From liufeng086 at 126.com Mon Jun 11 04:01:36 2012 From: liufeng086 at 126.com (wo) Date: Mon, 11 Jun 2012 16:01:36 +0800 (CST) Subject: [Disksim-users] Disksim-users Digest, Vol 79, Issue 2 In-Reply-To: References: Message-ID: <559ebacb.11572.137da8ff3c8.Coremail.liufeng086@126.com> Dear users, Have you ever used the intel trace generator named open storage toolkit to generate synthetic trace??Do you know how to generate the synthetic trace with it??Where to find the generated synthetic trace?? Thank you for your answer! At 2012-05-16 00:00:04,disksim-users-request at ece.cmu.edu wrote: >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. Re: How do I use trace file:WebSearch3.spc (Mingbang) > > >---------------------------------------------------------------------- > >Message: 1 >Date: Tue, 15 May 2012 10:58:53 +0800 >From: Mingbang >Subject: Re: [Disksim-users] How do I use trace file:WebSearch3.spc >To: disksim-users at ece.cmu.edu >Message-ID: > >Content-Type: text/plain; charset="gb2312" > >hello? > > of course you can no use the raw trace, as you said the format is >totally different. the format should be : > 21698.968000 0 835 4 0 > 21843.149000 0 59031 1 1 > 21875.917000 0 59031 1 0 > the first column is arrival time ,second is device no?3rd is lba? 4th is >block size , 5th is operation flag(0 is write ,1 is read ) > for WebSearch3.spc's trace: > the first column is device no , because flashsim can only support >single device , you should make all the device to be 0; 2nd is lba; 3rd is >byte number whick should be divided by 512 to be a block size; 4th >is operation flag; 5th is arrival time, because the arrival time interval >is too short , so it will appear the saturation wrong? and you >should amplify the arrival time interval? > if you want to use WebSearch3.spc's trace, you should change the >format. And I think awk is good tool for this . >Hope this can help you. > >by mingbang24 >-- >Wang Mingbang >Department of Electronic and Information Engineering >Beihang University >Phone Number: +86-15910556701 >Beijing 100191, P.R. China >-------------- next part -------------- >An HTML attachment was scrubbed... >URL: > >------------------------------ > >_______________________________________________ >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 79, Issue 2 >******************************************** -------------- next part -------------- An HTML attachment was scrubbed... URL: From liangxiongxu at gmail.com Fri Jun 15 11:59:42 2012 From: liangxiongxu at gmail.com (leon hsu) Date: Fri, 15 Jun 2012 08:59:42 -0700 Subject: [Disksim-users] Why only few trace(244) were handled when I tried to run 10000 traces. Message-ID: Dear All, I use the disksim && ssdsim for two months. I try to use them at the same time. But there is something weired. I use 10'000 traces or synthetic trace generator to test the system. Only a few traces at the beginning of trace file, more or less 200, were handled when the simulator stopped. I don't know why. Does anybody have similar experience? Please let me know. So appreciate. I attach the relevant files so that some helper could take a look. Also, here is the link to my project: https://www.dropbox.com/sh/nhavksi2jx5zfdg/Y8OFYsO4ux I write down all the problems on the "readme at first.txt". It is too long, so I put them into one file according to your suggestion. The folder HybridDiskSim is the project I am using. And my host operating system is Ubuntu 12.04 LTS. link to dropbox: https://www.dropbox.com/sh/nhavksi2jx5zfdg/Y8OFYsO4ux (1) When I use the synthetic trace for SSD, the number of maximum traces handled is 244. I don't know why. the command I used like: ./src/disksim example_ssd.parv example_ssd.outv ascii 0 1 I record the synthetic trace into the file synthetic_ssd.trace. The last item in this trace seems weird(Both SSD and HDD). It should be 0 or 1, but the value is 400000003 0.000000 1 31853640 8 3 0.132504 1 31853648 8 40000003 0.265008 1 31853656 8 40000003 0.397512 1 31853664 8 40000003 0.530016 1 31853672 8 40000003 0.662520 1 31853680 8 40000003 ........................... also, the ssd statistics is weird. Only 43 requests handled. Furthermore, all of 43 requests are read. No one write. SSD STATISTICS --------------------- ssd Total Requests handled: 43 ssd Requests per second: 0.850811 ssd Completely idle time: 531.769600 0.010522 ssd Response time average: 0.133301 ssd Response time std.dev.: 0.001549 ssd Response time maximum: 0.140120 ssd Response time distribution < 5 < 10 < 20 < 40 < 60 < 90 <120 <150 <200 200+ 43 0 0 0 0 0 0 0 0 0 ssd Overlaps combined: 0 0.000000 ssd Read overlaps combined: 0 0.000000 0.000000 ssd Critical Reads: 43 0.977273 ssd Critical Read Response time average: 0.133301 ssd Critical Read Response time std.dev.: 0.001549 ssd Critical Read Response time maximum: 0.140120 ssd Critical Read Response time distribution Leon <732.789.6683> Email: liangxiongxu at gmail.com -------------- next part -------------- An HTML attachment was scrubbed... URL: From ricardoreyyy at gmail.com Fri Jun 15 18:05:14 2012 From: ricardoreyyy at gmail.com (Ricardo Rey) Date: Fri, 15 Jun 2012 17:05:14 -0500 Subject: [Disksim-users] Dixtrac and topology questions Message-ID: Hi! I have a couple of questions: 1) How do I generate a ".diskspecs" file for some disk type? Do I have to use DIXtrac? Or is there some other way? And, about this... does anyone know if there is some repository of *.diskspecs somewhere? 