* Posts by Kebabbert

808 publicly visible posts • joined 22 Jul 2009

You won't find this in your phone: A 4GHz 12-core Power8 for badass boxes

Kebabbert
Happy

Re: SPARC is not even competitive to Power

"....Clearly the Oracle marketing team had an extra shot of espresso today. Statements about SPARC outperforming Power7/7+, copying SPARC features, SPARC is significantly less expensive than Power, oh yal and the old reliable "Oracle software licensing cost is 2X more on Power, etc, etc is laughable. These statements are misleading at best...."

Oh, this sounds like good old IBM FUD. Lots of unsubstantiated negative claims about a competitor, with no proofs - this is actually the very definition of FUD. Spreading negative rumours which have no bearing in reality. "Yes, I heard this horrible rumour, but I can not prove it is true" - because it is not true.

So, if you do claim that SPARC does not outperform POWER7, or SPARC is not cheaper than POWER - can you show us some hard facts that prove your stand point? Show us some performance benchmarks. And show us some pricing comparisons. Come on, I dare you! :)

If you can not show any hard proofs (which you can not because we are telling the truth, just check the benhcmarks we posted) then it is the IBM camp that is a bit over ambitious again. :)

I remember one of the diligent IBM supporters here, who has not been active as of late. I said that the x86 is faster than POWER6 in LINPACK benchmarks, in response he said "no, it isnt. The POWER6 is faster in LINPACK because the POWER6 has faster cores" or something weird. I cant recall the logic he used because it was wrong, I cant think like that. I pointed out that you need two POWER6 cpus to match one Intel Xeon in LINPACK, but he said "well, the POWER6 has faster cores, so it is faster in LINPACK". Sure, the Xeon had six cores, and scored twice as high and if you count core wise the POWER6 was faster, true. But that does not make the POWER6 the faster cpu, does it? I dont understand the logic IBMers use.

I also said "in this benchmark the SPARC T2 has higher throughput than the POWER6 system" to which he replied "it doesnt matter, the POWER6 has lower latency which is the important thing!". Later I showed another benchmark where the SPARC T2 had lower latency, to which he replied "it does not matter, because the POWER6 has higher throughput, which is the important thing!". I really dont understand the logic IBMers use.

And now here comes this "PowerMan@thinksis" claiming lot of weird stuff, without providing any benchmarks nor pricing examples. Out out the blue he says we are all wrong, without pointing out the errors. "Trust me on this, but you are wrong, I can not tell you where you are wrong, but I know you are wrong. I am a doctor, trust me". I really dont understand the logic IBMers use. Or, lack of logic that IBMers use. Or just marketing aggressiveness the IBMers display.

BTW, have you heard that AIX is going to be killed off? It will supposedly happen sometime in the future when cheap Intel x86 cpus are catching up on expensive POWER cpus. Because IBM only does high margin business, and if the x86 are cheaper than POWER with same performance, then IBM will shut down POWER and AIX too. AIX runs only on POWER, and without POWER servers AIX can do nothing. So the POWER future looks grim. Yes, it is true, I really heard it. I am talking out of the blue? I am spreading false rumours that I just made up, or can I prove this? Well, read it here yourself and see I am not spreading false made up rumours, like the IBM camp does. No, instead I back up my claims with hard undeniable facts, straight from IBM.

http://news.cnet.com/2100-1001-982512.html

"...Asked whether IBM's eventual goal is to replace AIX with Linux, Mills responded, "It's fairly obvious we're fine with that idea...It's the logical successor." A replacement "won't happen overnight," Mills said, but years of experience designing operating systems at IBM and other companies means developers know just where Linux needs to go. "The road map is clear. It's an eight-lane highway."

Kebabbert

Re: Catching up on SPARC T5

"...Software is licensed per core and even the old Power7 has more than twice the performance per core of the new T5..."

Yes, I do not rule out the possibility that POWER7 might be faster per core, on _some_ benchmarks, yes. However, on database workloads the SPARC T5 is faster both cpu wise and core wise. In fact, the SPARC T5 is the worlds fastest cpu today. For instance, in SPECint2006 and SPECfloat raw cpu power the SPARC T5 is faster, just check the benchmarks. The SPARC T5 is also 2.4x faster than POWER7 per cpu in real life TPC-C benchmarks, not only faster in theory.

The T5 has 16 cores, and the POWER7 has 8 cores. So if the T5 had the same oomph per core, T5 cpu would be 2x faster. But the T5 is 2.4x faster, which means the T5 cores are way faster than the POWER7. So, there might be some benchmarks where POWER7 is faster core wise (I dont know which, though). And there are benchmarks where T5 is faster core wise, including databases.

The SPARC T5 beats the POWER7, core for core in database workloads. So you dont need to license as many cores as you need on POWER7. Plus you have twice the number of cores in T5, so if you need even more performance you can get it, leaving POWER7 far behind. So the SPARC T5 hardware is way cheaper than POWER7, and the SPARC T5 is also cheaper in license costs.

What is best? A quad cpu with 4 strong cores, or a cpu with 64 cores, each core slightly slower than the quad core cpu? The 64 cpu will be faster, because it has more cores.

Kebabbert

Re: not a mention on instruction set

"...After a bit of research there seem to be a big step forward: IBM added hardware support for transactional memory! ..."

Tthe last in line. SPARC T5 was the first commodity cpu with transactional memory. The old Sun ROCK cpu also had transactional memory, but it didnt make it to delivery. But all that old ROCK research has found its way into current SPARC cpus. Intel Haswell has transactional memory too. But it is nice that IBM tries to catch up on other cpus.

Kebabbert

Catching up on SPARC T5

Well it is good that the POWER8 is up to 2.5x faster than the POWER7, in theory. The SPARC T5 is up to 2.4x faster than the POWER7 in real life benchmarks today:

https://blogs.oracle.com/BestPerf/entry/20130326_sparc_t5_8_tpc

When will the POWER8 be released? Next year? Same year as Oracle release the SPARC T6? The SPARC T6 servers will be twice as fast as the SPARC T5 servers. Oracle has just announced the M6 cpu on Hot Chips. And Fujitsu has also announced a new SPARC64 cpu there too.

Also, Intel's latest Xeon E7 cpus are faster than POWER7 in some benchmarks. So the POWER8 can replace up to 2.5 Intel cpus. (IBM claims the the POWER8 can replace up to seven Intel cpus, IBM marketing is a bit over ambitious)

.

It would be really funny if IBM focus on performance talking about POWER8, "it is so 2000ish to talk about performance. No one cares about performance anymore", said IBM. When Oracle released the SPARC T5 and crushed all other cpus, IBM said this:

http://www.forbes.com/sites/oracle/2013/04/02/big-data-performance-speed-cost/

“[Performance] was a frozen-in-time discussion,” Parris said in an interview Wednesday. “It was like 2002–not at all in tune with the market today. Companies today, Parris argued, have different priorities than the raw speed of chips. They are much more concerned about issues like “availability”–resistance to break-downs–and security and cost-effective utilization of servers than the kinds of performance numbers Ellison throws out."

So, when IBM could not compete anymore, IBM said no one is interested in performance. But when POWER8 finally catches up on SPARC T5, will IBM start to talk about performance again? That would be hilarious. :)

.

