Is Sun's Niagara Server Viagra? 190
argonaut writes "Ace's Hardware has an in-depth article on Niagara -- Sun's upcoming parallel server processor with 8 cores and 4 threads each. The article discusses the chip's radical architecture and what kind of performance can be expected from it in traditionally thread-heavy server applications like web hosting, databases, and other multi-user applications. Given the recent cancellation of the UltraSPARC V, it seems this is going to be Sun's new direction for its in-house CPU design efforts. Furthermore, both Intel and IBM are working on other highly parallel processors and AMD is expected to eventually introduce a dual-core Opteron. So, will more threads prop up Sun's performance?"
Nice title! (Score:3, Funny)
Best. Title. Ever.
Re:Nice title! (Score:3, Funny)
Re:Nice title! (Score:5, Funny)
I don't know. I still liked Lance Bass To Continue To Plague Earth's Surface [slashdot.org].
Re:Nice title! (Score:2)
Re:Nice title! (Score:2)
Re:Spell check Niagra - Niagara (Score:2)
And on Google, "Niagara" outnumbers "Niagra" almost 10:1.
GTRacer
- Niagara, Niagara was a good movie
Re:Nigga? (Score:2)
I think he/she/it (or whatever) meant the Nigerian spammers.
Niagara falls! (Score:5, Funny)
Re:Niagara falls! (Score:2)
Sun Sets By 2008 (Score:5, Informative)
The Niagara processor and its successor, Rock, are based almost entirely on the Hydra [stanford.edu] processor that Professor Kunle Olukotun developed at Stanford University. He co-founded the company, Afara Websystems, that Sun Microsystems purchased. If you want to know how Niagara works, just check out the Hydra processor.
The reason that Sun Microsystems abandoned the UltraSPARC V and successors is that the design teams who developed the UltraSPARC processors after the UltraSPARC II were just horrible. Normally, when engineers develop the microarchitecture and eventually the Verilog model of the chip, a documentation engineer documents all aspects of the chip. In the case of Sun Microsystems, there was no documentation engineer. Ultimately, on the very day that Sun released its processor to the market, no documentation existed.
Even Sun's own engineers did not have the documentation to develop the boards that would accept an UltraSPARC processor. The whole experience is incredibly stupid but true. Most engineers on the processor teams are Indians or Taiwanese, and they just "do not do documentation". Various Linux gurus complained about the lack of documentation needed to port Linux to the latest version of the UltraSPARC. Sun would have loved to produce the documentation if it existed. Unfortunately, it just did not exist.
UltraSPARC V had the same problem. The whole design process for the UltraSPARC V was a mess, and canceling the project fixed the mess.
Sun does not have the engineers with the skills to build a fat-core processor. So, Sun moved to thin-core processors like Niagara. They are easier to build and to document. They simply matched Sun's skill set, which is derived mostly from foreigners.
Unfortunately, for Sun, what is easy for Sun to design and build is also very easy for IBM and HP to design and build. If you IBM and HP engineers are reading this article, you are in luck. Just check out the Hydra processor, and you will know the 80% of microarchitecture of the Niagara processor. Fortunately, for you guys, building a Hydra-based processor that executes the Power instruction set architecture (ISA) or the HP ISA is much easier than building a processor that executes the SPARC ISA. Those damned 128-register register windows diminish the number of cores that can be squeezed onto the die.
I would like nothing more than to see Sun's processor department setting by 2008. Sun should not be in the business of designing processors. The UltraSPARC-III fiasco should have been a big clue.
If Sun were purely a software house, we'd have a chance of making a profit.
Re:Sun Sets By 2008 (Score:2, Interesting)
Firstly, Sun are absolutely right to keep hold of their processor technology. The market has long since grown up to the feeds-n-speeds contest, and realised that memory latency and I/O thro
Foreign workers (Score:2)
I have a hard time understanding how one's skill set is determined by their nationality. Ability to communicate in English, perhaps. But then again, I would say the same about any Bay Area born engineer that moves to Bangalore and has to communicate part time in Hindi.
In any case, I think there's a major difference between one's ability to write prose or design a chip. Communication among team members is crucial, of course. Bu
Re:Niagara falls! (Score:1)
Re:Niagara falls! (Score:2)
best-title-evah-my-arse dept. (Score:2, Funny)
What? (Score:5, Funny)
Geek porn! (Score:5, Funny)
Without warning, underwear tents pop up all over the country.
Viagra? (Score:5, Interesting)
Seriously though, why did the author have to use the term Viagra to simple mean 'performance boost'?