2) I would like to test some high-availability topologies with redundancy; something like this: topology disksim_iodriver ctlr0 [ disksim_bus bus1 [ disksim_disk disk0 [] disksim_disk disk1 [] ] disksim_bus bus2 [ disksim_disk disk0 [] disksim_disk disk1 [] ] ] Note that disk0 and disk0 are connected to both buses. Is it possible to build such an architecture? What I want to do is test some worst-case-scenarios, with some components failing, and how would the system respond to that. That is why I need this redundant topology. Thanks in advance - Rick -------------- next part -------------- An HTML attachment was scrubbed... URL: From dcjy at mail.ustc.edu.cn Sat Jun 16 10:09:42 2012 From: dcjy at mail.ustc.edu.cn (=?GBK?B?tPqzyb+h?=) Date: Sat, 16 Jun 2012 22:09:42 +0800 (CST) Subject: [Disksim-users] write is faster than read ? Message-ID: <22593278.3550871339855782460.JavaMail.coremail@mailweb> Hi, I found An interesting phenomenon when I used the disksim to simulate a atlas 10K disk. Assume that there is a fragment trace contains two requests. The first one's opcode is "read" and the second is "Write", then I found that the second request's response time is very samall. In other words, when the disk opration type changed form "R" to "W", the write opration can be completed more quickly, why? Could someone explain this phenomenon or have some reference material to recommend? Thank you very much. From zh-lian at 163.com Mon Jun 18 01:53:25 2012 From: zh-lian at 163.com (=?GBK?B?wO7V17ui?=) Date: Mon, 18 Jun 2012 13:53:25 +0800 (CST) Subject: [Disksim-users] Dixtrac and topology questions In-Reply-To: References: Message-ID: <7e0315aa.f6f3.137fe271dec.Coremail.zh-lian@163.com> For question 1,I don't think DIXtrac can help you .I have use it work on a disk about 300G,but it dosen't work.DIXtrac will work on small disk. At 2012-06-16 06:05:14,"Ricardo Rey" wrote: Hi! I have a couple of questions: 1) How do I generate a ".diskspecs" file for some disk type? Do I have to use DIXtrac? Or is there some other way? And, about this... does anyone know if there is some repository of *.diskspecs somewhere? 2) I would like to test some high-availability topologies with redundancy; something like this: topology disksim_iodriver ctlr0 [ disksim_bus bus1 [ disksim_disk disk0 [] disksim_disk disk1 [] ] disksim_bus bus2 [ disksim_disk disk0 [] disksim_disk disk1 [] ] ] Note that disk0 and disk0 are connected to both buses. Is it possible to build such an architecture? What I want to do is test some worst-case-scenarios, with some components failing, and how would the system respond to that. That is why I need this redundant topology. Thanks in advance - Rick -------------- next part -------------- An HTML attachment was scrubbed... URL: From vahldiek at mpi-sws.org Mon Jun 18 05:23:46 2012 From: vahldiek at mpi-sws.org (Anjo Vahldiek) Date: Mon, 18 Jun 2012 11:23:46 +0200 Subject: [Disksim-users] Dixtrac and topology questions In-Reply-To: <7e0315aa.f6f3.137fe271dec.Coremail.zh-lian@163.com> References: <7e0315aa.f6f3.137fe271dec.Coremail.zh-lian@163.com> Message-ID: <4FDEF3A2.1090505@mpi-sws.org> This is not correct. At least in my experience Dixtrac can work for disks with a capacity of 300GB. Most important are two things: 1) Disk has to be a SCSI disk 2) OS has to be 32-bit I finished extracting the model of 2 Maxtor disk (146GB & 300GB) including comparison graphs. You can find them here: http://ovahldy.blogspot.de/2011/03/recent-diskmodels-for-disksim.html For question 2: In my opinion what you suggest requires extensive programming. At least I haven't seen code which would deal with broken buses... For simplicity you may want to look at how to integrate disksim with a system simulator (in src/syssim). It may be easier to write your own outside special handling for broken busses. Anjo On Mon 18 Jun 2012 07:53:25 AM CEST, ??? wrote: > For question 1,I don't think DIXtrac can help you .I have use it work > on a disk about 300G,but it dosen't work.DIXtrac will work on small disk. > > At 2012-06-16 06:05:14,"Ricardo Rey" wrote: > > Hi! > > I have a couple of questions: > > 1) How do I generate a ".diskspecs" file for some disk type? Do I > have to use DIXtrac? Or is there some other way? And, about > this... does anyone know if there is some repository of > *.diskspecs somewhere? > > 2) I would like to test some high-availability topologies with > redundancy; something like this: > > topology disksim_iodriver ctlr0 [ > disksim_bus bus1 [ > disksim_disk disk0 [] > disksim_disk disk1 [] > ] > disksim_bus bus2 [ > disksim_disk disk0 [] > disksim_disk disk1 [] > ] > ] > > Note that disk0 and disk0 are connected to both buses. Is it > possible to build such an architecture? > > What I want to do is test some worst-case-scenarios, with some > components failing, and how would the system respond to that. That > is why I need this redundant topology. > > Thanks in advance > > - Rick > > > > > > _______________________________________________ > Disksim-users mailing list > Disksim-users at ece.cmu.edu > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users