I am doubting the POWER6 bandwidth numbers presented here. Back in the days, the POWER6 had 240GB/sec band width or so, according to IBM. I dug a bit, and found out that IBM had added all bandwidth in the cpu, L1 bandwidht + L2 bandwidth, etc. But you can not do like that. If there is a bottleneck of 1GB/sec in the cpu, the bandwidth will never be greater than 1GB/sec. This is a famous theorem in discrete math, called the "MAX FLOW = MIN CUT". It says, the greatest achievable bandwidth is the bottleneck. So, IBM marketing was a bit over ambitious. :)

It reminds me of when IBM claimed one Mainframe with 24 cpus can replace 1.500 of the x86 servers. It turned out that all x86 server idle at 1-2% and the Mainframe was 100% loaded! Also, one of the Mainframe cpus are way slower than a decent x86, so in no way it can replace a single high end x86 server doing work. IBM marketing was a bit over ambitious in this case too. :)

.

Funny how POWER is getting more similar to SPARC for every iteration. I remember when IBM mocked the new and radical SPARC Niagara cpu. Back in those days, all cpus had 1-2 cores. SPARC Niagara _started_ at 8 cores and went upwards. IBM said CPUs are better if they have one or two strong cores, than many lower clocked cores. So, POWER6 ran at 5GHz with 2 cores. And there was talk about future POWER cpus running even higher, 6-7GHz or more. Just like Intel Pentium4 prescott.

But today the POWER8 and all the other cpus (Intel, etc), all have many lower clocked cores and we dont see POWER cpus consisting of one single strong core anymore. That race is over. No more pushing one core to sky high Hz. So, it seems that IBM finally understood the way forward is many lower clocked cores. The same path Sun treaded 10 years earlier. If you push a cpu to very high Hz, it will use too much power, so the only way to keep power usage at bay, is to lower Hz. But in that case, you must compensate with more cores. So POWER8 has 12 lower clocked cores, each running 8 threads. Quite similar to SPARC T5 (16 cores each running 8 threads). There are no 1 core POWER8 running 2 threads at 8-10GHz, that was a dead end.

Kebabbert

Re: Putting up the good fight...

"....Unbelievable after all this time. IBM did pretty well with the Cell processor (256k core speed RAM next to each SPE), but we seem to have gone backwards since then...."

Well, you need thirteen (13) CELL cpus @ 3.2GHz, to match one single SPARC T2+ @ 1.6GHz in string pattern matching benchmarks. The CELL benchmark was done with heavy optimization in asm, loop unrolling etc. The SPARC developer did a school book implementation in C, and still was 13x faster. You need in total, 13 x 3.2 GHz = 42 GHz in aggregate power to match one 1.6GHz SPARC. The SPARC T2 uses it's GHz over 20x more efficient than a CELL cpu. Do you consider this fact, as CELL being good? Maybe that was the reason that CELL is discontinued and there are no CELL2 cpu. If CELL really was successfull, it would have lived. But no.

Oracle revs up Sparc M6 chip for seriously big iron

Kebabbert
Happy

Re: sales drones on here aside

"...Wow somebody doesn't understand the market worth a damn. The total market for proprietary UNIX boxes is less today than SUNs SPARC sales in the late dot com era by a significant amount...."

I agree on Unix is diminishing. But i distinguish between Unix servers (where Linux is getting in, however only for low end servers because there exist no high end Linux servers for sale), and database servers. Oracle is starting to create extremely fast database servers, that happen to run Solaris and SPARC. And all Oracle Enterprise customers running huge Oracle databases, will be very interested in the fast Oracle database servers. They might not be interested in SPARC, but they are interested in the database servers (which happens to use SPARC and Solaris). So, if a tiny fraction of all database customers wants to get better database servers, they must switch to the database servers from Oracle (using SPARC and Solaris).

:)

Kebabbert

Re: 256 socket Xeon

"....Yes you're talking about it, but no I'm afraid you don't know the difference between SMP and NUMA. Lets drill a bit deeper into your example, the M9000....Actually, we can start with the diagram on page 22 of the M5000, and the following sentence that says: "SPARC Enterprise M8000 and M9000 servers feature multiple system boards that connect to a common crossbar."

If you have a design where sockets on a system board only have access to limited local memory, and must traverse an interconnect, like a crossbar, to access memory on another system board, then that is a NUMA, or NUMA derived design. It's most certainly not SMP. An SMP design is where all CPUs have equal access to to all memory. The problem with that is it doesn't scale well, hence the reason why NUMA was invented...."

Yes, I do know all this. I was the one talking about NUMA and SMP, wasnt I? It seems you claim no 32 SMP servers do exist. If that is true, then maybe you accept that no Linux 32 cpu SMP servers exist. So again I am correct: there are no 32 cpu linux SMP servers.

The M9000 is not a true SMP, I know. But Sun worked hard to make it act like SMP. This manifests in that memory latency is quite bad on the M9000, but the latency is not that catastrophically bad. The latency is quite tight, with a small spread between best case and worst case. A true SMP server would have no difference, there would be no best case nor worst case latency. So, in effect the M9000 server is SMP.

If we look at a true NUMA system, such as the 8192 core Linux ScaleMP server with 64TB RAM. This server is a cluster running a single image of Linux. And like all clusters it has a very wide spread between best case and worst case latency:

http://forums.theregister.co.uk/forum/1/2011/09/20/scalemp_supports_amd_opterons/

"...I tried running a nicely parallel shared memory workload (75% efficiency on 24 cores in a 4 socket opteron box) on a 64 core ScaleMP box with 8 2-socket boards linked by infiniband. Result: horrible. It might look like a shared memory, but access to off-board bits has huge latency..."

So it does not really matter if a server is a mix of NUMA and SMP, if the latency is good (because the server is well designed). If a NUMA server had extremely good latency, it would for all intents and purposes act as a SMP server, and could be used for SMP workloads.

-The Sun M9000 has 500ns as worst case latency. And best case... maybe(?) 200ns or so. The M9000 did 2-3 hops in worst case, which is not that bad, you dont have to consider it as a problem when programming. In effect, it behaves as a SMP server.

-A typical Linux NUMA cluster has worst case... something like 10.000ns or even worse. The worst case numbers were really hilarious, and made you jump in your chair (was it even 70.000ns? I dont remember but it was really bad, the worst case numbers were representative for a typical cluster). In effect you can not program a NUMA cluster like it is SMP, you need to program differently. If you assume the data will be quickly accessed, and the data is far off in a Linux cluster, your program will grind to a halt. You need to allocate data to close nodes, just like cluster programming. And if you look at the use cases and all benchmarks on all Linux NUMA servers, they are all cluster HPC workloads. No one is used for SMP work.

This Oracle M6 server is an island of SMP servers, connected with NUMA connection. I am convinced Oracle is building on the decades of experience from the Sun server people, so the M6 server has very small difference between best and worst case latency. It will act like a SMP server, because databases are typical SMP workloads, and Oracle cares strongly about database servers. The Oracle M6 server will be heavily optimized to make sure you dont have to make more than 2-3 hops to access any memory cell in the entire 96TB RAM server - it acts like a SMP server fine for databases and other SMP workloads.

I suggest you study the RAM latency numbers for M9000 and for all Linux NUMA clusters. The differences are huge. 500ns worst case, vs 10.000s ns or was it 20.000ns?? One can be programmed like a SMP server, the other needs to be programmed as a cluster.

So, you are wrong again.

Kebabbert

Re: 256 socket Xeon

>>"...But Linux scaled awfully bad on 64 cpus...."

>"That used to be the case several years ago, but is not the case today. That's old FUD."

