×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

Gelsinger Shoots Down EMC On ARM

timothy posted about a year and a half ago | from the context-switching-is-costly dept.

Intel 57

Nerval's Lobster writes "EMC president and incoming VMware chief executive Pat Gelsinger most likely shot down any hope that the company's storage arrays would be built around the ARM architecture. Gelsinger, who also helped orchestrate the VMworld show in San Francisco this week, presented an Aug. 29 keynote at the Hot Chips conference in Cupertino, Calif. Afterward, an audience member told Gelsinger that as many as 25 percent of all servers could be shipped around the low-power ARM architecture, then asked if Gelsinger agreed with that estimate. EMC previously shifted its product lines to Intel processors. Gelsinger told the audience member that the situation is unlikely to change, even if ARM could deliver workloads at a fraction of the power of an X86 chip."

cancel ×
This is a preview of your comment

No Comment Title Entered

Anonymous Coward 1 minute ago

No Comment Entered

57 comments

Hot chips? (0)

Anonymous Coward | about a year and a half ago | (#41178965)

Sounds like a x86 show, allright!

duh (5, Informative)

afidel (530433) | about a year and a half ago | (#41178971)

EMC arrays are already pushing more than what four westmere cores can do and they don't even have some of the cool features that the new breed of all flash arrays are doing (global dedupe and inline block compression). It will be a LONG time before ARM can handle todays storage workloads, let alone all the cool stuff they should be adding.

Re:duh (5, Insightful)

Richard_at_work (517087) | about a year and a half ago | (#41179083)

Agreed - I read the summary and article more along the lines of "Company makes internal product design choice, ARM fan disappointed."

Why is it news worthy what processor EMC choose? Are we missing something here? Did someone accidentally delete the paragraph which offered up something juicy, like "Gelsinger went on to shout the virtues of Intel, while his own personal Intel sales representative gave him a blow job and stuffed hundred dollar bills into his pants" ?

Whoopdi-fucking-do that 25% of all servers could be based around ARM - does that immediately mean everyone that isn't using ARM should flock to it? Or does it mean that companies should continue to go on making internal decisions about their own products?

Ya well (4, Informative)

Sycraft-fu (314770) | about a year and a half ago | (#41179451)

ARM fanboys are convinced that ARM is in every way superior to Intel and if only all the stupid companies/users out there would realize it then the world could switch and start the glorious ARM revolution.

I've gotten pretty used to it on /. :P

Re:Ya well (0)

Anonymous Coward | about a year and a half ago | (#41179639)

Oh you have haven't replaced your Symmetrix with a RaspberryPi yet?

Re:Ya well (0)

Anonymous Coward | about a year and a half ago | (#41180485)

ARM fanboys are convinced that ARM is in every way superior to Intel and if only all the stupid companies/users out there would realize it then the world could switch and start the glorious ARM revolution.

Ah, a Linux analogy. I understand now.

Re:duh (5, Insightful)

hairyfeet (841228) | about a year and a half ago | (#41180275)

Here is what I don't get about these ARM fanbois...why is it that one product has to "win" and the other "lose". Its like what Jobs told the reporters when they kept trying to turn it into Apple VS MSFT "MSFT doesn't have to lose for Apple to win" and he was right, he was going after a different market.

With ARM you are talking extremely low power, idles in the sub-mw range. That low a power draw has a cost and that's IPC. No way in hell the newest ARMs can put out even half the IPC of a 6 year old Phenom quad much less the new Intel cores. On the flipside X86 chips can't drop to the sub-mw idle usage unless you really strip down the core design like Brazos or Atom.

Its different markets folks, its a scalpel and a chainsaw. This VS crap is just bullshit, different use cases require different levels of IPC, some x86 fits better, some ARM. That simple.

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41180279)

Why is it news worthy what processor EMC choose? Are we missing something here?
 
Because a lot of people around here like to shout from the rooftops that Linux is run on an embedded system that no one can modify even if they had the desire or skill. It's a badge of honor. For those of us who use tools as a means and not an end, it doesn't make much difference. But to the fanboy? They need all the backing they can get to justify their decisions.

Re:duh (4, Insightful)

TopSpin (753) | about a year and a half ago | (#41180391)

I was left to wonder who cared about what CPU EMC storage arrays use. Is there some EMC hacker culture with its own open source storage OS running on EMC devices? Where can I get one??