Re:Viagra? (Score:1)
Re:Viagra? (Score:1)
Obviously.
Re:Viagra? (Score:2, Funny)
Re:Viagra? (Score:2, Insightful)
ah, the smell of a new computer . .. (Score:5, Funny)
And why nerd dates are doomed to fail (Score:5, Funny)
"Honey, I need to open this box from Alienware first."
"Honey, I'm done playing counterstrike. We can go to bed now."
"Yeah right. I need to get to work."
Probably not, because (Score:2)
Will more threads prop up Sun's performance? (Score:3, Interesting)
Re:Will more threads prop up Sun's performance? (Score:5, Informative)
Of course, you'll only see a significant benefit when you've lots of threads in the run-ready state (which mostly happens when you have lots of threads, period). Given java's fondness for threads, and solaris' already outstanding handling of systems with thousands of threads, this seems like a smart optimisation choice.
So, with the necessary Solaris installed, your existing Tomcat running on your existing JVM will see all the benefits.
Re:Will more threads prop up Sun's performance? (Score:3, Interesting)
Not it won't. At least not so simply. It will see the benefits if there are enough concurrent threads running (as you said), and even that if these are not waiting for each other. So it will work for many clients at once.
I have my doubts that this architecture will help with most real world tasks - even real world server tasks - even with completely blown out of proportion threading like j
Re:Will more threads prop up Sun's performance? (Score:3, Informative)
It might be just the way I'm reading it but the only difference is that Intel started small (hyperthreading) and still currently rely on several physical processors. IBM's Power already has multiple cores, and this isn't the first time a dual core Opteron was mentioned.
Re:Will more threads prop up Sun's performance? (Score:2)
One core of the Power4 is nearly twice as fast as the fastest _available_ Sparc Proc.
See? Intel and IBM are late to the game because they didn't see the need to be earlier. The gap between Sparcs and other CPUs has been getting bigger and bigger. It will come to the point where even the most loyal customers won't be able to justify buying Sun equipment because they are so d
Re:Will more threads prop up Sun's performance? (Score:2)
Re:Will more threads prop up Sun's performance? (Score:4, Interesting)
Re:Will more threads prop up Sun's performance? (Score:2)
Weird analogy... (Score:5, Funny)
Viagra more like Intel (Score:2)
doh (Score:4, Funny)
AMD is expected to eventually introduce a dual-core Opteron
If Opteron rhymed with Levitra, I could make a pretty funny joke here...
No (Score:1, Funny)
Dual core Opteron? (Score:2)
(I figured that was going to happen 9 months ago when the 8-way systems weren't coming out... just waiting for the inevitable)
Of course this will be the direction Sun goes in. (Score:5, Interesting)
It's the same thing that's been happening for the last decade. As x86 slowly creeps in on Sun/IBM/Whatever's market, they have to come up with something "bigger".
Right now, the Opteron, with embedded memory controller and gobs of I/O, has really entered what was previously a niche market that Sun made very nice profits from.
So, now that particular cash-cow has fallen to the ravages of commodity parts, they're moving their sites even higher. Sun's never been the company to make $5 profit on each of 50 million computers, they'd much rather make $300,000 each on 1,000 computers.
steve
Re:Of course this will be the direction Sun goes i (Score:5, Interesting)
Re:Of course this will be the direction Sun goes i (Score:5, Interesting)
Re:Of course this will be the direction Sun goes i (Score:3, Insightful)
$34,000 - $20,000 = $14,000
That's $14,000 that they would lose in profits if they were to compete by matching the price of the commodity system.
Re:Of course this will be the direction Sun goes i (Score:2)
Re:Of course this will be the direction Sun goes i (Score:2)
I generally call it "commodity" if you can go out, buy the parts from different places, and put it together yourself. Although in the quad-cpu market you *usually* have to get the motherboard and case together, it still qualifies.
I have a quad-Xeon (P3) at the office, I bought the mobo/case from SuperMicro, and the other parts from various distributors. Sure, the price was around $15K by the time I got all of the doo-dads (10-disk RAID array...), but it was still all "commodity" hardware to me.
N
Re:Of course this will be the direction Sun goes i (Score:2)
Re:Of course this will be the direction Sun goes i (Score:5, Interesting)
This is not bigger. Taken to the extreme, this is like if Commodore was still in business and tried to sell computers with 2^32 6502 procs.
Look at the chart in the article to see how desperate Sun is:
They admit that existing Opterons Xeons not only kick the ass of their newest, existing architecture for a single thread, they also concede that even their not-existant future proc won't even be faster for single threaded apps.
