IBM is expected to announce a new generation of Unix servers on Tuesday that can run multiple OSes simultaneously, systems it believes powerful enough to let Big Blue topple rivals Sun Microsystems and Hewlett-Packard.
IBM attacks Unix rivals with Power5
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I highly doubt they will Kill Sun or HP. Sun has been doing 64-bit and UNIX for Years and they are known for their scalability. Also the price of a 2 processor server for 12,000$ isnt that much of a deal especially since suns 240 is just a tad bit cheaper.
Why would an enterprise that is not locked into IBM already even care about IBM’s expensive Power systems? There is no value proposition going with big bucks hardware.
If you go with x86, you can already run many kinds of Linux, Windows, BSD, and Solaris on the same machine. For far far far far far far far less money than IBM’s overpriced systems.
IBM is charging $11K+ for an entry level system that should cost the customer no more than $2K. The rest of that cost is just pure customer-gouging IBM profit.
If it weren’t for vendor lock-in, this whole “expensive UNIX” market would be dead already.
Ummm, I think you really missed the boat on this one. These are not intended for the website you run out of your basement. These servers are directed towards financial institutions and the like that have need for serious computing power. Those X86 systems can run all sorts of OS’s but not at the same time w/ practically no performance loss. And you’ve obviously never seen the power of AIX. The “value proposition” is that you have a rock solid system that isn’t built from cheap commodity parts. When time is money, you tend to look at the high end. But please by all means, run your business on some cheap linux system that you pieced together with parts from Ebay. It’s always nice to have an advantage over the competition.
Just a thought.
I think Sun Microsystems has best price model.
Because there is thousands of customers who need the power, the reliability, and the stability that a microprocessor like the Power5 can give. The x86 platform can simply not compete againns this kind of systems in the market of ultra-end computing, critical applications, etc….
And i guarantee you that you get for your money with those kind of systems!!!!
” IBM is charging $11K+ for an entry level system that should cost the customer no more than $2K. The rest of that cost is just pure customer-gouging IBM profit. ”
I guess a Porsche should cost the same as a KIA in your own personal reality, eh?
Dude.
All these computer companies really aren’t selling these multimillion dollar systems to idiots who’d be better off stringing together a couple of PCs.
You obviously have no experience of the industry or any real idea of what you are talking about if you say things like “There is no value proposition going with big bucks hardware.”
Please return to slashdot. You could make a positive contribution there.
A 16CPU p5 570 almost waste a HP 64CPU itanium2 machine in TPC-C: http://www.tpc.org/tpcc/results/tpcc_perf_results.asp?resulttype=no…
Where can I find some benchmarks for the Power5 comparing it to say a Power4 or UltraSparc III let alone an Itanium? Anyone know?
The day and age of big expensive UNIX boxes is coming to an end. The New Zealand stock exchange switched from expensive UNIX to cheap Linux boxes running Oracle 10g. And got a massive improvement in performance and reliability.
Any of IBM’s UNIX boxes costs far more than a cluster of Linux boxes… and is more expensive to provision, repair, support, etc. It is a simple no-brainer for a CFO to switch from IBM POWER UNIX to INTEL/AMD Linux.
I’m sure if you are a big fan of vendor lock-in, you will support IBM POWER UNIX. But for the companies that want better value and less vendor lock-in, they will go with something else other than IBM POWER.
Overall, IBM loves POWER UNIX because it brings in giant profits and giant services. Customers… well… they wish they weren’t locked into IBM!
as long as all this trickles down into my Mac within a few years… all the jargon is well, jargon to me i’m just a GA, but those is some big ‘ol words and hoity toity terms, so heck i’m excited!
About reliability, I know of some place where they use a Sun Ultra system as a file server and cheap Linux boxes as desktops connected to it, and from what I heard, the Sun machine is by far the most reliable of them all. And it has by far the best support.
And those TPC numbers are indeed interesting: the IBM 32×1.9Ghz Power4 beats the HP 64×1.5Ghz Itanium system.