Big storage needs lots of RAM. Intel has been providing 64 bit x86 CPUs, chipsets, memory controllers, etc. since 2004. I suppose EMC could license ARMv8 (the 64 bit extension of ARM that became available to licensees only about 12 months ago,) sign a contract with some foundry and design a system around 64 bit ARM. They certainly have the capital. If they did they would be the very first — there are no 64 bit ARMs being manufactured in volume anywhere yet.

One can imagine ARM having some success in big storage. ARM cores can be extended with custom silicon to integrate important algorithms and they can achieve very high core density. Much of storage is embarrassingly parallel so peak performance of CPU cores isn't terribly important.

Frankly I just don't think it matters much to big storage customers. They're paying for throughput, reliability, features and support, not an ISA. EMC could use 43 bit LISP processors soldered together by Taiwanese gnomes for all they care, as long as they can afford it and it performs.

Re:duh (2)

Guspaz (556486) | about a year and a half ago | (#41180501)

Well, one thing that most people are bizarrely forgetting is that Pat Gelsinger worked at Intel for 30 years, from 1979 to 2009... He was one of the main architects/designers of the 486 (parts of his work still being in modern Intel chips), and he later went on to be Intel's CTO...

Are people surprised that he would favour Intel chips after all that? I mean, seriously.

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41180843)

So do you have any evidence other than ad hominem that he didn't make this decision based on tehcnical merit?

Re:duh (2)

Belial6 (794905) | about a year and a half ago | (#41182159)

Really? You really can't understand why nerds would want to know what kind of processor a major technology provider uses? You don't consider it newsworthy unless there is a scandal? I find this new very interesting. If one want to know about blow jobs, there are much better sites than Slashdot.

Re:duh (1)

bws111 (1216812) | about a year and a half ago | (#41182515)

I don't consider it newsworthy unless it is news. A big part of the word 'news' is the word 'new'. There is absolutely nothing new in this article.

Re:duh (1)

Belial6 (794905) | about a year and a half ago | (#41184013)

Since the event reported on happening on Aug. 29, you are simply wrong.

Re:duh (1)

bws111 (1216812) | about a year and a half ago | (#41185449)

The event was new, but the event is not what is being reported on here.

This was not an announcement of change in direction or anything else. Somebody with an axe to grind asked the guy if they were going to use ARM, and he said 'probably not'. That is not news. News is when something changes (or maybe was expected to change, but didn't), not when it stays the same.

There is a reason why a 'news' reporter reporting that 'Generalissimo Franco is still dead' was funny - because it was not news.

There is no reason to assume that EMC would be changing the chips they are using, and some dope asking a question about it does not make the fact that they are continuing to use the same chips news.

Re:duh (4, Interesting)

Necroman (61604) | about a year and a half ago | (#41179105)

Unless they are targeting a lower end of the market. If you look at the low-end NetApp and Equilogic systems, I'm betting those could be (and may already be) powered by ARM chips.

One of EMC's competitors (Engenio, owned by NetApp now), had boxes in a variety of price ranges. The high-end boxes were all Xeons, while going down in price you would find PowerPC, and ARM chips (specifically XScale) inside.

Also, running on low-power chips is easier if they have a secondary chip to do RAID 5 and 6 calculations for them (or if it's built into the main CPU as an add-on module. Intel actually does this now with some of their Xeon chips [intel.com]).

Re:duh (2)

afidel (530433) | about a year and a half ago | (#41179163)

No, the low end NetApp uses an Intel processor and until recently it was the reason that nobody with any sanity would use their entry level array (the 2020/2050). There were all sorts of OS features you couldn't use because the mobile celeron in those boxes would bog down to the point of the storage becoming unavailable if they were turned on.

If there's one box in my infrastructure I don't want to be underpowered it's the storage controller because a performance problem there affects every other system attached to it.

Re:duh (1)

CajunArson (465943) | about a year and a half ago | (#41179215)

If a mobile celeron does not have enough horsepower then don't expect ARM to come up and replace it (unless this mobile Celeron is P3 or maybe P4 era mobile celeron).