Ok, you say, but it is faster for multithreaded apps. The only problem with that is that I bet that a recent multiproc Opteron/Xeon will give the future Sun architecture a run for its money.
And IBM/Intel won't have any problems building multicore procs, if they want. They just don't need to, at the moment.
IOW, looking at this chart one might ask himself why tSun even tries to build processors nowadays.
Re:Of course this will be the direction Sun goes i (Score:3, Interesting)
Re:Of course this will be the direction Sun goes i (Score:2)
'taint no CPU advances going to help Sun now (Score:5, Interesting)
1. overall system performance of their partitionable systems (i.e. the ones people will pay a premium for over low-end systems where Linux on Intel/AMD is killing them) is severely hampered by their 150MHz (Mhz!) backplane. Sun views this as a plus because it allows customers to run boards with differenc CPU speeds (e.g. a 750MHz board (5x backplane speed) and a 900MHz board (6x backplane speed)). So, board to board thruput suffers and overall scalability is reduced.
2. Their desire for greater hardware isolation between domains, down to only a 2 or 4 CPU board with whatever memory happens to be installed on those boards, severely limits the flexibility in providing workload management between logical servers (domains), as well as less flexibility to create / deploy fewer, smaller servers. IBM's LPAR architecture, and HP's VPARs, are kicking Sun's ASS!
Re:'taint no CPU advances going to help Sun now (Score:5, Interesting)
1. It is an easier upgrade path for customers. I think Sun learnt that it is easier to sell its customers incremental upgrades than to sell them brand new designs. Remember that the market they sell to (telco, financial) absolutely despises having to test all their mission-critical applications on new, unproven hardware. So while the slow backplane is a performance limitation, many customers may prefer stability to cutting-edge performance.
2. Wait for the 'Zones' in Solaris 10
Re:'taint no CPU advances going to help Sun now (Score:2, Interesting)
2. They HAD to come up with something to counter LPARs, etc... the market shifted and they got caught with their domain's down at their ankles... of course, no doubt IBM and HP could (and frankly, maybe have) come up with something akin to zones / containers as well, ON TOP OF h/w LPARs... the fact remains, better h/w flexibility
Re:'taint no CPU advances going to help Sun now (Score:2)
As for Sun Fires flying out the door, sales figures are excellent - look at the success of the V210, V240s, V440s, etc. All selling very well.
Re:'taint no CPU advances going to help Sun now (Score:5, Interesting)
2. As with all things, there are cost/benefits to every feature. I'm sure there are applications that are better suited with greater hardware independance. Still I'm not sure what you mean here, are you advocating more manageability between CPU's and different domains (which is good for managing severals VM's?)? With a processor that has eight cores, you'd assume that one would be able to put a vm on each one with that vm, having four hardware threads available. How is IBM/HP's offering different?
Re:'taint no CPU advances going to help Sun now (Score:2, Informative)
Re:'taint no CPU advances going to help Sun now (Score:4, Interesting)
If I understand the backplane as the CPU->CPU bus, then wouldn't a multi-core CPU reduce dependency on the backplane? For applications that require low latency and high throughput how can you get higher transfer rates that what's available on the CPU itself?
As per the second point: Wouldn't a multicore-multithread multi-CPU server offer more flexibility for load balancing and on-damand peak handling (ie, move CPU2 Cores 1-3 from mail/fileserver duties to httpd to handle slashdotting)?
It seems the differences you are stating are about the overhead of managing multiple physical CPUs, but with this new chip a 4 way could handle what a 16 or 32-way did before. Thus the intra-CPU differences between IBM/HP and Sun are fairly irrelevent. Maybe I'm missing your point.
Re:'taint no CPU advances going to help Sun now (Score:2)
It doesn't matter what kind of bus it is, the fact that a CPU has multiple cores probably will not help its bandwidth. Short form, it depends on how the cores are wired. If you have a bus into the package which goes into one core, and that core feeds the second core, then no, it's not going to help. For instance if you put two Opterons in a chip, and ran a hypertransport link between
Re:'taint no CPU advances going to help Sun now (Score:4, Informative)
2) Solaris 10 with N1 Grid containers [sun.com] should give Solaris the finer grain control that users want.
Re:'taint no CPU advances going to help Sun now (Score:2)
Taking an example of a 1280... You have a system bandwidth of 9.6GB/s, which is higher than anything IBM have. You purchase it today, with 8 UIII CPUs and in a year you find you need more power. You can then upgrade with four extra UIV CPUs, withouth taking the machine down, whilst making use of the latest CPU tech.
