From maobo1983 at 163.com Sun Sep 2 05:01:20 2007 From: maobo1983 at 163.com (hust_mb) Date: Sun, 2 Sep 2007 17:01:20 +0800 Subject: [Disksim-users] disk cache management References: <20070831160010.5CC6C8A3B@sos.ece.cmu.edu> Message-ID: <200709021701152968537@163.com> Hi,all In my experiment I found a very strange thing: No write hit in on-board disk cache! I use the validate trace as following: W Hit 5568004 8 1464.00000 53368.00000 ---request 1-- W Hit 5568012 8 1323.00000 54353.00000 ---request 2-- W Hit 1505868 4 1098.00000 32082.00000 ---request 3-- W Hit 1505868 4 879.00000 39461.00000 ---request 4-- # ./disksim cheetah9LP.parv file_1 validate trace 0 But no hit found in the file_1. Theoretically there should be a write combine with request 1 and 2. Also a write hit with 4 after 3. But indeed there is no. I tracked the disk_buffer_select_write_segment and disk_release_hda function. There is no conserve of write although I use the fast write mode. A new request own a segment for write and release it after completion. And I increased the segment counts in the cheetah9LP.diskspecs. There is litter increase(less than 1%) about the cache hit ratio! (with command ./disksim cheetah9LP.parv file_1 validate cheetah9LP.trace 0) Does this really the cache management for modern disk? I think the fast write mode enabled can conserve more small write and then write them optimizely (sometime sequentially write or with freeblock scheduling policy). Thank you for your discussion. ------------------ mb 2007-09-02 -------------- next part -------------- An HTML attachment was scrubbed... URL: From diwaker.lists at gmail.com Fri Sep 7 02:43:22 2007 From: diwaker.lists at gmail.com (Diwaker Gupta) Date: Thu, 6 Sep 2007 23:43:22 -0700 Subject: [Disksim-users] Interfacing disksim synchronously Message-ID: <891be9410709062343w7ee98d2dhd037e33f6ef9298e@mail.gmail.com> Hi *, I'm looking to integrate disksim in a full system simulator. I've read through disksim_interface*, the manual etc and everywhere one of the requirements listed is that the simulator support asynchronous operation -- that is, it should be able to proceed without blocking for the disk request to finish, and at some later time disksim would deliver the completion notification. Due to some constraints, I need to interface disksim in a synchronous fashion. That is, I'd like to pass in the request to disksim, and I'd like disksim to immediately return to me the time it _would_ take to finish the request. I can then inject the appropriately delay in the external simulator myself. Can anyone suggest what would be a good way to go about this? Is this even doable? Thanks, Diwaker -- http://floatingsun.net/ From michael.meeks at novell.com Fri Sep 7 09:38:20 2007 From: michael.meeks at novell.com (Michael Meeks) Date: Fri, 07 Sep 2007 14:38:20 +0100 Subject: [Disksim-users] run on 64bit systems ... Message-ID: <1189172300.21053.226.camel@localhost> So, When we apply this, disksim starts to run on 64bit systems. Is there -really- no-one collecting patches / build-fixes for disksim, wrt. rolling a new release ? I would have thought the combination of "build with recent gcc's" and "run on 64bit systems" would be worth others having ? :-) Fix from Radek Doulik FWIW. Regards, Michael. --- a/disksim-3.0/libparam/util.c +++ b/disksim-3.0/libparam/util.c @@ -805,7 +805,7 @@ int lp_add_param(struct lp_param ***b, i /* didn't find a free slot -- double the array */ int newlen = 2 * (*plen) + 1; (*b) = realloc((*b), newlen * sizeof(int *)); - bzero((int *)(*b) + *plen, ((*plen) + 1) * sizeof(int*)); + bzero((int **)(*b) + *plen, ((*plen) + 1) * sizeof(int*)); (*b)[(*plen)] = p; *plen = newlen; } -- michael.meeks at novell.com <><, Pseudo Engineer, itinerant idiot From swschlosser at gmail.com Fri Sep 7 15:38:59 2007 From: swschlosser at gmail.com (Steve Schlosser) Date: Fri, 7 Sep 2007 15:38:59 -0400 Subject: [Disksim-users] Interfacing disksim synchronously In-Reply-To: <891be9410709062343w7ee98d2dhd037e33f6ef9298e@mail.gmail.com> References: <891be9410709062343w7ee98d2dhd037e33f6ef9298e@mail.gmail.com> Message-ID: <4d362c350709071238j79e4b193pd793056b411d5809@mail.gmail.com> You should be able to use the disksim interface as-is, although it will preclude you from doing a few things. Basically your approach should be to issue a request using disksim_interface_request_arrive() and then call disksim_interface_internal_event() repeatedly until disksim calls your completion function. The time given to the completion function should be the time that the request finished. One limitation is that you can only ever have one request "outstanding" to disksim at a