the hardware chops or the marketing savvy.

Rumor 'round these parts is that the semiconductor mfg business is about to be sold to GloFo. IBM has about 5,000 people in the Essex Jct., VT plant (200 mm wafer line + design shops) just a little jumpy.

Captain Ned wrote:

just brew it! wrote:

Yes, but IBM is a pretty big gorilla too. Just because they've been focusing on the services sector lately doesn't mean they still don't have the hardware chops or the marketing savvy.

Rumor 'round these parts is that the semiconductor mfg business is about to be sold to GloFo. IBM has about 5,000 people in the Essex Jct., VT plant (200 mm wafer line + design shops) just a little jumpy.

Well that's sad, if true. Kind of along the lines of the implosion of Bell Labs. So will there be any significant SOI fabs outside of GloFo after this?

I can't wait until Kanter does a detailed write up on RTW. Wish Stokes was also still doing writeups on Ars.

I fell kinda bass-akwards into IT/Telcom/OPs, and never really took any CS courses (being the programming course on the VAX terminals that I despised and failed at), so i really enjoy them breaking down all the CPU tech for me. I dunno why I learning about all different types of CPUs,GPUs, and ASICs.

That said, I'm kinda jazzed about the POWER8. I have read, and understand some of the issues brought up in the comments. I just can't help feeling giddy, like a kid in a candy store, and imagining all the different possibilities to use it.


*Full Disclosure: I have only worked minimally with IBM equipment ( like 4 hours total), including helping setup some massive POWER Sytem (circa 2001) for a training class, and I had to read thru a bunch of service manuals, to figure out how to bring down and AS/400 and prep it for transportation. So, no, I'm not a shill.

I've been following POWER when it was still a consortium of Motorola, IBM, and Apple. I just think this stuff is neat. Like the ASICs in older generation consoles.

the wrote:

Unless you can find an old version of Windows NT for PowerPC, the POWER8 won't run any commercially released version of Windows.

Not even then. The PowerPC port of NT was designed for CHRP; I don't believe that PowerVM provides a CHRP-like base (I don't know for certain, mind, and I'm happy to be corrected if it does, but it would greatly surprise me.) It takes more than just support for the ISA to make a workable operating system - that's a necessary, but not sufficient, step.

(For the curious: I've done a little AIX admin work, and am certified - at a basic level - with AIX 6.1. Never done a huge amount with it, though; I know Solaris better, and even that was getting on over five years ago.)

sjl wrote:

the wrote:

Unless you can find an old version of Windows NT for PowerPC, the POWER8 won't run any commercially released version of Windows.

Not even then. The PowerPC port of NT was designed for CHRP;

Splitting hairs here, but Windows NT for PowerPC was actually designed for PReP. It was ported by Motorola and not Microsoft. Due to the lack of actual PReP hardware (ie Apple backed down from releasing Mac OS), it was dead in the market with just Windows NT being the sole major platform to get any traction on it.

sjl wrote:

I don't believe that PowerVM provides a CHRP-like base (I don't know for certain, mind, and I'm happy to be corrected if it does, but it would greatly surprise me.) It takes more than just support for the ISA to make a workable operating system - that's a necessary, but not sufficient, step.

Apple said that they couldn't support PReP even though they were originally on board. CHRP was an extension of PReP to further support more MacOS specific features. One of the main changes was that Open Firmware was to be required under CHRP. It was optional under PReP (and for reference Windows NT on PReP did use OpenFirmware to boot).

CHRP has been used by IBM for the POWER4 and POWER5 systems. IBM further extended CHRP with PAPR in 2006 which was used on the JS20/21 PowerPC 970 blades, POWER6 and later servers. Source: AIX 7.1 release notes. and its references of CHRP on modern POWER hardware.

The firmware and hardware is there to run Windows NT under PowerVM. The limitation at this point I see is if PowerVM can pass itself as a 32 bit platform as Windows NT 3.5/4 obviously has no 64 bit support. I believe that it can due to its ability to boot older 32 bit versions of AIX for backwards compatibility reasons. Little endian support is also needed for Windows NT. I just have no experience with PowerVM to know first hand.

sjl wrote:

(For the curious: I've done a little AIX admin work, and am certified - at a basic level - with AIX 6.1. Never done a huge amount with it, though; I know Solaris better, and even that was getting on over five years ago.)

Having worked at a remote hosting facility, it got to work with a wide abundance of hardware/software myself. I got to deal with pretty much everything under the sun there in terms of Linux and Windows as a jack of all trades technician. The rare stuff I've dealt with Solaris on SPARC + x86, and some AIX on POWER6 + POWER7 hardware there. Before that I worked in the IT side of the insurance industry where I learned the bare minimum about OS400 and zOS for some daily activities.

I have yet to use an Itanium system and neither OpenVMS or HPUX on any particular set of hardware.

just brew it! wrote:

Flatland_Spider wrote:

Probably not, no. The bulk of the server market is x86, and x86 is the path of least resistance. MS has the advantage of vendor lockin, and they can just port their stack over to whatever ISA they want.

But other than the now-abandoned MIPS and Itanium ports, they have chosen not to do so. (I'm not counting the mobile stuff, as that is not relevant to servers.)

The point being, this isn't going to cause MS or Intel to lower their prices. Intel has the bulk of the market versus Power's negligible marketshare, and MS produces software, which means this has no affect on them.

People running the MS stack are people running the MS stack, and they will pay what they need to keep that going. Thus, MS can name it's own price. If people start buy lots of PPC servers, MS can port Windows, and everything else, and they can keep charging the same amount. People jumping ship to another software ecosystem would get MS to lower prices, but MS devotees have been rather loyal.

Getting flanked by iOS and Android on ARM has to make MS management think twice about dismissing an ISA.

the wrote:

Splitting hairs here, but Windows NT for PowerPC was actually designed for PReP. It was ported by Motorola and not Microsoft. Due to the lack of actual PReP hardware (ie Apple backed down from releasing Mac OS), it was dead in the market with just Windows NT being the sole major platform to get any traction on it.

You're right, of course; I sit corrected. I knew that there had been an effort to make a common PowerPC platform, and a quick Google found CHRP; I'd forgotten about PReP as the first step on that path.

the wrote:

The firmware and hardware is there to run Windows NT under PowerVM. The limitation at this point I see is if PowerVM can pass itself as a 32 bit platform as Windows NT 3.5/4 obviously has no 64 bit support. I believe that it can due to its ability to boot older 32 bit versions of AIX for backwards compatibility reasons. Little endian support is also needed for Windows NT. I just have no experience with PowerVM to know first hand.

It would be an interesting exercise, but ultimately rather pointless - similar, in many ways, to getting VMS up and running on old VAX hardware. A good workout for the brain, but not particularly useful to actually get stuff done. Combined with the rather hideous (from the consumer perspective) price for even second hand POWER5 or later hardware, and I doubt we'll ever see it happen - never mind the fun in trying to get hold of a copy of the installation media.

So of course, now that I've said that, somebody is going to give it a go and get it to work...