Well, it was true a few years ago that Linux scaled awfully bad on the 64 cpu HP Itanium Superdome (or was it Integrity). Linux had something like ~40% cpu utilization on the HP-UX "Big Tux" server in official benchmarks from HP. So it was not FUD back then, it was true. FUD is basically: a negative lie with the purpose to thrash talk a product. But the bad scaling of Linux never was a lie, it was a fact. And facts are not FUD. You seem to believe that negative critisicism is FUD, but it is not. If the critiscism is true, and negative - it is relevant for the discussion.

You claim that Linux bad scaling was not bad, it was in fact good, but nobody would call ~40% cpu utilization good. So why are you trying to make it look like Linux scaled well back in the days? Linux never did. Why are you FUDing about this? I would never make up things like you seem to do.

And you claim that Linux scales well today. That bad scaling was old FUD. Well, do you have any benchmarks on 32 or 64 cpu Linux servers, that proves Linux scales better today? No you have not. Because there are no 32 cpu Linux servers for sale. Sure, you can compile Linux to the HP server, or to the SPARC server (Linux compiles for SPARC), but those are HP-UX and Solaris servers. And I also suspect that Linux scaling would be way bad on those. Solaris is fine tuned to scale well on this 96-socket monster. Linux is not.

So, you claim that Linux scaled well back in the days, because the rumours of "bad scaling was FUD". But Linux did not scale well back in the days, according to official HP benchmarks. This was true, and it is still true today. Why do you lie about this?

You now also claim that Linux scales well today, how do you know this? On what do you base your wishful thinking? There are no 32 cpu Linux servers to benchmark, so no benchmark does exist. Are you lying about this, too?

Kebabbert

Re: sales drones on here aside

I explained this three years ago, and write it again. Sun had 30.000 customers. Oracle has 340.000 customers. Enterprise customers. If Oracle can make a small fraction of them switch to SPARC and Solaris, then SPARC will be more wide spread than ever under Sun.

And Oracle is working hard to make Oracle Database run best on Solaris. The TPC-C world record for a single system is 8.3 million tmpc on a 8-socket SPARC T5 server. That is only on 8 sockets. Imagine running the Oracle database on this M6 monster, biggest and baddest on the market. With 96TB RAM and compressed RAM, it will give extreme performance. And Oracle will fine tune it to run databases.

Solaris had recently a tuning so it decreases latency of the Oracle DB with 17%. Oracle controls everything, hardware and the OS and Java and the Database, the entire stack, so Oracle can fine tune the entire system to run databases faster than anyone else. Certainly faster than Linux on 8-socket servers. And this control will continue to show in even better database performance. Oracle has only started.

No, I am not affiliated in any way with Oracle. I just happen to like good tech, I am a tech geek. If IBM had better tech, I would like that too. I liked the POWER7 ealier, it was really good back in the days.

Kebabbert

Re: 256 socket Xeon

"...No 32-way SMP Linux Server? Maybe not from Intel or Oracle but here's your Server from IBM (and that's no new Server btw.)..."

No, the IBM P795 is an AIX server. IBM or someone else, might have compiled Linux for it. But the P795 is an IBM AIX server. I doubt anyone of those huge extremely expensive P795 are using Linux. If you want Linux, you can get a cheap 8 socket server. No need to spend truckloads of money. Or you get a Linux cluster with 1000s of cores such as the SGI NUMA server.

There are no vendor designing and selling Linux 32 way SMP servers, there never has been such a big Linux server for sale. Compiling Linux to this huge 96 socket M6 server, does not make it a Linux server. No one sane would run Linux on it. It is designed to work with Solaris, which can handle the extreme scalability with almost 10.000 threads. Oracle sells Linux, but I doubt they will offer Linux on this huge M6 server.

HP compiled Linux for their... Superdome(?) or was it Integrity(?) server with 64 sockets and called it "Linux Big Tux". But Linux scaled awfully bad on 64 cpus. Later when HP started to offer Linux on that HP-UX server, it could only run Linux in a partition, and the largest Linux partition was... 16 cpus I think. Or was it 8 cpus? I dont get it. Sell a 64 socket server, and only offer a Linux partition with up to 16 cpus. If Linux could handle 64 cpus, it would have been offered. But probably larger partitions than 16 cpus would be too troublesome with too high a support cost. But still, the Superdome/Integirty is designed for HP-UX and is a HP-UX server. Linux on it, is an after thought and does not work well.

I wonder if the IBM P795 also runs Linux in partitions of up to 16 cpus, and no larger? Wouldn't surprise me.

BTW, the SGI Altix and SU servers are clusters. They are using NUMA. And NUMA is regarded as a cluster. NUMA is the same thing as Cluster. Read here if you dont believe me:

http://en.wikipedia.org/wiki/Non-Uniform_Memory_Access#NUMA_vs._cluster_computing

Kebabbert

Re: 256 socket Xeon

"...You also need to learn the difference between NUMA and SMP. You won't find anything anywhere that's SMP at 32 sockets, they are all ( various flavours of ) NUMA. Current SMP designs hit a brick wall at 8 sockets and the M6 is no exception, as this article clearly states...."

I agree that the Oracle T5 and M6 SMP servers are up to 8 sockets. I have never said anything else. The M6 server will glue several SMP servers via NUMA, yes.

But earlier SPARC servers was also SMP, for instance the Sun M9000 with 64 cpus. So, yes, I know the difference between SMP and NUMA, I am talking about it, am I not?

Kebabbert

Re: 256 socket Xeon

The SGI server is a pure cluster used for parallel HPC work, just study the case studies.

http://www.sgi.com/products/servers/uv/resources.html

The benchmarks are all parallel workloads, for instance SPECjbb2005 benchmark, it says "32 blades" in the picture. Those 32 blades are connected via a 7.5GB/sec NUMA link, which does not provide too much bandwidth, so it does not scale too well. I wonder how many hops there are worst case in this NUMA link? Latency could be quite bad, slowing everything down.

http://www.sgi.com/products/servers/uv/benchmarks.html

Here is another Linux cluster, the ScaleMP server. It too has 1000s of cores. And it use a single image Linux kernel. It uses a software hypervisor that tricks Linux kernel into believing it is running on a SMP server, not a HPC cluster:

http://www.theregister.co.uk/2011/09/20/scalemp_supports_amd_opterons/

I dont expect these Linux HPC clusters to run SMP workloads, like the Oracle server can do. Ive yet to find a 32 cpu Linux SMP server for sale, there are no Linux SMP servers for sale with 32 sockets. And has never been. But there are large 2048 core Linux clusters for sale, as these HPC servers show. But no 32 cpus.

Fujitsu to push 28 nanometer limits with Sparc64 X+

Kebabbert

Re: why would anyone buy this old technology?

Well, maybe you know that SPARC T5 is 2.4x faster than the POWER7 in TPC-C benchmarks?

https://blogs.oracle.com/BestPerf/entry/20130326_sparc_t5_8_tpc

The new POWER8 will be up to 2.5x faster than the POWER7, that is, only in par with SPARC T5. Next year the SPARC T6 servers will be 100% faster than todays T5 servers. And before the T5, the T4 was already the fastest cpu in the world, in some benchmarks.

SPARC T4 servers where the fastest in the world, in some benchmarks.

SPARC T5 servers are twice as fast as the SPARC T4 servers.

SPARC T6 servers are again, twice as fast as the SPARC T5 servers.

