Weekly Edition
Return to the Distributions page
| |||||||||||||
|
| |||||||||||||
SUSE Linux Enterprise Real Time 10 released
Novell has announced the availability of SUSE Linux Enterprise Real Time 10, "the only open source, enterprise-class real-time operating system available in the market today." "Enhancements to SUSE Linux Enterprise Real Time 10 include the latest enterprise-hardened open source technologies that reduce system latency or delay and improve predictability, such as CPU shielding, priority inheritance, sleeping spinlocks, interrupt threads, high-resolution timers and the latest OpenFabrics Enterprise Distribution for commodity high-speed interconnects, OFED 1.2.5. As a result, customers gain time advantage over competitors to make more money or avoid financial losses."
(Log in to post comments)
Marketingspeak Posted Nov 28, 2007 16:02 UTC (Wed) by epa (subscriber, #39769) [Link] Why does every press release or product description have to include useless filler like '...which helps you meet your business objectives' or '...giving your business a competitive advantage' or, in this case, 'make more money'? Does that influence anyone? Even if the answer is yes, there is no need for LWN to quote such nonsense verbatim.
Marketingspeak Posted Nov 28, 2007 19:11 UTC (Wed) by Tar (subscriber, #2456) [Link] Straight out of Bullshit Generator.
Marketingspeak Posted Nov 28, 2007 20:24 UTC (Wed) by johnkarp (guest, #39285) [Link] Isn't the #1 rule of sales is to tell prospective customers what they want to hear? That seems to be what they're doing. The audience in this case is management types, so they tell them what they want to hear, such as 'make more money' and 'competitive advantage'. The technical filler like "CPU shielding, priority inheritance, sleeping spinlocks, interrupt threads, high-resolution timers" is there merely as supporting points. I do agree that lwn should focus more on the technologies involved, rather than the marketing.
Realtime for the Enterprise? Posted Nov 28, 2007 16:32 UTC (Wed) by alex (subscriber, #1355) [Link] I'm confused. What problems does this solve for Enterprise servers? Are you going to start setting database tasks to RTPRIO? Real Time has it's uses in the constrained life of the embedded segment. Arguably it's also useful for audio set-ups and other desktop uses. However it doesn't increase throughput (in fact it invariably reduces it). If your enterprise server can't keep up with the load you make it beefier by adding processors/memory/whatever. All you'll achieve by setting a service to real time priorities is your ability to login via ssh disappearing once your complex app goes into an infinite loop without any scheduling.
What problems it solves Posted Nov 28, 2007 16:45 UTC (Wed) by corbet (editor, #1) [Link] See, for example, the Credit Suisse portion of the 2007 kernel summit customer panel discussion. These folks are counting milliseconds so they can get their trade orders in before somebody else beats them to it.
What problems it solves Posted Nov 28, 2007 17:15 UTC (Wed) by alex (subscriber, #1355) [Link] I'd forgotten about that article. Thanks John. I'm guessing the sort of apps they are running with these realtime priorities are well bounded dedicated applications. The mind boggles at the reaction time needed for the markets. I can't help but wonder how much money could be lost by two sympathetic market AI's getting into a bad feedback loop before someone could pull the plug.
What problems it solves Posted Nov 28, 2007 19:52 UTC (Wed) by nix (subscriber, #2304) [Link] How much could be lost? A lot. There is speculation (backed by rigorous guesswork as is typical with the markets) that this sort of feedback loop helped cause a number of recent funancial ructions, including the February falls (at least I think it was February). Lots of very similar models all making similar decisions at blinding speed, whoops... (The more recent mortgage meltdowns are different, though, violations of the much older rule `don't loan lots of money to people who can't pay it back' and the relatively new but in retrospect bleeding obvious one `diversifying risk by splitting it into lots of tiny packets and bundling lots of those packets together doesn't help if those packets share a lot of properties so that something can knock a lot of them over at once'. But, of course, it would take an unlikely entirely hypothetical event like, say, a rise in interest rates to do that.)
More on rt and finance Posted Nov 29, 2007 4:55 UTC (Thu) by dmarti (subscriber, #11625) [Link] Brian Stevens from Red Hat also mentions financial customers (podcast) as a target for realtime.The main userspace stuff that depends on this today is proprietary stuff such as Tibco and IBM MQseries (Websphere MQ), but there's an open source project around real-time messaging too: Apache QPid that's based around in-house code from one of the big financial firms.
Performance/latency tradeoff Posted Nov 28, 2007 17:36 UTC (Wed) by epa (subscriber, #39769) [Link] Notable from that article is that according to Credit Suisse, the realtime patches improve both latency and performance for their app.
