Problem with XP scheduler?
in [Win32 Kernel]
|
From: Maxim S. Shatskih on 31 Dec 2009 05:22 > in the 90s ... sequent claimed that it did much of the "windows" scaleup > work for shared-memory multiple processors (getting NT running > efficiently on more than two-processor multiple processor machines. Can you point at particular legacy of this Sequent's effort in Windows? For me, looks like that the "so-so more-or-less" SMP support was in Windows from the beginning, and the major improvements in it were done only in 2008/2008 R2 timeframe. -- Maxim S. Shatskih Windows DDK MVP maxim(a)storagecraft.com http://www.storagecraft.com
From: Maxim S. Shatskih on 31 Dec 2009 05:24 > opened and have done several hours worth of work and click back > on my firefox icon and do a refresh of about 10 pages in a rapid > succession, and my box simply freezes I think this is due to Firefox playing bad games with Windows shell, which is - in such a scenario - the only component which freezes. IE is better in this. IE crashes sometimes though - mainly in Flash.ocx. -- Maxim S. Shatskih Windows DDK MVP maxim(a)storagecraft.com http://www.storagecraft.com
From: jmfbahciv on 31 Dec 2009 08:57 Maxim S. Shatskih wrote: >> in the 90s ... sequent claimed that it did much of the "windows" scaleup >> work for shared-memory multiple processors (getting NT running >> efficiently on more than two-processor multiple processor machines. > > Can you point at particular legacy of this Sequent's effort in Windows? [inserting <CRLF>s required] > > For me, looks like that the "so-so more-or-less" SMP support >was in Windows Nope. > from the beginning, and the major improvements in >it were done only in 2008/2008 R2 timeframe. > /BAH
From: jmfbahciv on 31 Dec 2009 08:58 tanix wrote: > In article <9418f497-efe8-44b9-b9ab-71fc99cf4dc7(a)j19g2000yqk.googlegroups.com>, J de Boyne Pollard <J.deBoynePollard(a)Tesco.NET> wrote: >> DB> My real complaint with this whole discussion is that >> DB> multitasking means different things to different people. =A0 >> DB> If you are an old timer like me, there were major >> DB> screams 30 years ago when Unix claimed to be >> DB> multitasking. At that time most companies described >> DB> what we now consider as multitasking as >> DB> multiprocessing and that tasks were what we currently >> DB> call threads. =A0And 30 years ago, Unix did not support >> DB> threads. >> >> Yes (although with either definition the trolling here is false). >> More fundamentally, "task" means different things to different >> people. In some cases, a "task" is what in the Windows NT model would >> be a "thread". In other cases, a "task" would be a Windows NT >> "process". In yet other cases still, a "task" is what in the Solaris >> model would be a "light weight process". Teminology is and long has >> been, most definitely, an issue in this area. If one has clear >> definitions of "thread", "process", and "processor", then "multi- >> thread", "multi-process", and "multi-processor" have clear meanings. >> But since "task" is often ambiguous, "multi-tasking" is as well. >> Let's not get started on the difference between "concurrent" and >> "multi-programmed" or what "job" means ... (-: > > Well, that is all fine and dandy. > But... > What it boils down to, no matter what you call it, > is something that can have potentially deadly effect on operation > of your entire system, and I see it all day long every day with > windows. > > When I have at least 10 programs running and about 30 windows > opened and have done several hours worth of work and click back > on my firefox icon and do a refresh of about 10 pages in a rapid > succession, and my box simply freezes, what it means to me is > that I lost ALL sorts of work, and I lost ALL sorts of program > states, and it is going to take me half an hour to restore all > things, and that is if I am lucky. Some work might have been > lost for good. > > I could care less if firefox freezes by itself but I can continue > with everything else. I could simply hand kill that process from > the task manager need be as I have done gazillion of times. > > But when I loose nothing less then OS, than it is a TOTALLY > different issue, and the issue is OS, and not some screwed up > firefox or some driver it happens to be using. > > Simple as that. Does the monitor take a dump? /BAH
From: Anne & Lynn Wheeler on 31 Dec 2009 10:07
"Maxim S. Shatskih" <maxim(a)storagecraft.com.no.spam> writes: > Can you point at particular legacy of this Sequent's effort in Windows? > > For me, looks like that the "so-so more-or-less" SMP support was in > Windows from the beginning, and the major improvements in it were done > only in 2008/2008 R2 timeframe. re: http://www.garlic.com/~lynn/2009s.html#54 Problem with XP scheduler? http://www.garlic.com/~lynn/2009s.html#55 Problem with XP scheduler? Sequent had 16-way & 32-way ... and had done a lot of work on Unix to improve Unix scaleup in more than 2-way & 4-way operation (dynix). before we left ibm in early 90s, we had done some work with SCI ... for possibility for ha/cmp scaleup ... although at the time we were dealing with processor chip that didn't provide for any cache consistency .... and all scaleup had to be "cluster". old reference to jan92 meeting in ellison's conference room on ha/cmp cluster scaleup http://www.garlic.com/~lynn/95.html#13 other posts mentioning ha/cmp product http://www.garlic.com/~lynn/subtopic.html#hacmp later, two of the people (at the jan92 meeting) had left and show up at small client/server startup responsible for something called "commerce server" (the startup had