Right, because we all know that the NZX is at the forefront of global commerce. Show me any other large financial institution and I’d be willing to bet that you’ll find big iron there. Your one example is not really compelling evidence for your case.
So I guess Google should quit using cheap Intel/Linux boxes because they *obviously* aren’t getting enough computing power out of them…..
Just because there are a lot of stupid people who have historically thrown money at big vendors like IBM doesn’t mean it’s impossible to successfully deploy services on cheaper alternatives.
Even if you showed me a list proving that thousands of big companies were wasting money with big vendors like this, it wouldn’t convince me….the fact that there is even *one* successful business using the alternative not only proves you wrong, but proves that all of those business are wrong as well.
>>The x86 platform can simply not compete againns this kind of systems in the market of ultra-end computing, critical applications, etc…<<
How about the AMD Opteron?
Of course, much of expense/power of commercial quality servers come from RAIDs and the like. IDE drives are fine for one user, but not much more.
Still, it seems like you might be able to put together a decent opteron based system for than $12K.
Because if you read a bit more about Google you’ll see that the infrastucture they chose to run is just that, loads of cheap boxes they don’t care about if they die… They just add new ones. They don’t even care about removing dead ones!
So from your point point of view, because one business does something that seems to work all businesses should just do the same thing… otherwise they’re stupid… You’re just over simplifiying the situation really.
Do you think big businesses want to have a bunch of hippies in the backroom assembling boxes together with junk parts bought from eBay for a tenner and then happily rejoicing in installing the linux flavour of the month?!?! Are you deluding yourself or what?
Even if they were to use linux they would want to have a proper support contract for hardware & software, maintenance, training, etc… the likes IBM, Novell, Redhat could provide and that won’t be cheap anyway.
Just come down from your cloud mate.
So I guess Google should quit using cheap Intel/Linux boxes because they *obviously* aren’t getting enough computing power out of them…..
Just because there are a lot of stupid people who have historically thrown money at big vendors like IBM doesn’t mean it’s impossible to successfully deploy services on cheaper alternatives.
Give me a break. What google uses on its webservers/indexing servers is irrelevant. The real critical point at google is storage/retrieval which is where the big SANs come in. That’s their ‘big iron’. And if you think the SAN vendors (IBM is one btw although I think I heard google uses EMC) won’t bleed you dry you’re sorely mistaken.
Ok,
If you think google’s cluster is even remotely the same concept as the big iron that runs things like banks, you’re stupid. Google’s cluster is built for scaling and latency. NOT reliability and accuracy. A machine just died? oh well, we’ll index those webpages again with the next spider web crawl.
What’s that, a machine just died and you lost all your financial records? No sweat. NOT the same problem definition is it?
“So I guess Google should quit using cheap Intel/Linux boxes because they *obviously* aren’t getting enough computing power out of them….. ”
I sure hope you are not a decision maker in your company, Google’s search application runs best on many / clusters and it would not be very bright to buy a Power 5 cluster.
Do you think that Intel designed the Itaniam for the fun of it, they just woke up one day and said lets just design something from scratch and throw away billions in research …. They did it to compete in this market.
AMD are the only ones that have brought something to the market that is somewhere in between but even they will admit they do not play in this league.
I will admit though that companies like Dell who have a really good support network are putting pressure on the value proposition of these large Unix vendors.
Maybe Google wasn’t the best example…I was really just trying to find the extreme case.
My irritation with this argument is there are a lot of people out there that can’t accept the fact that x86/Linux based solutions really can be used to replace historically expensive solutions. Google is just one case (even if they don’t use it for everything). More and more are moving this direction every day.
Don’t misunderstand me. I’m not claiming that this will work for *evey* computing need…just that there are many many situations where it can and does…very successfully.
Yep, there is indeed a good use for dinosaurs.
It’s called a museum.
And that is where all the big iron from IBM and others is going to end up. The big iron way of computing is extremely high cost and generally poor energy efficiency.
A large cluster of small inexpensive smart nodes is going to offer reliability, scale, and speed that the big iron simply cannot match.
As the node size gets smaller, the communications gets faster, and the individual node speed increases, clusters will start making big iron look like the obsolete junk that it is.