And the SPARC T5 is the worlds fastest cpu today. Much faster than the POWER7 cpu. The SPARC T5 crushes the POWER7 in cpu benchmarks, among other official benchmarks:

https://blogs.oracle.com/BestPerf/entry/20130326_sparc_t5_speccpu2006_rate

So, let us rephrase the question, why would anyone buy a 10x more expensive/performance IBM POWER server? Yes, IBM is up to 10x more expensive the SPARC servers. The IBM customers are being wildly overcharged for the performance they get.

If you are doing databases (which most companies do) the choice is clear, the fastest database servers in the world, are the Oracle SPARC servers. Just check the official benchmarks.

Ex-Sun Micro CTO reveals Greenbytes 'world-beating' dedupe

Kebabbert

Greenbyte uses ZFS + OpenSolaris

Greenbyte has written the dedupe engine in ZFS. According to home users, the dedupe engine in ZFS is not that good, actually. Its main disadvantage is it eats RAM, something like 1GB RAM for each 1TB disk is recommended. Maybe this is normal requirements though for dedupe? Maybe all enterprise storage vendors that offers dedup always have plenty of RAM in their solutions? Anyway, for a home user, ZFS dedupe is not recommended, unless you have plenty of RAM. ZFS in itself works fine with 4GB RAM, with that little RAM, a very small diskcache ARC will be used, which degrades performance to disk speed - which is less optimal.

http://www.theregister.co.uk/2012/10/12/greenbytes_chairman/

NASDAQ halts stock trading, citing data-feed glitch

Kebabbert

Re: The horror

What kind of trading system did you see the code for? Not the NASDAQ system? What was the name of the system?

EMC, you big tease! At last, the specs for million-IOPS VNX2

Kebabbert

Nexenta does 1.6 million IOPS

Nexenta, the OpenSolaris distro running ZFS, does 1.6 million IOPS. I dont know the price, but I suspect it is quite low

http://www.marketwatch.com/story/nexenta-systems-powered-storage-solution-achieves-16-million-iops-2013-03-19

Kebabbert

Re: nothing special

ZFS does dedupe, but it is not that good. For instance Tegile and Greenbyte use OpenSolaris + ZFS with a rewritten proprietary dedup engine to power their enterprise storage servers:

http://www.theregister.co.uk/2012/06/01/tegile_zebi/

http://www.theregister.co.uk/2012/10/12/greenbytes_chairman/

GreenBytes guts its arrays, turns self into chompable doughnut

Kebabbert

ZFS + OpenSolaris

Greenbyte is using ZFS for their storage. ZFS has a deduplication functionality that is lacking in performance, so Greenbyte has rewritten the ZFS dedup engine. It would be really cool if Greenbyte would release their dedup engine into the open source ZFS world, but that will not happen I am afraid. But it would be wicked cool and make ZFS even better.

Competition is always good for us customers, I hope they succeed in putting pressure on the competitors so prices will be lowered.

http://www.theregister.co.uk/2012/10/12/greenbytes_chairman/

Torvalds frustrated at missing simultaneous release

Kebabbert

Re: Linux has bigger things to worry about

"...Note the last paragraph:

As Linux supports a larger number of different devices "out of the box" than any other operating system, and it supports these devices on more different processor architectures than any other operating system, this proven type of development model must be doing something right :) ..."

Maybe you missed all the reports of Linux upgrades breaks the install? Linux device driver model is broken. As soon Linux releases an upgrade, all hardware vendors need to modify and recompile all their device drivers. HP reportedly spends millions on this. Does this sound right to you? HP should migrate to FreeBSD instead.

Kebabbert

Re: Linux has bigger things to worry about

Your Linux install is broken after an upgrade? You are not alone. Bassbeast explains:

"Sadly friend that is ALL you will get because ultimately the broken driver model has become a religious element, a way to "prove the faithful" by how much they will get behind an obviously and demonstrably bad design.

Quick, how many OSes OTHER than Linux use Torvald's driver model? NONE. How many use stable ABIs? BSD,Solaris, OSX,iOS,Android,Windows, even OS/2 has a stable driver ABI.

I'm a retailer, I have access to more hardware than most and I can tell you the Linux driver model is BROKEN. I can take ANY mainstream distro, download the version from 5 years ago and update to current (thus simulating exactly HALF the lifetime of a Windows OS) and the drivers that worked in the beginning will NOT work at the end.

And before anybody says "Use LTS" that argument doesn't hold water because thanks to the again broken design by Torvalds most software in Linux is tied to the kernel so if you want more than a browser and Open Office? You WILL be forced to upgrade because "this software requires kernel x.xx" or be left behind with older non supported software. With Windows with the exception of games that require a newer version of DirectX (which is rare, most have a DX9 mode for this very reason) you can install the latest and greatest on that 10 year old XP machine and it JUST WORKS.

Again let me end with the simple fact that after NINE YEARS I'm retiring the shop netbox. That is TWO service packs and at LEAST 3000 patches and not a single broken driver, NOT ONE. If Linux wants to compete then it actually HAS to compete, not give us excuses which frankly math can prove doesn't work. Look at the "Let the kernel devs handle drivers" excuse. You have 150,000+ drivers for Linux, with a couple of hundred new devices released WEEKLY..how many Linux kernel devs are there again? if you pumped them full of speed and made them work 24/7/365 the numbers won't add up, the devs simply cannot keep up...which is of course one of the reasons to HAVE a stable ABI in the first place, so that the kernel devs can work on the kernel while the OEMs can concentrate on drivers.

Sorry for the length but this one really irks me, if you like running an OS that is rough because of reasons? Go right ahead, i wish you nothing but luck. But when you compare that broken mess to either OSX or Windows I gotta throw down the red flag and call bullshit, its not even in the same league. Oh and do NOT bring up embedded or servers as that is "moving the goalposts" and honestly i don't care how cool your OS is at webserving, I'm not selling webservers and that isn't the subject at hand. Linux is broken ON THE DESKTOP and that is what we are discussing so try to stay on topic.

I'll leave you with this, if one of the largest OEMs on the entire planet can't get Linux to work without running their own fork, what chance does the rest of us have?"

http://www.theinquirer.net/inquirer/news/1530558/ubuntu-broken-dell-inspiron-mini

SGI backs away from public clouds, chases HPC and big data

Kebabbert

Re: Why

"....No use arguing with Kebabbert about this - he knows that he's right. The fact that there are a variety of architectures available doesn't matter to him. The fact that SGI state that their 512 ( or whatever it it now) core systems will run SUSE or Redhat unmodified as a single system image doesn't apparently count either...."

As I said, ScaleMP has their 2048-8192 core server running unmodified Linux kernel in one single system image. But, ScaleMP is running a software hypervisor that tricks the Linux kernel into believing it is a single system image. The ScaleMP server is just a cluster. Just as the SGI server I have mentioned.

If you state that "there are a variety of architectures available", can you show us some links then? I have never seen a 32 cpu Linux server for sale. But obviously you claim they exist, several different architectures, you say. Then show us some links to those 32 cpu Linux servers, then. I am not talking about 1000s of cpus in a cluster like the SGI, I am talking about SMP servers. Who has ever seen a Linux 32 socket server for sale? I have never seen anyone. Whenever I talk about this, Linux fans always say 32 socket Linux servers exist for sale - but no one has ever showed me a link. Not. A. Single. Link. Never. Ever.