As a point of reference, my 2.5 year old Atom server has over 5 times the single-threaded performance of my Raspberry Pi... and the Atom has 2 cores. Yes I know that my Raspberry Pi does not represent the fastest ARM solution available, but the faster ARM setups ain't 5x faster on a per-core basis either.. and that's just to catch up to an obsolete Atom...

Re:duh (3, Informative)

jkflying (2190798) | about a year and a half ago | (#41179469)

It's worse than that when you need something that uses lots of cache, like Java. I'm getting ~100x better performance on my i5 laptop than my Pi when doing Java FPU benchmarks, and this is with the Raspbian (hardware FPU) release.

Re:duh (1)

Necroman (61604) | about a year and a half ago | (#41181231)

Well, the true NetApp stuff, yes, you're probably right.

I was talking about the Engenio group (which NetApp purchased last year from LSI). They sold through channel partners like IBM, Oracle (via StorageTek who Sun bought), and others. Their low end systems used to run XScale processors, but that was probably back in 2003 or 2004.

Re:duh (1)

arth1 (260657) | about a year and a half ago | (#41179293)

Also, running on low-power chips is easier if they have a secondary chip to do RAID 5 and 6 calculations for them (or if it's built into the main CPU as an add-on module. Intel actually does this now with some of their Xeon chips [intel.com]).

Incidentally, that's one place where I see a good market for arm - as generic RAID controller CPUs, replacing WYBIWYG solutions with open raid firmware that's extensible and maintainable.
Need to support RAID 60? No problem, there'll be a module for that. Need to tweak for lowest possible worst case time at the expense of average time? No problem.

Re:duh (3, Informative)

tlhIngan (30335) | about a year and a half ago | (#41179319)

Unless they are targeting a lower end of the market. If you look at the low-end NetApp and Equilogic systems, I'm betting those could be (and may already be) powered by ARM chips.

One of EMC's competitors (Engenio, owned by NetApp now), had boxes in a variety of price ranges. The high-end boxes were all Xeons, while going down in price you would find PowerPC, and ARM chips (specifically XScale) inside.

I could swear the NAS appliance sitting on my desk at home had the EMC logo on it. And I know it has an ARM processor in it, specifically a Marvell XScale chip. It runs a modified version of Linux, but it's an EMC box (and even has some approved for VMWare thing on it).

So yes, EMC has gone ARM on the low end, specifically the stuff they market under their consumer brand as Iomega ("An EMC company").

Plenty fast for the home user, probalby sufficient for a mom and pop company, but will be woefully insufficient for anything larger. But nothing wrong - ARM makes it cheap and decently performing.

Re:duh (1)

Anonymous Coward | about a year and a half ago | (#41179763)

Which honestly makes you think. If you're aiming at performance and higher markets, instead of home-consumer, would you keep an ARM development team, or re-purpose your developers to keep working on high performance devices instead?

As someone pointed out, it's just a business decision, and if they think that's the path to move up faster, that's their decision. Not seeing why it's so much /. worthy except, as someone else claimed, this is only a troll to bring ARM fans to complain and cry about it.

Re:duh (1)

dbIII (701233) | about a year and a half ago | (#41186783)

Well, I don't use any EMC stuff because for some loads (eg. lots of nodes hitting storage at once) you need a bit of CPU power, and a few cores help a lot, so currently a traditional file server gets the job done more effectively. It appears they are aiming for that niche now.

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41180641)

EMC doesn't make anything that small. EMC recently came out (in the past 1-2 years) with their small versions of the VNX; that's as small as they get.

What you have sounds almost exactly like the Buffalo Terastation that I have at home - it has a Marvell ARM chip in it, has a 'Certified for VMWare' sticker on the box, and runs Linux. It's a real slow piece of crap. It's fine for home usage, but it's not something I'd ever try to deploy as an enterprise solution. The throughput is ridiculously slow - you can get nowhere near a gigabit out of it - you're lucky if you can even hit 100 megabits. I only have it because its a freebie from work (they didn't want it because it was a slow pile of crap).

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41180983)

Forgot to mention that the VNXe (smallest EMC NAS available) has quad core Xeons.

Re:duh (1)

afidel (530433) | about a year and a half ago | (#41183785)

No, EMC bought Iomega, they do have a prosumer/SMB brand now (kind of like Cisco buying Linksys to give them a product in the same market).