Now if vendors would just get a clue and stop making the big iron… and stop forcing it on customers… and start innovating instead… wow things would be better much quicker than the glacial pace that vendor lock-in allows…
Businesses go for reliability, not the latest greatest. Reliability costs $$$ and thats what IBM, Sun, HP etc. sell.
The $12,000 figure is somewhat deceptive as it includes a year of support and those 2 processors are dual core (i.e. it’s a quad processor system).
Find an x86 machine of that quality, expanability, performance* and with that grade of support you’ll find it’ll cost a rather large pile of money as well – if such x86 machines exist at all.
*Try finding any x86 with competitive performance and I think you’ll likely be dissapointed. Yes, thoretical performance may be as good if not better but the POWER5’s 36MB cache and monster bandwidth will make a large difference in real lfe business applications.
>>Now if vendors would just get a clue and stop making the big iron… and stop forcing it on customers… and start innovating instead… wow things would be better much quicker than the glacial pace that vendor lock-in allows…<<
I’m not sure if you understand how things work on the corporate scale. For a major corporation, $12K is nothing. Also, for major corporate system – in most cases it would be insane not to move a glacial pace. You don’t replace decades worth of custom built legacy applications overnight. It’s not like upgrading your home PC.
Upgrading from an old machine to new machine is relatively easy if they are the same platform ( eg. 486 -> Pentium4)
compared to upgrading from old machine to a different
platform. ( eg. DOS/Win/586 -> mac G5)
It takes quite a lot of work/effort to do so.
So changing from old mainframes or mini-computers platform
to something different like Windows/x86/linux would
require alot of work, time, testing, money, etc..
You better hope that it is worth it in the long run…
My irritation with this argument is there are a lot of people out there that can’t accept the fact that x86/Linux based solutions really can be used to replace historically expensive solutions.
Sorry, thats not a fact and like preivious posters have pointed out Linux/x86 and Windows/X86 cannot play where the Big Iron boxes play. Now it doesn’t make sense to run some of the less critical business needs like DNS or low priority web serving on a Million dollar Super Dome. Linux or better yet BSD would be a good choice for that.
“like preivious posters have pointed out Linux/x86 and Windows/X86 cannot play where the Big Iron boxes play”
Look. You guys are really quick to jump up and defend your favorite ‘Big Iron’ solutions, but not one of you has ever ‘pointed out’ any proof that x86/Linux solutions can’t compete. Your only defense is that *most* businesses are using these expensive old solutions, and that doesn’t prove anything.
As previous posters have already ‘pointed out’ with REAL examples:
“The New Zealand stock exchange switched from expensive UNIX to cheap Linux boxes running Oracle 10g”
You want proof that x86/Linux can play where the big boys play? New examples like this one are popping up everyday.
Wake up people! If you’re going to argue about this, come up with a real rebuttal. Not this ‘Big Iron is more reliable’ garbage. You want to prove me wrong? Show me an example where it failed. I have yet to hear anything but success stories.
Seems like you all don’t know a whole lot about google. There are no SANS, or big iron as you put it, that come into play. Read about google fs, which is all on cheap x86 boxes. They don’t lose data when a machine dies since it’s replicated over many different machines.
Look. You guys are really quick to jump up and defend your favorite ‘Big Iron’ solutions, but not one of you has ever ‘pointed out’ any proof that x86/Linux solutions can’t compete. Your only defense is that *most* businesses are using these expensive old solutions, and that doesn’t prove anything.
Look, I work for a global 500 company and we have got Sun, IBM and Linux. Those few Linux boxes are mostly web servers and test machines. If we could substitute the $$$ machines with inexpensive Linux boxes we would.
You want to prove me wrong? Show me an example where it failed.
In the kind of business these machines are used in failure is not an option. The only ‘failure’ there would have been was on a low profile test box in a basement somewhere.