also invented something called "SSL"). we were brought in because they wanted to payment transactions on their server .... and the result is now frequently called "electronic commerce". Part of the "electronic commerce" work was something called a "payment gateway" ... which acted as interface between webservers on the internet and financial networks for payment transactions (we periodically refer to it as original SOA). The initial "payment gateway" was an HA/CMP configuration with several other boxes around the edges providing various kinds of integrity, diagnostic, and security functions. misc. past posts mentioning "electronic commerce" and "payment gateway" work: http://www.garlic.com/~lynn/subtopic.html#gateway the small client/server startup had also a growing presence on the internet ... with servers deliverying their client & server products. They were using (unix) servers from companies in the silicon valley area .... that were quickly overloaded and required installation of more & more servers ... with their own unique host name ... and internet customers were asked to selective specify different URL host names when connecting (trying to spread out the internet load, this was before some of the front-end router work that was done at google). then they brought in a large sequent box and the problems went away (not just large sequent box, and dynix smp scaleup work ... but dynix also had some amount of tcp/ip protocol scale-up work) At the time NT had some SMP support ... but running on 8-way didn't show any increased thruput than running on 4-way (and little thruput improvement on 4-way compared to 2-way). Somehow sequent was involved to get NT running on their 32-way box ... and do a lot of scaleup work to show increasing thruput as configurations scaled passed 4-way (aka more than four processors) in SMP configuration. Later in the 90s, sequent was doing a 256-way smp SCI-based machine (NUMA-Q) and doing further work on their (dynix) unix to scaleup to 256-way (although I don't know of any work on NT for 256-way). In that timeframe, Steve Chen (from cray & chen supercomputers) was CTO at sequent ... and we did some consulting for him. Now while sequent did a lot of work on NT kernel to show increasing thruput as processors increased past two ... (at the time) NT thruput still didn't match Unix products (on the same hardware). at one point, there was a joint project that redmond had for putting a large web-based service on the internet and we were brought in to do some work on it. We showed that NT still didn't have thruput necessary to support the fully deployed operation ... and the redmond group decided that I would be the person to explain to their CEO why a UNIX platform would have to be used for the deployment. Before that actually took place, the executive running the group ... decided that (instead) the web service would have a staged roll-out ... the web population supported would never exceed the scaleup capability of NT (and therefor it would not be necessary to use a UNIX platform ... and I wouldn't have to explain to their CEO why NT wasn't being used). now there are two somewhat related issues (but not identical) .... whether NT showed increasing (SMP) thruput as the number of processors increased ... and NT thruput compared to other SMP implementations on the same hardware. sequent wiki page: http://en.wikipedia.org/wiki/Sequent_Computer_Systems 23May95 article: Sequent Unveils New High-End Systems for Windows NT; Based on Industry-Leading Platform Symmetry... http://www.allbusiness.com/technology/software-services-applications-computer/7126055-1.html from above: The WinServer systems add higher-end Windows NT-based performance to the existing integrated solutions Sequent provides to support customers' business requirements in decision support, online transaction processing and messaging. WinServer systems offer customers the benefits of proven Symmetry hardware, the industry's most mature and technologically advanced SMP platform, which has been installed with the UNIX operating system at thousands of Sequent customer sites around the world. .... snip ... 9Nov92 article: JUST WHEN SEQUENT THOUGHT IT WAS SAFE... http://www.businessweek.com/archives/1992/b329273.arc.htm from above: Microsoft Corp. has picked Sequent multiprocessing technology for Windows NT, the advanced operating system software it is readying for 1993. .... snip ... steve chen wiki page: http://en.wikipedia.org/wiki/Steve_Chen_%28computer_engineer%29 note the reference to SCI in the above ... is different than the SCI technology used by Convex, SGI, Sequent, DG & others for scalable shared memory multiprocessors ... wiki page: http://en.wikipedia.org/wiki/Scalable_Coherent_Interface above makes mention that the standardization effort eventually morphed into current InfiniBand ("which is similar to SCI in many ways") infiniband wiki: http://en.wikipedia.org/wiki/InfiniBand for some topic drift ... only marginally related posts (security, not SMP): http://www.garlic.com/~lynn/2009l.html#20 Cyber attackers empty business accounts in minutes other earlier posts mentioning the above: http://www.garlic.com/~lynn/2009.html#60 The 25 Most Dangerous Programming Errors http://www.garlic.com/~lynn/2009g.html#18 Top 10 Cybersecurity Threats for 2009, will they cause creation of highly-secure Corporate-wide Intranets? http://www.garlic.com/~lynn/2009h.html#28 Computer virus strikes US Marshals, FBI affected http://www.garlic.com/~lynn/2009i.html#22 My Vintage Dream PC -- 40+yrs virtualization experience (since Jan68), online at home since Mar1970 |