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41188597)

I don't see it anywhere on emc.com but I'll take your word for it.

I'd like to see the performance numbers for it. Would you be caught dead deploying this for a vmware farm?

Re:duh (1)

afidel (530433) | about a year and a half ago | (#41189393)

Depends on the size, EMC showed off the px6 using six SSD's at last years EMC World and they booted 100 VDI clients in about a minute and a half off it.

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41179461)

I'm still trying to figure out who would really *care* about what cpu the vendor uses in their appliance, as long as the appliance (storage array in this case) has the performance you need in your environment. FFS, they could be using a proprietary cpu core implemented in FPGA's to drive the thing for all most people care - as long as A) they can get service/parts, and B) it performs.

Re:duh (0)

Anonymous Coward | about a year and a half ago | (#41179533)

No one other than fanbois so the article writer is trying to bait them to drive up page clicks.

HAH! (-1)

Anonymous Coward | about a year and a half ago | (#41178989)

VMware can't run ARM because Windows is x86 only!

CHECKMATE, Nerds!

Ugh (0, Offtopic)

Desler (1608317) | about a year and a half ago | (#41179097)

Another day another thinly-vieled attempt at getting people to drive up page hits to the ghost towns of SlashBI/Cloud/etc. Just post the story here rather than trying redirect people to your buzzword sites.

Re:Ugh (2)

PCM2 (4486) | about a year and a half ago | (#41179193)

Another day another thinly-vieled attempt at getting people to drive up page hits to the ghost towns of SlashBI/Cloud/etc. Just post the story here rather than trying redirect people to your buzzword sites.

Not to mention that the "buzzword sites" are all advertising anyway. A lot of the posts on SlashBI are by Mike Vizard, [kingfishmedia.com] who is currently a member of the "content strategy team" at King Fish Media, a company that has trademarked the phrase "own your own media channel." [kingfishmedia.com] Basically, Slashdot has sold itself out as a propaganda channel for tech vendors, disguised as news.

Re:Ugh (0)

Anonymous Coward | about a year and a half ago | (#41179275)

If I was an advertiser on SlashBisexual I'd be pretty pissed knowing that page hits are artificially inflated by such ridiculous shit.

Re:Ugh (0)

Anonymous Coward | about a year and a half ago | (#41179901)

SlashBisexual?

I thought it was SlashBoysLove

Re:Ugh (2)

Jeng (926980) | about a year and a half ago | (#41179467)

At least it is not another fucking video, nor is it an ad laden blog that references another blog that references something that somewhat resembles the summary.

Well, if it is ad laden, I don't see it.

Re:Ugh (2)

Desler (1608317) | about a year and a half ago | (#41179549)

The SlashBI/Cloud/Whatever sites are nothing but ads and slashvertisement articles. That's why the vast majority have zero comments.

Re:Ugh (1)

Jeng (926980) | about a year and a half ago | (#41180541)

nothing but ads and slashvertisement articles

And that is different from every other submission how?

As I said, at least it is not a link to a blog, that links to another blog, etc.

There are no advertisements on the screen I went to, not counting the story itself though. That is actually an improvement over most stories presented on slashdot.

Re:Ugh (0)

Anonymous Coward | about a year and a half ago | (#41180619)

If they want traffic, they should embrace the slashBI name. Is there a person alive who sees that and actually thinks "Business Intelligence"?

Re:Ugh (1)

DavidRawling (864446) | about a year and a half ago | (#41187425)

Absolutely - I see SlashBI and I think "great idea, let's slash our business intelligence [further] by getting the MBAs to read this site". I doubt that that's the reaction for which the MBA crowd aimed, however.

In related news... (4, Funny)

HockeyPuck (141947) | about a year and a half ago | (#41179187)

ARM fan is pissed off that IBM zSeries won't be using ARM processors either.

Re:In related news... (0)

Anonymous Coward | about a year and a half ago | (#41182545)

... it's System Z, now. We changed the name again ...

sorry.

niG6a (-1)

Anonymous Coward | about a year and a half ago | (#41179217)

to be about doing area. It is the OpenBSD leader WTheo started work on

Keep the conspiracy alive (1)

Anonymous Coward | about a year and a half ago | (#41179585)

Remember! Gelsinger was a Senior VP and worked at Intel for thirty years.