Yeah right, how do you think they ‘replicate’ this data ?
http://www.craigslist.org/pen/sof/34321021.html
Google: SAN/High Availibility Architect, Ads Site Reliability Eng. (mountain view)
Reply to: [email protected]
Date: 2004-06-21, 9:20AM PDT
SAN / High Availability Architect, Ads Site Reliability Engineering : This position is located in Mountain View, CA.
Sure, I spose they’d maybe have some big storage solutions where it’s aplicable, but again read up on google fs. It handles hundreds of terabytes all on cheap little x86 machines that are expected to fail. And replication on google fs has nothing to do with any other machines or SANS or whatever. When a file gets written multiple copies of that file are distributed throughout the cluster, 10, maybe 20 times, so that there’s no single point of failure.
ehhh i dont work in a fortune500 or anything of the sort. I have however just got my first enterprise grade computer – HP Proliant dl380 – dual proc 2.4 ghz xeons, hot swap ultra3 scsi drives.
There is no way to describe how much kick ass this thing has over my previous server boxes (beige boxes, intel clearwater cases with intel sata raid controllers etc.)
The server is valued at $18k AU compared to the $2-3k i used to spend on the previous servers. Whether i am running linux or windows on it is irrelevant – the software that comes with it is half the value (java based monitoring software – ILO software HP/CPQ management software). Redundancy, failover and fault detection is built right through this thing – raid setup time – 0, i have spent hours waiting for crappy ass SATA IDE raid cards to build up a raid set for instance.
I can only imagine what true super computer or big-iron stuff is capable of – 10 OS’s running off a single CPU!!!! the bandwidth those suckers would push through for inter processor comms etc.
Dont knock it till you try it – regardless of OS, the big-iron runs some very serious bang for the buck and is in no way dead or dying. Its like trying to argue that your average PC comes even close to the quality of the components you get in a g5.
Look. You guys are really quick to jump up and defend your favorite ‘Big Iron’ solutions, but not one of you has ever ‘pointed out’ any proof that x86/Linux solutions can’t compete. Your only defense is that *most* businesses are using these expensive old solutions, and that doesn’t prove anything.
Ever try to hot plug pci cards on a x86/linux system? try it on current P series IBM hardware. How about Dynamic partitioning with a Regatta i can add CPU and IO on the fly while the system is running. Or ahow about an end to end soultion to house a corporate 100GB Oracle database? HP Server(64×128 Superdome), HP Cache Centric Disk Arrays(XP512), HP SAN switches, HP Operating system…one phone number for support, 4 hour response time!
List one Linux X86 soultion provider that can do this.
Let me splain something to you since you have obviously never worked in a large corporate datacenter…downtime is measured in Dollars lost not hours/minutes/seconds. Companies pay for premium for Hadware/OS/service becasue in the long run it is cheaper then having 2000 accountants sitting around on their ass because the system is down.
“Companies pay for premium for Hadware/OS/service becasue in the long run it is cheaper then having 2000 accountants sitting around on their ass because the system is down.”
Well said.
As much as I love microsoft in terms of their advancement of the general computer community, the previous posters are right on the money.
I work in ERP as a system administrator of a Peoplesoft installation and large, serious companies rarely consider os’s like Linux or windows server to run their financial data.
The key word there is financial data.. downtime can happen, but you’d better have a hell of an explanation for it.
We use solaris 9 at work on a 240(R??) and it roxxors. AS/400 kicks ass too.
Windows continues to be a superior application server but in terms of over-all reliability the old school servers still own it.
– Microsoft Fanboy
POWER5 isn’t much of a RISC anymore, its so d@mn complex.. the most complex 64-bit CPU I’ve ever experienced. POWER5 is deffinately dangerous and scalable but it’s also very expensive. POWER is in a leauge of it’s own….. If i were sun or HP I’d deffinately be scared but HP And sun have out-done IBM Before and now since sun is going to be using SPARC64 instead of UltraSPARC in it’s highend systems and will jointly develope sparc with fujitsu, i think that IBM should watch out for that. IBM may be big.. HP may have the alpha engineers but lack of archetecture roadmap (Only itanium) … but sun has the APL…. So I think I wouldn’t buy HP stuff unless if i wanted an IA-64 server–they have good pricing for that stuff… So It comes out to sun/fujitsu vs. IBM….. It’s really hard to tell sun are the kind of people that will hang on to technology and try to support existing customers than IBM does.. so I think for a longterm investment id go with sparc.. but power really is much more powerful… this is if sun pulls thru it’s next quarterly earnings release.