Show us the money, where are those 32 socket Linux servers? This is too weird, Linux supporters say all the time Linux 32 socket servers exist, but no one has ever showed me a link. No one. Never. Isnt that a bit weird? Every time I have requested links, but to no avail. No links. And STILL they claim such huge 32 cpu Linux servers exist for sale. I dont get it. If those servers does not exist - why in earth do they insist they exist? On top of that, they try to ridicule me, try to make me look like a ignorant computer nerd. "Dont try to argue with Kebabbert on this one, because such Linux servers exist, but I will not show you any links, and no one has ever seen such links - but it is obvious that Kebabbert is mad when he claims there are no such Linux servers, because they exist. Somewhere. I heard. But dont ask me for links. But trust me, they exist. I think. In an alternate Universe. Or on Mars."

Why are those Linux supporters like that? Do they believe in Santa Claus too? "Yes he exist, but I cant prove that but trust me. If you doubt it, you are a mad man".

Kebabbert

Re: Why

"...Be impressed, 32-CPU systems shipped years ago, I've seen a few personally. IBM sells a system with 8 6-core CPUs which came out in 2010, Sun had a 4 x 8-core system already by 2010. You could buy 12-core CPUs from AMD by 2010, and run 48 cores using a 4-socket motherboard...."

I am talking about 32 sockets. 32 cpus. Not 32 cores. I have never seen a Linux server with 32 sockets. Have you? If you have, please show us some links.

Kebabbert

Why

Don't SGI manufacture a big bad SMP server with 16-32 cpus instead of clusters with 1000s of cores or cpus? Enterprises are interested in big bad SMP servers, not clusters. For instance, the IBM P795 weighs ca 2.000kg and it has up to 32 cpus.

Until SGI sells a big bad SMP server with as many as 32 cpus, I won't be impressed. Anyone can make a huge cluster with 1000s of cores and tons of RAM. Until anyone offers a Linux server with 32 cpus for sale, I wont be impressed. The problem is that Linux can not scale beyond 8 cpus today. No one has ever offered a 16 or 32 cpu Linux server for sale. SGI could be the first to make Linux scale beyond 8 sockets? And then SGI could target the Enterprise market, instead of chasing HPC scientific number crunching specialized companies, running large Linux clusters.

Just look at the ScaleMP solution, which is very similar to SGIs cluster called Altix, UV, etc. ScaleMP also has 1000s of cores and runs a single image Linux kernel, in their large servers. But, ScaleMP uses a software hypervisor which tricks the Linux kernel into believing it is running on a single SMP server, when in fact, the kernel runs on a cluster. But using a software hypervisor is not the same thing as a running on a true SMP server. The largest SMP servers have 32 or sometimes, even 64 cpus. The largest HPC clusters have 1000s of cores and tons of RAM.

Did Linux drive supers, and can it drive corporate data centers?

Kebabbert

Re: No commodity Linux

But Lars, the SGI server with 512 cpus or more, and Linux SGI Altix servers with 2048-8192 cores, are all clusters. They are running a software hypervisor tricking Linux into running single image. They are clusters, not a single fat SMP server.

Can you show me links to a 16 or 32 cpu SMP Linux server for sale? No? Why? Maybe Linux has problems handling 16-32 cpus? Or, what is your explanation?

Linux servers have 1-8 sockets. Actually, it is Dell/HP/IBM/Oracle selling ordinary x86 servers with 1-8 sockets. and Linux servers have 2048-8192 cores and 100s or 1000s of cpus. But all those big servers are clusters. There are no 16-32 cpu Linux servers for sale. Not a single one. Never has been. Linux can not scale to 16 cpus, nor 32 cpus. Linux runs fine on clusters, such as the ScaleMP server, SGI Altix server, etc. And Linux runs not so good on 8-socket servers. But no one sells 16-32 cpu Linux servers. No one. Why? Show me a link to a 32 cpu Linux server. You cant, because Linux does not scale.

Kebabbert

Re: No commodity Linux

"...Now I'm confused. Are we talking sockets, cores, or cpus? What do YOU mean by cpu?..."

What is the difference between a cpu and a socket? I said there are no 32 cpu Linux SMP servers for sale, and has never been. Or can you link to such a server? IBM P795 has 32 cpus, or sockets. But is has 256 cores. Oracle M5-32 has 32 cpus, or sockets. HP Superdome has 64 cpus (or is it Integrity?) - I am not updated on HP servers.

My point is still valid: Linux will not venture into the big SMP server area, with 32 cpus. The old IBM P595 server with 32 cpus, costed $35 million list price. Large SMP servers are high margin business, that is where the big money is.

Kebabbert

Re: No commodity Linux

So, no. In the Enterprise they mostly use large SMP servers with up to 64 cpus, which Linux can not handle. Linux does not scale beyond 8 cpus today in a single fat SMP server. There are no 16 cpu Linux SMP servers for sale, nor 32 cpu Linux servers. Because Linux does not scale.

But there are many 2048/4096/8192 core Linux servers for sale - they are all clusters. Clusters are easy to scale to, running embarassingly parallell workloads.

So, there are 1-8 socket Linux servers for sale (HP / IBM / Oracle offers such x86 servers). And there are 2048-8192 core Linux servers for sale (SGI / ScaleMP / etc). But there are no 16/32 cpu Linux SMP servers for sale. The big money is in large SMP servers with 32 cpus, they cost millions. For instance, the IBM P595 server used for the old TPC-C record costed $35 million, list price. $35 million. I am not kidding. It has 32 cpus. If someone could manufacture a 32 cpu SMP server and make Linux scale, they would sell loads of such servers to the enterprise, investment banks, telcos, etc. They could charge a measly $1 million or so. Here is a niche, a business opportunity. But no one does that. Because Linux can not scale to 32 cpus.

Linux even has problems scaling to 8 sockets, if you look at some official benchmarks. For instance SAP benchmarks on 8 socket x86 servers, Linux had 87% cpu utilization, whereas Solaris had 99% cpu utilization. On 4 socket server, Linux reached 97% cpu utilization in SAP benchmarks. With fewer sockets, Linux performs better. Maybe on 12 cpus, Linux would have 75% cpu utilization? And on 16 socket, Linux would have 60% cpu utilization? I actually saw a benchmark on something called Big Tux, HP compiled Linux to their 64 cpu Superdome/Integrity(?) SMP server, and Linux had ~45% cpu utilization. I understand why HP does not sell Superdome/Integrity servers running Linux, the scaling would be to bad. 45% cpu utilization on 64 cpus is not something anybody would buy.

Until Linux scales better, it will not venture into the large SMP server domain in the Enterprise (vertical scaling). It will never rival IBM P795, Oracle M5-32, HP Superdome, etc. But on clusters, Linux scales very well (horizontal scaling). Google horizontal scaling, vs vertical scaling. Until Linux can not scale vertically, it will not venture into the large SMP server domain, with up to 32/64 cpus.

Kebabbert

No commodity Linux

Supercomputers are not running commodity Linux. In fact, they run a heavily modified stripped down Linux kernel, which has not much in common with a standard Linux kernel. If you can strip out everything except the bare minimal skeleton, then you can speed up the kernel say, 10%. This makes the entire supercomputer faster by up to, 10% - which is a lot in this domain. These people will do everything to speed up the kernel. It is like comparing a stripped down racing car with nothing in it (no protection, no nothing), to a commodity Volvo. Its a big difference.