With IBM you'd have to pay a premium for old CPUs or buy an entire new machine...
Re:'taint no CPU advances going to help Sun now (Score:2)
You've been bleating about not being able to scale bandwidth - I've outlined a situation which clearly shows the value of Sun's approach, which you don't seem willing to address. Fair enough.
Re:'taint no CPU advances going to help Sun now (Score:2)
Multi-threading and selectable IO channels (Score:5, Interesting)
especially for server software that threads.
I'm especially intrigued on how this will
work with the Java NIO (new IO) libraries,
which handle pooled selectable IO channels.
Niagra and Java NIO together looks like
a really fast way to do mass serving...
Can anyone comment on threads sharing IO?
Re:Multi-threading and selectable IO channels (Score:3, Informative)
Re:Multi-threading and selectable IO channels (Score:3, Funny)
May I suggest this slight modification:
This Sun Niagra processor looks promising,
especially for software that threads.
I'm especially intrigued on how this
will effect Java IO overhead.
Memory subsystem? (Score:5, Insightful)
I hope that they've made some vast improvements or they're gonna have some serious issues feeding that beast. Systems now, even the Opteron which is among the better mem controllers around for a commodity processor, still have issues with wait states. Uberthreading it and dumping more cores on the chip will only make the situation worse unless they do a serious upgrade of the memory controller.
If they do not, why pay bazillion bucks for a processor that is idle for most of the time?
Re:Memory subsystem? (Score:3, Interesting)
Re:Memory subsystem? (Score:5, Informative)
It's interesting that you should mention that, because one of the early multi-threaded processors (at Tera) was specifically designed to solve that problem. The theory was, and still is, that if one thread has to stall it's OK because there are still plenty of others that can keep running from cache. So no, you won't have N threads all running without waits and yielding N threads' worth of performance, but you'll still have enough live threads to give you more performance than you'd have with a single-threaded core.
Only time will tell which way it will really go. Most likely, there will be some workloads on which this approach works extremely well, some on which it provides no benefit, and a few on which you would have been better off with a "fat" single-thread CPU design. One thing to remember is that if the system has X threads, cache pollution and memory bandwidth are going to be problems either way. The fact that the multi-thread processor can still get some work done on some threads even while others are blocked waiting for memory will probably allow it to maintain an advantage over a faster single-thread processor that blocks completely more often.
Re:Memory subsystem? (Score:5, Interesting)
1) Memory latency will be a bigger and bigger bottleneck in systems as cpu frequencies scale
2) Technology will not allow memory latency to keep pace with cpu frequency.
See ace's previous interview [aceshardware.com]
A snippet:
Chris Rijk [Ace's Hardware]: Stalled on waiting for data, basically.
Dr. Marc Tremblay: Yes. In large multiprocessor servers, waiting for data can easily take 500 cycles, especially in multi-GHz processors. So there are two strategies to be able to tolerate this latency or be able to do useful work. One of them is switching threads, so go to another thread and try to do useful work, and when that thread stalls and waits for memory then switch to another thread and so on. So that's kind of the traditional vertical multithreading approach. The other approach is if you truly want to speed up that one thread and want to achieve high single thread performance, well what we do is that we actually, under the hood, launch a hardware thread that while the processor is stalled and therefore not doing any useful work, that hardware thread keeps going down the code as fast as it can and tries to prefetch along the program path. Along the predicted execution path [it] will prefetch all the interesting data and by going along the predicted path [it] will prefetch all the interesting instructions as well.
Re:Memory subsystem? (Score:2)
Why? Niagara sounds good. That's reason enough for many other products on the market.
Re:Memory subsystem? (Score:2)
So do I have to see a doctor ? (Score:4, Funny)
WARNING! (Score:5, Funny)
Re:WARNING! (Score:2)
Seek the nearest sorority house.
Unfortunately (Score:4, Insightful)
No matter how impressive the architecture happens to be, it will be evaluated by comparing it with a rack of P4's.
How many Power server systems does IBM ship compared with x86 systems?
Revenue per system might be better for high end systems, but the volume - the market size - is just not growing.
This could be HUGE (Score:5, Interesting)
But that's a big "if."
Re:This could be HUGE (Score:2, Interesting)
Re:This could be HUGE (Score:2)
I disagree... (Score:2)
My only concern: (Score:4, Funny)
My god what have we done.... (Score:2, Funny)
Re:My god what have we done.... (Score:2)
Some things you just don't admit to people man.
