From allen.l.1314 at gmail.com Sun Aug 2 03:22:15 2009 From: allen.l.1314 at gmail.com (Luis Allen) Date: Sun, 2 Aug 2009 15:22:15 +0800 Subject: [Disksim-users] Can DiskSim use SSD as a cache for a disk? Message-ID: <9a17909c0908020022p711777fta7379498a96a055f@mail.gmail.com> Hi everyone, I'm using DiskSim for some SSD test, and the SSD model is implemented by Microsoft. Now I'm wandering if we can use an SSD as a cache for a disk in DiskSim. I couldn't findout in the mannual. So, please help me. Thank you very much. Best Regards, Luis From courageous at gmail.com Sun Aug 2 18:06:13 2009 From: courageous at gmail.com (Krithika) Date: Sun, 2 Aug 2009 18:06:13 -0400 Subject: [Disksim-users] Request arrival rate Message-ID: Hi, I want to run a number of traces of different characteristics, in a raid system I simulated in disksim. The problem is, when I use many traces which was collected from a system with more disks than my current simulated setup, the disksim is not able to handle the request rate of the trace. The simulation stops with a message "Stopping simulation because of saturation" which means the requests are fed in at a rate faster than the storage system can handle it. Is there a parameter in disksim that asks it to stop feeding in more requests when the current requests are not yet complete ? I guess we call it a open-loop simulation (??)... Basically my goal is to use the simulator to any traces irrespective of the request rate and the configuration of the system -- when I feed a more intense trace with say 1500 reqs/sec, and the storage system can handle only 500 reqs/sec, the simulator should still run correctly by controlling the rate of feeding in the trace requests to the storage layer.. Any help would be greatly appreciated.. Thanks a lot! Krithika. -------------- next part -------------- An HTML attachment was scrubbed... URL: From dmaung at earthlink.net Mon Aug 3 14:12:20 2009 From: dmaung at earthlink.net (David Maung) Date: Mon, 3 Aug 2009 11:12:20 -0700 Subject: [Disksim-users] Request arrival rate In-Reply-To: References: Message-ID: <002101ca1465$f8ae37e0$ea0aa7a0$@net> > The simulation stops with a message "Stopping simulation because of saturation" I ran into this error myself. It is a hard limit. In disksim_lororg.c, you can find the offending code as: if (currlogorg->stat.outstanding >= MAX_QUEUE_LENGTH) { fprintf(stderr,"Stopping simulation because of saturation: simtime %f, totalreqs %d\n", simtime, disksim->totalreqs); fflush(stderr); fprintf (outputfile, "Stopping simulation because of saturation: simtime %f, totalreqs %d\n", simtime, disksim->totalreqs); fprintf (outputfile, "last request: dev=%d, blk=%d, cnt=%d, %d (R==1)\n",curr->devno, curr->blkno, curr->bcount, (curr->flags & READ)); disksim_simstop(); } MAX_QUEUE_LENGTH is 10000 and is not based on any underlying cache configuration or anything. For example, I have configured a memory cache of 8192 blocks, but outstanding requests can still exceed 10000, perhaps due to multiple requests for the same blocks, etc. I would be inclined to comment out this code, if, as in my case, you actually find it to be invalid for your given configuration. David Maung From dmaung at earthlink.net Mon Aug 3 14:47:37 2009 From: dmaung at earthlink.net (David Maung) Date: Mon, 3 Aug 2009 11:47:37 -0700 Subject: [Disksim-users] Can DiskSim use SSD as a cache for a disk? In-Reply-To: References: Message-ID: <002501ca146a$e3404790$a9c0d6b0$@net> I think it would work. Section 3.4.10 of disksim4-manual.pdf describes using one device as a cache for another. A sample file to start with is disksim-4.0/valid/synthcachedev.parv. I would: 1) modify that file to add an SSD device (perhaps copying the "ssdmodel_ssd SSD {...}" section from a .parv file in the ssdmodel/valid subdirectory), 2) add it to the topology under BUS1, and 3) change the "cache device" line to refer to the SSD device, and give it a test. >I'm using DiskSim for some SSD test, and the SSD model is implemented >by Microsoft. Now I'm wandering if we can use an SSD as a cache for a >disk in DiskSim. I couldn't findout in the mannual. So, please help >me. Thank you very much. > >Best Regards, > >Luis From n080di at yahoo.com Tue Aug 4 09:47:17 2009 From: n080di at yahoo.com (Nobody Nobody Nobody) Date: Tue, 4 Aug 2009 06:47:17 -0700 (PDT) Subject: [Disksim-users] Request arrival rate In-Reply-To: <002101ca1465$f8ae37e0$ea0aa7a0$@net> Message-ID: <786894.89612.qm@web55104.mail.re4.yahoo.com> Not sure if commenting out is a good idea. I tried to increase the MAX_QUEUE_LENGTH value. But it appears that disksim either hangs or takes a really really long time to finish if I do this. You can see the offending while loop by attaching a debugger to it. Basically, the loop seems to be making