From myidpt at gmail.com Mon Jul 5 15:34:28 2010 From: myidpt at gmail.com (Yonggang Liu) Date: Mon, 5 Jul 2010 15:34:28 -0400 Subject: [Disksim-users] Has anyone used DIXtrac? Message-ID: Hello all, I'm pretty new to Disksim. Recently I met some problems when I was running DIXtrac on CentOS 5.1 to model the 8GB SCSI disk on VMWare. I followed what DIXtrac says to extract the local disk model. The first problem is when doing the disk layout mapping, it shows "Cannot handle defect list format (0)". So I looked into the code and commented out the "read_defect_list" function in that hope that it still works correctly. But after then I met the second problem. When I'm running pass0.py, the program kept running all the time (it didn't finish after several days). I think it may because the python program has used up too big memory, since it stores the entire content of "layout.mappings" to the object "lout". But the layout.mappings file is usually in the scale of gegabyte, which the memory can not deal with. To verify my guessing, I tried the small layout.mappings files, and this time pass0.py works well. These problems have confused me for a couple of weeks. If anyone have used DIXtrac, could you please help me? I will really appreciate it. Thank you very much! -- Yonggang Liu Advanced Computing and Information Systems Laboratory University of Florida -------------- next part -------------- An HTML attachment was scrubbed... URL: From ganger at ece.cmu.edu Mon Jul 5 16:12:46 2010 From: ganger at ece.cmu.edu (Greg Ganger) Date: Mon, 5 Jul 2010 16:12:46 -0400 (EDT) Subject: [Disksim-users] Has anyone used DIXtrac? In-Reply-To: References: Message-ID: Perhaps I am misunderstanding, but you should not expect DIXtrac to successfully characterize a virtual disk. Greg On Mon, 5 Jul 2010, Yonggang Liu wrote: > Hello all, > > I'm pretty new to Disksim. Recently I met some problems when I was running > DIXtrac on CentOS 5.1 to model the 8GB SCSI disk on VMWare. > > I followed what DIXtrac says to extract the local disk model. The first > problem is when doing the disk layout mapping, it shows "Cannot handle > defect list format (0)". So I looked into the code and commented out the > "read_defect_list" function in that hope that it still works correctly. But > after then I met the second problem. When I'm running pass0.py, the program > kept running all the time (it didn't finish after several days). I think it > may because the python program has used up too big memory, since it stores > the entire content of "layout.mappings" to the object "lout". But the > layout.mappings file is usually in the scale of gegabyte, which the memory > can not deal with. To verify my guessing, I tried the small layout.mappings > files, and this time pass0.py works well. > > These problems have confused me for a couple of weeks. If anyone have used > DIXtrac, could you please help me? I will really appreciate it. Thank you > very much! > > -- > Yonggang Liu > Advanced Computing and Information Systems Laboratory > University of Florida > From myidpt at gmail.com Mon Jul 5 19:12:33 2010 From: myidpt at gmail.com (Yonggang Liu) Date: Mon, 5 Jul 2010 19:12:33 -0400 Subject: [Disksim-users] Has anyone used DIXtrac? In-Reply-To: References: Message-ID: Thank you so much for your reply. I will try it on real SCSI disks. On Mon, Jul 5, 2010 at 4:12 PM, Greg Ganger wrote: > > Perhaps I am misunderstanding, but you should not expect DIXtrac to > successfully characterize a virtual disk. > > Greg > > > > On Mon, 5 Jul 2010, Yonggang Liu wrote: > > Hello all, >> >> I'm pretty new to Disksim. Recently I met some problems when I was running >> DIXtrac on CentOS 5.1 to model the 8GB SCSI disk on VMWare. >> >> I followed what DIXtrac says to extract the local disk model. The first >> problem is when doing the disk layout mapping, it shows "Cannot handle >> defect list format (0)". So I looked into the code and commented out the >> "read_defect_list" function in that hope that it still works correctly. >> But >> after then I met the second problem. When I'm running pass0.py, the >> program >> kept running all the time (it didn't finish after several days). I think >> it >> may because the python program has used up too big memory, since it stores >> the entire content of "layout.mappings" to the object "lout". But the >> layout.mappings file is usually in the scale of gegabyte, which the memory >> can not deal with. To verify my guessing, I tried the small >> layout.mappings >> files, and this time pass0.py works well. >> >> These problems have confused me for a couple of weeks. If anyone have used >> DIXtrac, could you please help me? I will really appreciate it. Thank you >> very much! >> >> -- >> Yonggang Liu >> Advanced Computing and Information Systems Laboratory >> University of Florida >> >> -- Yonggang Liu Advanced Computing and Information Systems Laboratory University of Florida -------------- next part -------------- An HTML attachment was scrubbed... URL: From yyangpan at gmail.com Sat Jul 10 16:06:09 2010 From: yyangpan at gmail.com (Yangyang Pan) Date: Sat, 10 Jul 2010 16:06:09 -0400 Subject: [Disksim-users] How to run one trace file repeatedly? Message-ID: Hi all, I want to run the simulation for a long time, however I trace file is not large enough and I do not want to use the synthetic trace. Does anyone know how to run a trace file repeatly? Also, if you have any better trace sources for SSD, would you please send me the link? I will be very appreciated for your help. Thank you very much. Best, Yangyang -- All the Best ----------------------------------------------------------- Yangyang Electrical and Computer&System Engineering Department Rensselaer Polytechnic Institute, 12180,Troy.NY -------------- next part -------------- An HTML attachment was scrubbed... URL: From shil0704 at mail.ustc.edu.cn Tue Jul 13 05:19:48 2010 From: shil0704 at mail.ustc.edu.cn (Shi Liang (Charles)) Date: Tue, 13 Jul 2010 17:19:48 +0800 (CST) Subject: [Disksim-users] Stopping simulation because of saturation, what is the ture reason????? In-Reply-To: References: Message-ID: <31753722.5911279012788808.JavaMail.coremail@mailweb> Hi, all I use disksim to run a intensive trace. There is a very strange problem that I come to: As we know, the controller type 3 can have cache there. I fI use type 3 controller with a cache to run the trace, there is a sudden stop of simulation with following message: "Stopping simulation because of saturation: simtime 143242.140000, totalreqs 103433" I know this message comming from file disksim_logorg.c with the outstanding value bigger than 10000. But when I use the type 1 controller, which is no cache, but directly send the request down or up. It can successfully run without any problems. In my view, with a cache, the throughput shoud be bigger, and it should not saturation compared with no cache. This is strange. If any one comes to this question, pls share with me, thank you very much. SHI Liang USTC of China and CityU of HK (Someone has shared that this is because throughput, but with no cache, it is OK. I think this is not the ture reason.) From shil0704 at mail.ustc.edu.cn Fri Jul 16 23:35:27 2010 From: shil0704 at mail.ustc.edu.cn (Shi Liang (Charles)) Date: Sat, 17 Jul 2010 11:35:27 +0800 (CST) Subject: [Disksim-users] Disksim-users Digest, Vol 58, Issue 3 In-Reply-To: References: Message-ID: <11199196.278301279337727180.JavaMail.coremail@mailweb> Hi, all, The problem has been solved. The reason for the time saturation is that request size in the intensive trace is larger than that of cache can solve. Two appoarches can be applied to solve the problem: 1 change the size of max request size 2 change the large size request into small sequential ones. I think these solutions can be good for those who think that throughput of disksim is not OK for high throughput applications. SHI Liang > -----Original email----- > From: disksim-users-request at ece.cmu.edu > Sent Time: Jul 14, 2010 12:00:04 AM > To: disksim-users at ece.cmu.edu > Cc: > Subject: Disksim-users Digest, Vol 58, Issue 3 > > Send Disksim-users mailing list submissions to > disksim-users at ece.cmu.edu > > To subscribe or unsubscribe via the World Wide Web, visit > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users > or, via email, send a message with subject or body 'help' to > disksim-users-request at ece.cmu.edu > > You can reach the person managing the list at > disksim-users-owner at ece.cmu.edu > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Disksim-users digest..." > > > Today's Topics: > > 1. Stopping simulation because of saturation, what is the ture > reason????? (Shi Liang (Charles)) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 13 Jul 2010 17:19:48 +0800 (CST) > From: "Shi Liang (Charles)" > Subject: [Disksim-users] Stopping simulation because of saturation, > what is the ture reason????? > To: disksim-users at ece.cmu.edu > Message-ID: <31753722.5911279012788808.JavaMail.coremail at mailweb> > Content-Type: text/plain; charset=us-ascii > > Hi, all > > I use disksim to run a intensive trace. There is a very strange problem that I come to: > As we know, the controller type 3 can have cache there. > I fI use type 3 controller with a cache to run the trace, there is a sudden stop of simulation with following message: > "Stopping simulation because of saturation: simtime 143242.140000, totalreqs 103433" > I know this message comming from file disksim_logorg.c with the outstanding value bigger than 10000. > > But when I use the type 1 controller, which is no cache, but directly send the request down or up. It can successfully run without any problems. > > In my view, with a cache, the throughput shoud be bigger, and it should not saturation compared with no cache. > > This is strange. If any one comes to this question, pls share with me, thank you very much. > > SHI Liang > USTC of China and CityU of HK > > > (Someone has shared that this is because throughput, but with no cache, it is OK. I think this is not the ture reason.) > > > ------------------------------ > > _______________________________________________ > Disksim-users mailing list > Disksim-users at ece.cmu.edu > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users > > > End of Disksim-users Digest, Vol 58, Issue 3 > ******************************************** From shil0704 at mail.ustc.edu.cn Fri Jul 23 02:50:41 2010 From: shil0704 at mail.ustc.edu.cn (Shi Liang (Charles)) Date: Fri, 23 Jul 2010 14:50:41 +0800 (CST) Subject: [Disksim-users] Idle process has been running for more than 50 seconds In-Reply-To: References: Message-ID: <30429694.613921279867841707.JavaMail.coremail@mailweb> Hello, everyone, I come to a new problem. I wander if anyone comes to it before. When I use disksim with its synthetical trace input parameter, I come to the result that: Idle process has been running for more than 50 seconds and then disksim stop. This sentence is in the disksim_pfsim.c. An interesting problem that I found is when I use the cache in the smart controller with setting the bit granularity with 4, this result would sure to happen. Hope anyone can help me. Thank you. SHI Liang (Charles)