Performance/latency tradeoff Posted Nov 29, 2007 0:55 UTC (Thu) by djabsolut (guest, #12799) [Link] Do you mean reduce latency and improve performance? Sorry for the nitpick, but the "improve xyz" phrase, where xyz is actually better off being reduced, is popping up a little too often lately (e.g. "improve emissions", "improve fuel consumption" ...)
Performance/latency tradeoff Posted Nov 29, 2007 5:27 UTC (Thu) by njs (subscriber, #40338) [Link] Err, the opposite of "reduce" is "increase", not "improve". "improve" means "make better", and indeed, for the examples you cite, smaller is better.
Performance/latency tradeoff Posted Nov 30, 2007 2:44 UTC (Fri) by djabsolut (guest, #12799) [Link] We seem to have a different understanding of the grammar here. By using a phrase such as "makes better fuel consumption" it is not immediately clear as to whether the consumption is increased or decreased. For all we know, in a different context, increased fuel consumption can actually be better (e.g. in the context of say a olympic cyclist sprinting, faster fuel consumption usually means faster speed). Anyway, this is all becoming rather tangential...
SUSE Linux Enterprise Real Time 10 released Posted Nov 28, 2007 16:39 UTC (Wed) by jengelh (subscriber, #33263) [Link] Pick a openSUSE 10.3, slap a kernel-rt on it (there are even rpms for it!). If that is going to have the same effect as NSLERT10, save me the enterprise talk.
a few dollars more Posted Nov 28, 2007 17:21 UTC (Wed) by ccyoung (guest, #16340) [Link] let them make their money opposite of the "if you need to ask you can't afford it" - if you know to ask you don't need it.
SUSE Linux Enterprise Real Time 10 released Posted Nov 28, 2007 18:40 UTC (Wed) by axboe (✭ supporter ✭, #904) [Link] Good luck getting support for that.
SUSE Linux Enterprise Real Time 10 released Posted Nov 28, 2007 23:43 UTC (Wed) by garloff (subscriber, #319) [Link] The beauty of open source, no? You can always get the software for free, use it, adapt it, try it out, whatever you like. That's great and that's why many of us are in. Distributors live on offering you services around it, like maintenance and support. If you find these useful (and many many enterprise customers do), then you pay for it. You might decide to pay anyone who promises to deliver this though for many the natural choice is the distributor who has assembled and developed the software and thus can be expected to be skilled to provide the best services. The RealTime offer is in no way different than the other enterprise Linux offers for RedHat or Novell.
SUSE Linux Enterprise Real Time 10 released Posted Nov 28, 2007 20:08 UTC (Wed) by xkarel (guest, #49324) [Link] What a nonsense marketing speak. What about RTEMS, eCos and others?
SUSE Linux Enterprise Real Time 10 released Posted Nov 28, 2007 20:44 UTC (Wed) by jondkent (guest, #19595) [Link] If I remember correctly SUSE real time is the rt-kernel patch _plus_ additions from Concurrent. So its not so simple as adding a rt-kernel to bog standard SUSE and there you go. I worked in this area and saving msec in latency is becoming a serious hot topic. The latest stupid term to be knocked about is ultra-low latency, whatever that really means, but next year this _is_ the area that most financial trading systems are going to be concentrating on. I actually prefer Red Hat's approach to all of this, its more bespoke and sensible with less "look at us" that SUSE has. But both are relevant. Jon
SUSE Linux Enterprise Real Time 10 released Posted Nov 29, 2007 10:47 UTC (Thu) by trochej (guest, #35052) [Link] Don't forget OpenSolaris (it has RT scheduling class for years now) and QNX (recently partially opensourced?). This the availability of SUSE Linux Enterprise Real Time 10, "the only open source, enterprise-class real-time operating system available in the market today." is bullshit that stinks mile saway.
SUSE Linux Enterprise Real Time 10 released Posted Nov 29, 2007 14:01 UTC (Thu) by corbet (editor, #1) [Link] The existence of a realtime scheduling class (something which is quite easy to implement) should not be confused with the ability to provide realtime response guarantees (which is rather harder). Linux, too, has had realtime scheduling classes for a long, long time.I'm surprised by the responses to this announcement. Novell/SUSE is offering a product which some of its customers will doubtless find useful - there is a market for this kind of thing. As more and more business is done through electronic agents, I suspect this market will grow. So there's something useful here; the rest is just standard press release babble.
SUSE Linux Enterprise Real Time 10 released Posted Nov 30, 2007 9:20 UTC (Fri) by trochej (guest, #35052) [Link] What I was implying is that Solaris is really a soft RealTime system. If you start processes with RT priorities they are really RT. Solaris doesn't aim for hard RT and I believe that SuSE doesn't either. If you want more information about Solaris RT capabilites Solaris internals is a good source. I am not surprised. I remember that Novell/SuSE did some considerable work and that there is a market for such product. I just dislike a marketing bullshit. And this is a marketing bullshit.
|
|
||||||||||||