Pat Gelsinger, former CTO of Intel (3, Informative)

charnov (183495) | about a year and a half ago | (#41179613)

In other news former CTO of Intel who has huge amounts of stock options says Intel chips are awesome! Seriously though, our tiny little SAN maxes out 8 Xeon cores and 16 GB of ram while running less than 30 heavy VMs (80,000 IOs on average). I don't see ARM in this space for a while.

Re:Pat Gelsinger, former CTO of Intel (0)

Anonymous Coward | about a year and a half ago | (#41179697)

agree but on the other hand i do see FPGA here

Re:Pat Gelsinger, former CTO of Intel (0)

Anonymous Coward | about a year and a half ago | (#41180933)

FPGAs are around. Texas Memory System products use xilinx and lattice devices in their products, for example.

The disk drives use more power than the CPU (4, Insightful)

Animats (122034) | about a year and a half ago | (#41179775)

If you have a rackmount case full of big disk drives front-ended by a CPU, the CPU isn't using a big fraction of the power. Nor does it constitute a large fraction of the cost. ARM is a 32-bit architecture. If you have a few terabytes in your disk array and 10Gb Ethernet going in and out. you might want more than 4GB of RAM in front of it.

This sounds like some ARM fanboy thing.

Re:The disk drives use more power than the CPU (0)

Anonymous Coward | about a year and a half ago | (#41179839)

Re:The disk drives use more power than the CPU (3, Informative)

CajunArson (465943) | about a year and a half ago | (#41179907)

Grandparent poster is right on 32 bits and anonymous coward is a stupid marketing drone. Arm recently released a specification for chips that will use 64 bit addressing... when they finally ship in 2015 or so, making them about 12 years late to the party* after the launch of the Athlon 64.

* Yes I know that 64 bit was around much much longer than 2003, but I'm talking about the consumer space here.

Re:The disk drives use more power than the CPU (2)

Quila (201335) | about a year and a half ago | (#41180509)

Agree with the point, but we're not talking about consumer space. In the embedded high-performance market like this, 64-bit was standard back in the 90s. The 64-bit DEC Alpha was common for Raster Image Processors and was in early NetApps. 64-bit MIPS found quite a lot of use in this area too.

So ARM will be more like two decades late to the 64-bit party.

Re:The disk drives use more power than the CPU (2)

Desler (1608317) | about a year and a half ago | (#41179933)

In your rush to one up them conveniently left off this part from the quote:

As of March 2012, only the ARMv8-A ("application") profile has been defined, and no implementations have been announced.

So there are no implementations of ARMv8 so his statement is effectively correct for any ARM chip or SoC you can currently purchase.

how is emc on arm anything to hope for???? (0)

Anonymous Coward | about a year and a half ago | (#41179807)

ARM nowadays stands mostly for locked down tablet crap, phones and locked down consumer equipment .. the cost of cpu is marginal in a storage controller
compared to what the entire unit goes on sale for and there are extreme performance requirements,
this story just doesnt make sense .. my guess the spin doctors at geeknet were paid to put this out.

ARM (-1)

Anonymous Coward | about a year and a half ago | (#41179917)

Ain't Running Much

Not all servers are big (1)

wvmarle (1070040) | about a year and a half ago | (#41186621)

Comments here all of course look at big stuff, servers that handle huge amounts of storage, that serve dozens of VMs for remote users, run busy web sites.

Most servers are not like that. EMC may be an example of the high-end stuff but most servers in this world are low-end. They have to serve files and e-mail to maybe a dozen users, they have to store the media catalog of a four-person family, that kind of things. That's where ARM may be very useful.

Check for New Comments
Slashdot Account

Need an Account?

Forgot your password?

Don't worry, we never post anything without your permission.

Submission Text Formatting Tips

We support a small subset of HTML, namely these tags:

  • b
  • i
  • p
  • br
  • a
  • ol
  • ul
  • li
  • dl
  • dt
  • dd
  • em
  • strong
  • tt
  • blockquote
  • div
  • quote
  • ecode

"ecode" can be used for code snippets, for example:

<ecode>    while(1) { do_something(); } </ecode>
Sign up for Slashdot Newsletters
Create a Slashdot Account

Loading...