time, meaning that you won't see any effect from scheduling, overlapping of seek and transfer, etc. This may not matter to you. Also, your simulator's notion of time will be different from disksim's view of time, since your repeated calls to disksim_interface_internal_event() will cause disksim's internal time to advance into the future, relative to your simulator's time. Good luck. -steve On 9/7/07, Diwaker Gupta wrote: > Hi *, > > I'm looking to integrate disksim in a full system simulator. I've read > through disksim_interface*, the manual etc and everywhere one of the > requirements listed is that the simulator support asynchronous > operation -- that is, it should be able to proceed without blocking > for the disk request to finish, and at some later time disksim would > deliver the completion notification. > > Due to some constraints, I need to interface disksim in a synchronous > fashion. That is, I'd like to pass in the request to disksim, and I'd > like disksim to immediately return to me the time it _would_ take to > finish the request. I can then inject the appropriately delay in the > external simulator myself. > > Can anyone suggest what would be a good way to go about this? Is this > even doable? > > Thanks, > Diwaker > -- > http://floatingsun.net/ > _______________________________________________ > Disksim-users mailing list > Disksim-users at ece.cmu.edu > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users > From ali_teke84 at yahoo.com Sun Sep 16 10:33:05 2007 From: ali_teke84 at yahoo.com (Ali Teke) Date: Sun, 16 Sep 2007 07:33:05 -0700 (PDT) Subject: [Disksim-users] Fw: Welcome to the "Disksim-users" mailing list Message-ID: <192455.58076.qm@web53304.mail.re2.yahoo.com> ----- Forwarded Message ---- From: "disksim-users-request at ece.cmu.edu" To: ali_teke84 at yahoo.com Sent: Monday, September 10, 2007 2:39:03 PM Subject: Welcome to the "Disksim-users" mailing list Welcome to the Disksim-users at ece.cmu.edu mailing list! To post to this list, send your email to: disksim-users at ece.cmu.edu General information about the mailing list is at: https://sos.ece.cmu.edu/mailman/listinfo/disksim-users If you ever want to unsubscribe or change your options (eg, switch to or from digest mode, change your password, etc.), visit your subscription page at: https://sos.ece.cmu.edu/mailman/options/disksim-users/ali_teke84%40yahoo.com You can also make such adjustments via email by sending a message to: Disksim-users-request at ece.cmu.edu with the word `help' in the subject or body (don't include the quotes), and you will get back a message with instructions. You must know your password to change your options (including changing the password, itself) or to unsubscribe. It is: ... ____________________________________________________________________________________ Looking for a deal? Find great prices on flights and hotels with Yahoo! FareChase. http://farechase.yahoo.com/ -------------- next part -------------- An HTML attachment was scrubbed... URL: From maobo1983 at 163.com Sat Sep 22 09:17:59 2007 From: maobo1983 at 163.com (hust_mb) Date: Sat, 22 Sep 2007 21:17:59 +0800 Subject: [Disksim-users] About store data out of disksim References: <20070907160010.970C08A65@sos.ece.cmu.edu> Message-ID: <200709222117567969770@163.com> Dear Professor Ganger, I am a Ph.d student in China. Now I encountered a problem in my research job that I want to keep the data for the request in the disksim. And reuse them later. But disksim can't do this. As one paper in FAST'02 title "Timing-Accurate Storage Emulation" argued to do such a job. Is it possible for me to get the source code used in the paper? Thank you very much! ------------------ Maobo CS, HUST. China From ganger at ece.cmu.edu Sat Sep 22 14:51:20 2007 From: ganger at ece.cmu.edu (Greg Ganger) Date: Sat, 22 Sep 2007 14:51:20 -0400 (EDT) Subject: [Disksim-users] Re: About store data out of disksim In-Reply-To: <200709222117567969770@163.com> References: <20070907160010.970C08A65@sos.ece.cmu.edu> <200709222117567969770@163.com> Message-ID: You will be much better off building a simple "store actual data in a big file" component of your own. Good luck, Greg On Sat, 22 Sep 2007, hust_mb wrote: > Dear Professor Ganger, > > I am a Ph.d student in China. > Now I encountered a problem in my research job that I want to keep the data for the request in the disksim. And reuse them later. But disksim can't do this. As one paper in FAST'02 title "Timing-Accurate Storage Emulation" argued to do such a job. Is it possible for me to get the source code used in the paper? > > Thank you very much! > ------------------ > Maobo > CS, HUST. China > > > > >