Other than that… IBM is doing PR crap like they’ve always done. POWER may be great but it won’t be that hard for intel and sun/fujitsu to catch up… as for AMD Opteron.. the 64-bit Xeon will likely be a great alternative (I think the whole opteron thing is for short term profitability personally which it won) Sigh I love those opterons though I use e’m here.
about hotswapable drives and what not.. the CPU really does not make a difference in this field.. x86 systems can actually do this.. I’m sure you can add x86 extentions to something to a power CPU and it could probably run x86 apps as well. i think just because a cpu is just ‘x86’ compatible dosen’t mean its horrible.. although most x86 cpus really are peices of crap
“If it weren’t for vendor lock-in, this whole “expensive UNIX” market would be dead already.”
Last I checked the leader was Sun microsystems and there are multiple companies who make motherboards, processors, notebooks, servers, etc. Last I checked the SPARC archetecture is completely open.. hmm… I wonder what all this ‘vendor lock in’ was all about… actually i think its the only major arch. that’s actually open unsure though
POWER is powerpc compatible or last I checked, motorola makes powerPC as well. so your not completely locked in with power either although you don’t have all that much freedom as there aren’t any companies that make big boxes for POWER other than ibm.. er i hate ibm
Once upon a time, there were quite some alternatives to Intel for workstations and servers: Alpha, Sgi, PowerPC, UltraSparc, PA-RISC. Of those, only those architectures that boot using OpenFirmware (PPC and Sparc) have survived.
Now is this a coincidence, or will Itanium die too, because of the lack of OpenFirmware?
Don’t forget, these machines have wider traces and hardware and software, plus ECC ram to check, and repair damage from cosmic rays. So, you could replace some of these systems in the Test/development area, but not in Production.
Cosmic ray hits have been estimated as occuring once a month per server.
“POWER5 isn’t much of a RISC anymore, its so d@mn complex.. the most complex 64-bit CPU I’ve ever experienced. POWER5 is deffinately dangerous and scalable but it’s also very expensive. POWER is in a leauge of it’s own….. ”
RISC was never supposed to be equal with simplicity at any cost!!!! Geez, please read on the actual foundations behind RISC and what it means. It only means that a RISC machine’s ISA is composed of instructions which try to do whatever they can so they can pretty much achieve an effective CPI of 1. In fact the POWER architecture has always been far more complex than almost any of it contemporaries. The instructions are simple, the machine itself does not have to be.
Think of RISC as the microcode for the CISC machines (in fact that is how Patterson et al came up with the idea) you just get rid of a middle man of sorts… obviously you need an increased instruction bandwidth… and that is where a lot of the complexity of modern RISCs lays
“Other than that… IBM is doing PR crap like they’ve always done. ”
… as opposed to Intel? Gee wiz, this P4 I got really makes the internet go faster. Yeah… right. And of course eventually other people will catch up, I assume that under that guise most high performance systems would have never taken off. Why would we but this machine now? Eventually in a couple of years other cheapers systems will offer the same performance. Nice way of being kicked out of business by your competitors
“”If it weren’t for vendor lock-in, this whole “expensive UNIX” market would be dead already.””
“Last I checked the leader was Sun microsystems and there are multiple companies who make motherboards, processors, notebooks, servers, etc. Last I checked the SPARC archetecture is completely open.. hmm… I wonder what all this ‘vendor lock in’ was all about… actually i think its the only major arch. that’s actually open unsure though”
“POWER is powerpc compatible or last I checked, motorola makes powerPC as well. so your not completely locked in with power either although you don’t have all that much freedom as there aren’t any companies that make big boxes for POWER other than ibm.. er i hate ibm”
FYI, Hitachi uses POWER4 in their high end machines.