For instance, IBM's supercomputer Blue Gene which officially runs Linux, does not. It uses Linux to distribute the work loads to the different compute nodes. These nodes runs a minimalistic special OS designed to just do number crunching and nothing else. So I would actually exclude IBM Blue Gene from the list, because it does not run Linux. But Linux is popular today, so everyone say they run Linux - even when they don't.

Linux running on these supercomputers are running HPC workloads, that is, large clusters running embarassingly parallell workloads (google this). This is not a evidence of Linux scaling well, but a proof of Linux being customizable. It is easy to tailor Linux to do a highly specialized task like number crunching and shut down all services and daemons, and strip out all code. It is a simple kernel. To do only one thing, is easy to optimize for. Unix are mature and complex kernels, not easy to tailor or strip or understand.

It is another thing to run a large single SMP server serving 1000s of users doing many different tasks, running many different programs. That is a very complex environment and difficult to scale well to. These large SMP servers, typically have 16 or even 32 cpus (like the huge IBM P795 or Oracle M5-32). Some have 64 cpus (Fujitsu M4-10s). But not more than that. SMP servers have 64 cpus at maximum. And there are no Linux SMP servers for sale. Because Linux can not handle 32 or 64 cpus in a SMP setting because that is difficult to scale well to.

Large servers like the SGI Altix server running Linux, are using 2048 or even 4096 cores and 64TB RAM, but it is actually a cluster, running a software hypervisor tricking Linux into believing it is a single fat SMP server. Which is not. If you study the SGI Altix UV customers, they are all running HPC number crunching workloads. No one is using such servers for SMP workloads. ScaleMP is another company doing this, using a hypervisor tricking Linux kernel into believing it is running a single fat SMP server, when it in fact is running a cluster:

http://www.theregister.co.uk/2011/09/20/scalemp_supports_amd_opterons/

"...Since its founding in 2003, ScaleMP has tried a different approach. Instead of using special ASICs and interconnection protocols to lash together multiple server modes together into a shared memory system, ScaleMP cooked up a special hypervisor layer, called vSMP, that rides atop the x64 processors, memory controllers, and I/O controllers in multiple server nodes. Rather than carve up a single system image into multiple virtual machines, vSMP takes multiple physical servers and – using InfiniBand as a backplane interconnect – makes them look like a giant virtual SMP server with a shared memory space..."

Happy 20th birthday, Windows NT 3.1: Microsoft's server outrider

Kebabbert

Re: Sun Microsystems called NT a Mainframe OS for a PC

"...Please support your assertion that Solaris is the most wide spread Unix with actual data - 10 million downloads mean nothing, especially if no timeframe is given...."

I dont have links right now. But this is so well established that I dont bother. Everybody knows it. This site sometimes presents the state of Unix, and other reports says the same. Just read them. It goes something like this: IBM made the most money (but they are selling few expensive servers). Solaris sells most servers (but cheaper). bla bla. Just read the reports.

But if you think, Solaris runs on x86, there are lot of Solaris distros on x86: Nexenta is a storage vendor company selling Enterprise servers, Tegile is a Enterprise storage company competing with NetApp, Greenplum, Coraid, etc etc - there are lot of x86 distros and hardware vendors selling open sourced Solaris products. Then we have Oracle Solaris which is closed, it is wide spread on x86. And of course, all those SPARC servers too. In comparison, POWER servers are few. And Itanium too. This is so well established I dont bother, google it yourself. Everybody knows this.

Kebabbert

Re: Sun Microsystems called NT a Mainframe OS for a PC

Solaris is the most wide spread Unix. There are something like 10 million downloads of Solaris 10, alone. It runs on x86, that is, lot of servers. It also runs on all SPARC machines.

IBM AIX runs on POWER servers. There are not many of them, in comparison. HP UX runs on Itanium, maybe there are fewer Itanium servers than POWER servers?

The most innovative Unix today, is Solaris. Everybody wants or copied Solaris tech, such as ZFS, DTrace, SMF, Crossbow, Containers, etc. You name it. For instance, let us talk about the lesser known DTrace:

-The Linux clone of DTrace is called Systemtap.

-IBM AIX clone is called Probevue

-Mac OS X has ported it

-FreeBSD has ported it

-NetApp engineers talked about porting it, on a blog post. (NetApp ONTAP is a FreeBSD derivative)

-VMware clone is called vProbes

-QNX has ported it

These are just the OSes on top of my head, that I know. If I google a bit, maybe I could add some more OSes that ported/cloned DTrace. It is almost like every major OS has got DTrace, in one way or another. DTrace is a must have, they think.

So, can you name some cool IBM AIX tech that everybody has cloned or copied? Or cool HP UX tech? Or cool Linux tech? No? There are no must-have tech? Then surely Solaris is the most innovative Unix today. And the most wide spread Unix too. Your post is in other words, totally off, split from reality.

Kebabbert

Re: humph

MS asked for Commodore to license Amiga OS? Never heard this before. Do you have links, or is just hear say?

Kebabbert

Linux graphics runs in the kernel, too.

"...The OS was very stable until NT shoved the device drivers into the kernel (with NT4) and allowed graphics to crash the box.. which is ironic because Kernel graphics is one of the reasons Linux is better for GPGPU & HPC..."

You do know that Linux today, runs it graphics in the kernel?

Kebabbert

Re: Misleading

But as of today, Windows does not run the graphics in the kernel anymore. The graphics is running outside the kernel. Windows7 is the most stable Windows Ive tried. It works well. Sure, it is not as stable as Unixes, but it is stable enough.

Strange enough, Linux has lately moved the graphics into the kernel to gain more speed. This move has made Linux more unstable.

NEC tag teams with HP on high-end x86 servers

Kebabbert

You got it wrong. Sgi is using a hypervisor to trick the linux kernel into believing it is running a single smp server, when it is in fact a cluster. ScaleMP also use the same approach, they also have a 2048 cpu linux server, but they are also running a cluster. These clusters have awful latency to nodes far away, they may look like single image, but it is not. They are only used for parallell HPC workloads. Google on ScaleMP too, and see it also uses a software hypervisor tricking linux kernel into running a single server. These servers are clusters, and unusable as smp single servers.

UPDATE: here is a link on the similar clustered server ScaleMP sells (2048 cores, 64TB RAM):

http://www.theregister.co.uk/2011/09/20/scalemp_supports_amd_opterons/

"....Since its founding in 2003, ScaleMP has tried a different approach. Instead of using special ASICs and interconnection protocols to lash together multiple server modes together into a shared memory system, ScaleMP cooked up a special hypervisor layer, called vSMP, that rides atop the x64 processors, memory controllers, and I/O controllers in multiple server nodes. Rather than carve up a single system image into multiple virtual machines, vSMP takes multiple physical servers and – using InfiniBand as a backplane interconnect – makes them look like a giant virtual SMP server with a shared memory space...."

Kebabbert

SGI UV is a cluster.

You got it wrong. Sgi is using a hypervisor to trick the linux kernel into believing it is running a single smp server, when it is in fact a cluster. ScaleMP also use the same approach, they also have a 2048 cpu linux server, but they are also running a cluster. These clusters have awful latency to nodes far away, they may look like single image, but it is not. They are only used for parallell HPC workloads. Google on ScaleMP too, and see it also uses a software hypervisor tricking linux kernel into running a single server. These servers are clusters, and unusable as smp single servers.

IBM gooses 'business class' mainframes with z12 engines

Kebabbert

Re: Why?