Sun is dead (Score:3, Funny)
The Rock will flatten "Niagra" (Score:5, Informative)
Sony, IBM, and Toshiba Have the Cell Processor (Score:4, Interesting)
The Cell processor follows the same design idea, if not more radical. Sony has cross licensed it to IBM and Toshiba. Toshiba is already planning on using Cell in high end processing.
The big question is if bandwidth constraints will choke these massively parallel superscalar processors.
Sun no longer belongs in the processor design (Score:3, Insightful)
Now all we need... (Score:2, Funny)
Get cheap $U|\N n-i-a-g-a-r-a serrvrs (Score:5, Funny)
c l i c k here [64.124.140.199]
babble james tycoon motherfucker nabob carlin reilly bubblehead chomsky allergy morning comment plastic bellybutton bookmark gate askslashdot Produce diplopod digress superheterodyne derriere whiskery Antithetical dovekie anthropic unshaped cresol perfectly Pothook slaveholder unzip hotbox athodyd occur verderer Bilander Dacron imprecate bulbar costmary sciolism coco Refusal sclerous unequal missal erratic redroot monotypic
Yes, you should.... (Score:2)
For those complaining about Sun CPU performance... (Score:2, Informative)
Why Niagra will suck (Score:5, Informative)
Current Oracle licensing schemes require that clients pay PER CPU CORE, for multi core processors. This screws anyone that uses Sun boxes, because the cores are US2 based. So the Oracle client has to pay heaps of cash to use, effectively, a 5 year old processor design. In addition, Oracle licensing requires that if your server has the capacity to hold more than 4 processors (eg cores) thes you have to pay the "enterprise" rates.
So in conclusion, the price of Oracle on a 2 cpu Xeon, AMD, or Ultra sparc 3 is about $6000. The price for Oracle on a 2 cpu Niagra (8 cores each) will be $320,000. Only an idiot will use this cpu (or this database). Since a lot of companies have a huge investment in Oracle, they will have no choice but to switch to x86 hardware. Sun is going to kill themselves with this design, despite the fact that the design, in itself, will greatly improve the throughput of their servers.
Oracle licensing is heavily slanted toward intel arcitecture, they have always penalized people for using risc based processors.
Re:Why Niagra will suck (Score:2)
I just browsed through all the licensing docs I am privy to and they make no distinctions between multiple cores and a single CPU. Licensing costs are always referenced as PER CPU not CPU CORE.
Re:Why Niagra will suck (Score:5, Informative)
"For the purposes of counting the number of processors which require licensing, a multicore chip with "n" processor cores shall be counted as "n" processors."
I guess they have it in for Sun.
Re:Why Niagra will suck (Score:2)
Bastards.
Re:Why Niagra will suck (Score:3, Informative)
BTW, it's not jusr the database licensing but also 9i RAC licensing ($20K/CPU list).
No wonder IBM's Power 4 looks so much better. Plus it comes with a non-lame volume manager and filesystem (no Veritas tax).
I'm not an IBM shill...just saying that Sun is at a serious competitive disadvantage for big databases.
Re:Why Niagra will suck (Score:3, Interesting)
Re:Why Niagra will suck (Score:2)
All I know is the little box prices.
Oracle licensing sux, the people that figure it out at Oracle are obviously insane. I don't think they have any idea about technical realities. Or, possibly, they want to destroy Sun so that they can produce 1 fewer port o
Re:Why Niagra will suck (Score:3, Interesting)
Maybe an Oracle techie/insider can debunk these rumors coming from the sales droids.
1) 10g was made for linux, all other versions are a "port"
2) At Oracle you need Larry's signature on the PO if you want to order a Sun box.
If either of these rumors are true, that's pretty bad news for Sun.
Re:Why Niagra will suck (Score:2)
Re:Put on the Sun glasses, you're blinded. (Score:2)
What irks me is that the multi-core cpu is just another technique to improve performance. Oracle is forcing their clients to freeze their database hardware at 1995 perfomance levels or face huge license fees.
I for one am going to use MySQL's cluster option instead.
An Interesting Plan (Score:3, Interesting)
It's ironic to see how positions have changed. Intel and AMD are developing multi-core CPUs for use in 4+ way systems, while Sun develops a CPU that is SMP incompatible. Of course Sun is also working on Rock, and hoping it can compete with a Xeon as a single cpu, while still scaling for 100 CPU Infernos (or whatever they are going to call them.)
Re:Imagine a beowulf cluster of... (Score:1)