progress, but it's hard to say for sure since it seems to run forever. --- On Mon, 8/3/09, David Maung wrote: From: David Maung Subject: Re: [Disksim-users] Request arrival rate To: disksim-users at ece.cmu.edu Date: Monday, August 3, 2009, 2:12 PM > The simulation stops with a message "Stopping simulation because of saturation" I ran into this error myself.? It is a hard limit.? In disksim_lororg.c, you can find the offending code as: ???if (currlogorg->stat.outstanding >= MAX_QUEUE_LENGTH) { ? ? ? fprintf(stderr,"Stopping simulation because of saturation: simtime %f, totalreqs %d\n", simtime, disksim->totalreqs); ? ? ? fflush(stderr); ? ? ? fprintf (outputfile, "Stopping simulation because of saturation: simtime %f, totalreqs %d\n", simtime, disksim->totalreqs); ? ? ? fprintf (outputfile, "last request:? dev=%d, blk=%d, cnt=%d, %d (R==1)\n",curr->devno, curr->blkno, curr->bcount, (curr->flags & READ)); ? ? ? disksim_simstop(); ???} MAX_QUEUE_LENGTH is 10000 and is not based on any underlying cache configuration or anything.? For example, I have configured a memory cache of 8192 blocks, but outstanding requests can still exceed 10000, perhaps due to multiple requests for the same blocks, etc.? I would be inclined to comment out this code, if, as in my case, you actually find it to be invalid for your given configuration. David Maung _______________________________________________ Disksim-users mailing list Disksim-users at ece.cmu.edu https://sos.ece.cmu.edu/mailman/listinfo/disksim-users -------------- next part -------------- An HTML attachment was scrubbed... URL: From n080di at yahoo.com Tue Aug 4 13:11:43 2009 From: n080di at yahoo.com (Nobody Nobody Nobody) Date: Tue, 4 Aug 2009 10:11:43 -0700 (PDT) Subject: [Disksim-users] 64 bit, patch, git Message-ID: <989799.2822.qm@web55105.mail.re4.yahoo.com> Was this done? Is the patched version available somewhere? Thanks. On Fri, 2007-07-27 at 18:46 -0600, Yipkei Kwok wrote: > It might be the problem with the compiler version. Which may be fixed by my patch: https://sos.ece.cmu.edu/pipermail/disksim-users/2007-July/000168.html As yet un-replied to AFAICS ;-) OTOH - though it now compiles cleanly, and all the fixes were obvious in one way or another; I've as yet to hook DiskSim up and test it live; so YMMV. If I get no response wrt. the patch inclusion process, I guess I'll just check it into a git/ repository somewhere and post a link to that. HTH, Michael. -- michael.meeks at novell.com <><, Pseudo Engineer, itinerant idiot -------------- next part -------------- An HTML attachment was scrubbed... URL: From michael.meeks at novell.com Wed Aug 5 07:12:13 2009 From: michael.meeks at novell.com (Michael Meeks) Date: Wed, 05 Aug 2009 12:12:13 +0100 Subject: [Disksim-users] 64 bit, patch, git In-Reply-To: <989799.2822.qm@web55105.mail.re4.yahoo.com> References: <989799.2822.qm@web55105.mail.re4.yahoo.com> Message-ID: <1249470733.17740.115.camel@linux-h3ht.site> On Tue, 2009-08-04 at 10:11 -0700, Nobody Nobody Nobody wrote > Was this done? Is the patched version available somewhere? git clone git://gitorious.org/iogrind/mainline.git That includes a patched version of disksim with several fixes and a simpler API that we use in iogrind. Of course, really - we need a separate github repo setting up for that beast, and it needs a -lot- of cleanup: the riddling with global variables, and the general state of the code is quite extraordinary ;-) HTH, Michael. -- michael.meeks at novell.com <><, Pseudo Engineer, itinerant idiot From yyzhang at cs.wisc.edu Mon Aug 24 13:57:41 2009 From: yyzhang at cs.wisc.edu (Yiying Zhang) Date: Mon, 24 Aug 2009 12:57:41 -0500 Subject: [Disksim-users] compiling error on 64bit machine Message-ID: <0fcdb008b39e2b0e5726c45a2d425c1e.squirrel@webmail.cs.wisc.edu> Hi, When I try to compile disksim on a 64bit machine running 2.6.9-67.0.4.ELsmp, it gives me the following error: make[1]: *** No rule to make target `/usr/include/gnu/stubs-32.h', needed by `mech_g1_seektime.o'. Stop. It compiles fine on a 32bit machine. Any help will be appreciated. Thanks From markus at bluegap.ch Wed Aug 26 03:47:35 2009 From: markus at bluegap.ch (Markus Wanner) Date: Wed, 26 Aug 2009 09:47:35 +0200 Subject: [Disksim-users] compiling error on 64bit machine In-Reply-To: <0fcdb008b39e2b0e5726c45a2d425c1e.squirrel@webmail.cs.wisc.edu> References: <0fcdb008b39e2b0e5726c45a2d425c1e.squirrel@webmail.cs.wisc.edu> Message-ID: <20090826094735.35196s2eysi2difr@mail.bluegap.ch> Hi, Quoting "Yiying Zhang" : > When I try to compile disksim on a 64bit machine Yes, disksim currently doesn't compile on 64bit machines. I did some corrections and get it to compile, but I didn't publish any patch up until now. Thank you for reminding me, I'll try to get that together. Regards Markus Wanner