My point is that using software emulation (which is 5-10x slower than running native code) you get the same performance as this IBM mainframe, using an old 8-socket x86 server. Running native code would yield 5-10x higher performance for the x86 server. So, if you recompile your Mainframe application on the x86 server, you get 5-10x higher performance than this Mainframe. Which is in par with the largest IBM Mainframe.

What does this tell you about the Mainframe cpus? They are quite slow?

Kebabbert

Why?

"...With all ten engines fired up, that came to 3,139 MIPS of aggregate performance after SMP overhead was taken off the top..."

You get 3.200 MIPS under software emulation using an old 8-socket x86 server for a fraction of the price. It is much cheaper and more flexible.

http://en.wikipedia.org/wiki/Hercules_%28emulator%29#Performance

Snowden leak: Microsoft added Outlook.com backdoor for Feds

Kebabbert

OpenBSD is secure and safe

Everybody should switch to OpenBSD then. Linux has so high code turn over, which makes it impossible to catch up the audit. When you audit one line, three new lines have been added. The code base is completely rewritten every... half a year(?). In a very short period. It is like all those device drivers that need to be updated as soon as Linus Torvalds breaks them in an ABI upgrade (which happens all the time). HP spends millions of USD to update every device driver when Linux has been upgraded. And Long Term Distros does not help, because when you install a new software, it needs new updated libraries, forcing you to upgrade them, which forces you to upgrade everything else.

OpenBSD has very strong audit all the time. I suspect NSA hates OpenBSD and actively tries to diminish it. Linux suits NSA very well, with the very high code turn over.

Maybe Richard Stallman was right when he insisted on taking extreme precaution before surfing on the web. People here called him moron for doing that. Maybe RMS is clever-er than both you and me.

Oracle cranks up SuperCluster with Sparc T5 engines

Kebabbert

Re: Why are there no Sparc desktops anymore?

Correctly written Solaris software should only require a recompile between x86 and SPARC.

You can develop on x86 and recompile for SPARC. There are numerous open source Solaris distros out there, for isntance SmartOS, OmniOS, OpenIndiana, etc.

Of these, SmartOS and OmniOS seem most compelling. SmartOS is developed by Joyent, the company behind nodejs and the newly released Manta, which is similar to hadoop but with a unique twist, so it is much easier to analyse Big Data than ever before:

http://www.joyent.com/blog/hello-manta-bringing-unix-to-big-data

SmartOS also has several defected Solaris kernel devs from Sun, such as creator of ZFS, creator of DTrace, etc. They have probably the best Solaris knowledge outside Oracle. The CloudOS SmartOS uses KVM so you can run different Zones for higher security (all these zones can now analyse big data, see above)

http://www.joyent.com/technology/smartos

OmniOS seems to be very popular too.

Kebabbert

Re: hear those crickets

It is funny to see how IBMers try to tackle Oracles benchmark records.

When Oracle released the T5 SPARC servers, IBM said talking about performance is so 2000ish and no one cares about performance today.

http://blogs.wsj.com/digits/2013/03/27/ibm-fires-back-at-oracle-after-server-attacks/

“This was a frozen-in-time discussion,” Parris said in an interview Wednesday. “It was like 2002–not at all in tune with the market today.” Companies today, Parris argued, have different priorities than the raw speed of chips."

I promise that if IBM succeeds with the POWER8 (which seems unlikely because new POWER cpu generations are not that faster, whereas SPARC is 100% faster every generation) and get some records, then IBM will boast about how important performance is. Something like "with high performance you can replace many servers with only one single server, therefore performance is very important". But now that Oracle has the performance crown, IBM is saying that performance is not important anymore. *chuckles*

.

.

And IBM supporters such as "asdf" above, can not really attack the numbers, instead they attack pointless meaningless things. Such as "Oracle pretending to be a hardware company". I dont get it, why choose that attack? "asdf" makes it sound as: today Oracle has performance world records, even as they are not trying hard to get into hardware, imagine if Oracle tried hard. In other words, "asdf"'s attack is poorly chosen. Instead, I advice you, "asdf" to attack the colour of the SPARC servers or something better than this choice which is really bad. Because it makes Oracle's servers look really fast. Which, BTW, they are. Fastest in the world. So, 10x performance/price compared to IBM is real.

Kebabbert

Re: Bussiness head

Of course you can use SPARC T5 cpus for HPC, because they are the fastest cpus on earth, regarding SPECint and SPECfloat benchmarks:

https://blogs.oracle.com/BestPerf/entry/20130326_sparc_t5_speccpu2006_rate

But, it is probably better to use cheap Intel x86 servers in a large cluster such as the SGI Altix Linux servers. x86 servers are cheap and fast so you will get most bang for the buck using x86. If you study my link, you will see that x86 are faster than IBM POWER7 cpus on these cpu benches too. And cheaper.

Kebabbert

"...proclaim as fastest system on the planet...." Not really. Oracle only claims they are 10x faster/price than IBM POWER7+ systems. But it is true that the previous generation SPARC T4 servers are much faster than IBM POWER7 systems on certain benchmarks. And these T5-8 servers have double the nr of cpus, and double the nr of cores, ergo four times faster than the fastest T4 servers.

Supercomputer efficiency jumps, but nowhere near exascale needs

Kebabbert

As I have understood it,

Blue Gene are using PowerPC cpus, not POWER cpus. Because PowerPC are more tailored to low end, whereas POWER are tailored to high end where power consumption is not a big deal. For instance the IBM Blue Gene supercomputer, uses 750MHz PowerPC cpus. That was not impressive back when it was new. But the difficulty lies in using all cores effectively. No one has really succeeded in doing that effectively. Linear scaling does not exist. When / if we ever achieve linear scaling, then it makes sense to upgrade to better cpus. But today, cpu performance is not the problem. The connection is the problem. As has always been.

Blue Gene uses tailor made Linux to distribute the workload out to each node, in the cluster. Then a special OS takes over and does the actual number crunching. So I would not say that Blue Gene runs Linux, even if it listed as a Linux server.

RBS Mainframe Meltdown: A year on, the fallout is still coming

Kebabbert

Re: AC: This wasn't a hardware flaw. This wasn't a mainframe flaw.

"...While great herds of commodity servers surely have a place, it is unlikely that they have a complexity, managability, or reliability advantage over mainframes for large workloads. It is quite possible that they also have no overall cost advantage...."

Google thinks it is more cost effective and reliable to have herds of commodity servers for their services, than a few large servers. Google has designed their system so it doesnt matter if a server breaks, they just insert a new cheap server. If you have 100 cheap servers that can switch roles anytime, you will have a better uptime than a single mainframe. Distributed beats centralized both in terms of costs and reliability and performance.

But sometimes, you can not distribute. Some workloads are inherently centralized and can not be parallellized. Google has embarassingly parallell workloads, so they can just use lot of COTS servers.

For banking, Mainframes are the norm. They can be replaced by a fleet of COTS servers, but no one has done it yet, as I know of. There will be years of R&D. Maybe it is easier to just use a Mainframe.

Another note. In banking, Mainframes are the norm. They are slow, have extremely slow cpus, but good I/O. Banking is doing lot of account updates like calculate interest rate, etc. Nothing sexy. Old, boring, dusty stuff. More, accountant stuff. It doesnt matter if your account gets updated 0.1 s later or so. Latency is not important. Much work is done in batches, and can wait another hour. Mainframes have good throughput, bad latency.

In finance, you never use Mainframes. In finance, you typically use Linux/Unix. You are doing real time high performance calculations. HFT. Quant. Math. Algorithmic trading. Linux has low latency, so that is important in some fields. Mainframes are too slow to do finance as they have very weak cpus, much weaker than x86 cpus. For instance, large stock exchanges typically run Linux/Unix. No stock exchange runs on Mainframes, they have too bad latency for that.

Banking: accountant stuff, boring. Needs to be reliable and just work. No requirements for bleeding edge performance. Simple calculations.

Finance: math, high performance stuff, complex calculations. Needs to be reliable and high performance. If you are faster than any one else, you get your order filled and you can earn money.

Girls, beer and C++: How to choose the right Comp-Sci degree for you

Kebabbert

Re: Does anybody know ...?

I dont understand why you bad mouth Dominic. He tries to give good advice to help us, and what does he get in return? Sure, he displays some attitude in his articles, but they are very amusing and interesting to read. And his articles gets lots of comments, that must mean something? Regarding attitude, there are several people here sharing the same trait (myself included), so we should not focus much on that. If you dont like his articles, dont read them. I think it is interesting to read about the job market in City. More on that, Dominic! :)

Back to topic. I think we must distinguish between comp sci and engineering. If you are doing bread and butter programming you are probably an engineer. If you are doing stuff which requires theoretical knowledge (like writing a compiler, designing new algos) you are more a comp sci guy, and you must have a solid math foundation to do that. If you dont need math in your work, you are into engineering. Real comp scis often have an PhD and can tackle difficult problems. Of course, some people dont need a PhD to do that, but many have.

An comp sci degree should focus heavily on math and theory (logic, discrete math, compiler theory, AI courses, algorithms, paradigms such as functional programming, prolog, database theory, etc). That stuff never gets out of time and is indispensible for a theorist(?). During all the years there should always be one programming course rolling so the students always have one foot in programming. The last years should have some limited coverage of some applied tech, see below. Maybe 2-3 short courses.

And engineer degree should have some basic theory, lightly covered in 2-3 courses in total. And the focus should be learning applied tech such as SQL, C#/Java/C/C++, network programming, internet programming (HTML5/Javascript), maybe one functional language, etc. Basically what companies ask for. They should learn to build three layered logic servers with front end and back end.

Regarding teaching programming, in my experience you should always start with learning basic concepts first. Be simple, there is no need to go through pointers at the first course. It is better to get used to programming and the flow of logic, with if, for, etc. So I strongly believe all programming courses should start with Python, the easiest language. Large parts of the Spotify backend is developed in Python. When people knows how to program and write small programs, only then they should start with C/C++. All large software projects should avoid C/C++ in courses, because the students need to see their project deployed in time, so use a simple language. Python is so small you can keep the entire language in your head (much like C) but there is no difficult syntax. It just works. See what Eric Raymond, ("Cathedral and the Bazaar", Linux hacker) writes about Python. Very enlightening.

http://www.linuxjournal.com/article/3882

Myself focused heavily on algorithms and discrete math in my first M.Sc. I read only theoretical stuff for one of the best mathematicians in the world. He was not a good teacher though, because he is a genius and everything is simple to him. We struggled hard with the concepts and had to solve problems in the exams, which where published in refereed math journals. One friend of mine, even got an assigment which was an open problem! My friend could not solve the problem, so he asked the professor about the solution, neither of them could solve it. So the professor emailed one of the experts in the field who said it was an open problem. The classes where very difficult and my university is one of the best in the field of algorithms and discrete math, and we regularly beat Stanford, MIT, Harvard, etc in programming contests.

My second degree is in applied math. So I switched field to algorithmic trading. I could not have done this switch if I where an engineer (then I would have built the systems). Solid theory allows you to switch fields. I always studied academic subjects (maths, algos) instead of techniques (C#, Java).

AMD announces 'world's first commercially available 5GHz CPU'

Kebabbert

Re: "the world's first commercially available 5GHz CPU,"

"..Of 120 engines, up to 101 are available for processes to use (the others are reduntant, run in lockstep with others, or used for IO or similar)..."

So is one "engine" equivalent to a "core"? It would be much easier if IBM talked about "cpus" and "cores" instead of "books" and "engines".

One book is one cpu? And one engine is one core? Is this true?

Kebabbert

Re: "the world's first commercially available 5GHz CPU,"

"...Everyone knows that a 5GHz Power CPU will run circles around a 5GHz X86 equivalent...."

Well, here you see that an old Westmere-EX x86 cpu at 2.4 GHz is just ~10% slower than a 3.55 GHz POWER7 at a SAP benchmark. If you clocked the old Westmere-EX up to the same speed as a POWER7, the Westmere-EX would be 28.4 % faster.

http://www.anandtech.com/show/4285/westmereex-intels-flagship-benchmarked/3

It seems that x86 have improved very fast. Now the latest x86 is several generations newer, and faster. Whereas the POWER7 has only been upgraded one generation in the same time frame: to the POWER7+. The POWER7+ is only slightly faster than the POWER7. So, if you clocked the latest x86 up to the same speed as a POWER7+, the x86 would surely outperform the POWER cpus.

It seems that you assertion is not valid in modern times. Back then, the POWER cpus where indeed faster than x86. But today x86 has much more R&D resources than POWER has, and x86 improves faster.

(We should not mix in Oracle SPARC, because SPARC gets 100% faster for every generation. This is far better than POWER or x86. Even the four cpu T4 servers outperformed the double number of POWER7 cpus in some benches. The eight cpu T5 servers wipes the competition).

Kebabbert

Re: "the world's first commercially available 5GHz CPU,"

"The current generation (zEC12) mainframe chips are 5.5GHz; 6 core/chip, 6 chips/module. Each core can be doing 6 things simultaneously (two integer units, two load-store units, one binary and one floating point/decimal unit)."

And it has a huge cache, right?

http://en.wikipedia.org/wiki/IBM_zEC12_%28microprocessor%29

"there are 2 dedicated companion chips called the Shared Cache (SC) that each adds 192 MB off-die L4 cache for a total of 384 MB L4 cache. L4 cache is shared by all processors in the book."

So, what is a "book"? How many cpus does the zEC12 mainframe have? 24 cpus, and 4 of them are dedicated to the OS? So what is a book, then?

NetApp musters muscular cluster bluster for ONTAP busters

Kebabbert

DTrace?

NetApp engineers discussed porting DTrace to ONTAP, on a blog. Have DTrace been ported yet to ONTAP? ONTAP is based on FreeBSD, which already has ported DTrace, so it should be easy to do.

IBM to port KVM hypervisor to Power-Linux iron

Kebabbert

P795 64-cpus??

I thought the P795 has 32 sockets, right? I am quite sure it has 32 cpus. Fujitsu and HP have 64 socket Unix servers today, but not IBM.

IBM needs to reinvent to POWER servers, and by that, I dont mean to transform them to Linux servers. x86 are cheaper and have almost the same performance.

Windows NT grandaddy OpenVMS taken out back, single gunshot heard

Kebabbert

Open sourced?

I have worked with sysadmins that thinks Unix to be a bit unstable, but OpenVMS has always been rock solid. Especially the clustering is praised with uptimes of decades. I am a geek, fan of great tech. And there are people whom I respect, that says OpenVMS is better than Unix, so I am a supporter of OpenVMS. An OpenVMS cluster have far better uptime than a Mainframe they say. I have only dabbled a bit with OpenVMS at work, and the shell seemed a bit arcane. But that does not really matter if it is a great OS. I wish it was open sourced so the community could take over and make it live, and never die.