Number of Tasks in Distributed Computing
in [Matlab]
|
From: Amir Homayoun on 11 Jan 2010 06:18 Dear Titus Thanks. Well. My MATLAB is 7.4.0.287 (R2007a) and my MATLAB Distributed Computing Engine and Distributed Computing Toolbox are both Version 3.1 and it seems that I don't have PCT installed on my computer, which is weird. Now I can realize one difference between Parallel Jobs and Simple Jobs. It seems that for earlier than 2009 MATLAB editions, it is possible to have more than 4 parallel jobs. What are the other differences between parallel job and simple job? Have a good time Amir "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message <hif09i$75i$1(a)fred.mathworks.com>... > Hi Amir, > > if you are using R2009a or before, you are limited to 4 workers (that's a > limitation of the Parallel Computing Toolbox). If you need more than this, > you will need the MATLAB Distributed Computing Server. > This limit of 4 has been changed to 8 in the current release (R2009b). > > Titus > >
From: Titus Edelhofer on 12 Jan 2010 06:15 Hi Amir, two comments: - your DCT is what now is called PCT (the Distributed Computing Toolbox has been renamed to Parallel Computing Toolbox to better reflect the fact, that not "only" distributed applications can be written but also parallel applications - you can split your job into as many pieces as you like. What makes the performance interesting is, how many of these you can compute in parallel. If you have the MATLAB Distributed Computing Engine, then you can have at least 8 workers available. But you would have to use it: with "sched = findResource('scheduler', 'type', 'local');" you are using the Distributed Computing Toolbox only, without the Engine. You would need to do something like "sched = findResoruce('scheduler', 'type', 'jobmanager');" to use the Engine instead of "only" the Toolbox. With the toolbox only (local scheduler) you are restricted to the 4 workers mentioned before ... Titus "Amir Homayoun " <a.h.javadi(a)gmail.com> schrieb im Newsbeitrag news:hif1db$j3s$1(a)fred.mathworks.com... > Dear Titus > > Thanks. Well. My MATLAB is 7.4.0.287 (R2007a) and my MATLAB Distributed > Computing Engine and Distributed Computing Toolbox are both Version 3.1 > and it seems that I don't have PCT installed on my computer, which is > weird. > Now I can realize one difference between Parallel Jobs and Simple Jobs. It > seems that for earlier than 2009 MATLAB editions, it is possible to have > more than 4 parallel jobs. What are the other differences between parallel > job and simple job? > > Have a good time > Amir > > > "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message > <hif09i$75i$1(a)fred.mathworks.com>... >> Hi Amir, >> >> if you are using R2009a or before, you are limited to 4 workers (that's a >> limitation of the Parallel Computing Toolbox). If you need more than >> this, you will need the MATLAB Distributed Computing Server. >> This limit of 4 has been changed to 8 in the current release (R2009b). >> >> Titus >> >> >
From: Amir Homayoun on 12 Jan 2010 07:52 Dear Titus Thanks. I tried both "sched = findResource('scheduler', 'type', 'jobmanager');" and "sched = findResource('scheduler', 'configuration', 'jobmanager');" and I received the following error message, Warning: Could not contact any job manager lookup process. You may not have started a job manager, or multicast protocols may be failing on your network. If you are certain that a job manager is running, try findResource with a 'lookupURL' input. > In findResource>iCreateAccessor at 305 In findResource>iFindJobManagers at 160 In findResource>iFindScheduler at 263 In findResource at 139 I copied the file "distcompUserConfig.m" to my working directory (current directory) and tried again but still nothing. I looked up in the forum and I saw that another person had the same problem in 2007. Before I continue let me tell you about my system setup. I have one Mac computer with 8 cores and I want to run 8 parallel jobs on the same machine. It is not a network connection. There, in 2007, you mentioned that the distributing computing engine must be started using "mdce" command. I navigated to ...toolbox/distcomp/bin and entered the following command from MATLAB command window !mdce install and it returned me the following error message /bin/bash: mdce: command not found I tried to run the same command from terminal but I think I don't enter the right command and it doesn't find the "mdce" file, although I see it using the "ls" command. I simply write "mdce install" and it gives me the following error message -bash: mdce: command not found Sorry for taking your time Have a good time, Amir "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message <hihljs$fbc$1(a)fred.mathworks.com>... > Hi Amir, > > two comments: > - your DCT is what now is called PCT (the Distributed Computing Toolbox has > been renamed to Parallel Computing Toolbox to better reflect the fact, that > not "only" distributed applications can be written but also parallel > applications > - you can split your job into as many pieces as you like. What makes the > performance interesting is, how many of these you can compute in parallel. > If you have the MATLAB Distributed Computing Engine, then you can have at > least 8 workers available. But you would have to use it: with "sched = > findResource('scheduler', 'type', 'local');" you are using the Distributed > Computing Toolbox only, without the Engine. You would need to do something > like "sched = findResoruce('scheduler', 'type', 'jobmanager');" to use the > Engine instead of "only" the Toolbox. With the toolbox only (local > scheduler) you are restricted to the 4 workers mentioned before ... > > Titus > > "Amir Homayoun " <a.h.javadi(a)gmail.com> schrieb im Newsbeitrag > news:hif1db$j3s$1(a)fred.mathworks.com... > > Dear Titus > > > > Thanks. Well. My MATLAB is 7.4.0.287 (R2007a) and my MATLAB Distributed > > Computing Engine and Distributed Computing Toolbox are both Version 3.1 > > and it seems that I don't have PCT installed on my computer, which is > > weird. > > Now I can realize one difference between Parallel Jobs and Simple Jobs. It > > seems that for earlier than 2009 MATLAB editions, it is possible to have > > more than 4 parallel jobs. What are the other differences between parallel > > job and simple job? > > > > Have a good time > > Amir > > > > > > "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message > > <hif09i$75i$1(a)fred.mathworks.com>... > >> Hi Amir, > >> > >> if you are using R2009a or before, you are limited to 4 workers (that's a > >> limitation of the Parallel Computing Toolbox). If you need more than > >> this, you will need the MATLAB Distributed Computing Server. > >> This limit of 4 has been changed to 8 in the current release (R2009b). > >> > >> Titus > >> > >> > > >
From: Titus Edelhofer on 12 Jan 2010 08:00 Hi, are you sure, you have purchased the MATLAB Distributed Computing Engine? Because this comes with the mdce script ... I remember, that the directory where this script is, was slightly different (maybe subdirectory?). Please search in <MATLAB>\toolbox\distcomp for the mdce script... Titus "Amir Homayoun " <a.h.javadi(a)gmail.com> schrieb im Newsbeitrag news:hihr9i$iv2$1(a)fred.mathworks.com... > Dear Titus > > Thanks. I tried both "sched = findResource('scheduler', 'type', > 'jobmanager');" and "sched = findResource('scheduler', 'configuration', > 'jobmanager');" and I received the following error message, > Warning: Could not contact any job manager lookup process. You may not > have started a job manager, or multicast protocols may be failing on your > network. If you are > certain that a job manager is running, try findResource with a 'lookupURL' > input. >> In findResource>iCreateAccessor at 305 > In findResource>iFindJobManagers at 160 > In findResource>iFindScheduler at 263 > In findResource at 139 > > I copied the file "distcompUserConfig.m" to my working directory (current > directory) and tried again but still nothing. > > I looked up in the forum and I saw that another person had the same > problem in 2007. > Before I continue let me tell you about my system setup. I have one Mac > computer with 8 cores and I want to run 8 parallel jobs on the same > machine. It is not a network connection. > There, in 2007, you mentioned that the distributing computing engine must > be started using "mdce" command. I navigated to ...toolbox/distcomp/bin > and entered the following command from MATLAB command window > > !mdce install > > and it returned me the following error message > > /bin/bash: mdce: command not found > > I tried to run the same command from terminal but I think I don't enter > the right command and it doesn't find the "mdce" file, although I see it > using the "ls" command. I simply write "mdce install" and it gives me the > following error message > > -bash: mdce: command not found > > Sorry for taking your time > > Have a good time, > Amir > > > "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message > <hihljs$fbc$1(a)fred.mathworks.com>... >> Hi Amir, >> >> two comments: >> - your DCT is what now is called PCT (the Distributed Computing Toolbox >> has been renamed to Parallel Computing Toolbox to better reflect the >> fact, that not "only" distributed applications can be written but also >> parallel applications >> - you can split your job into as many pieces as you like. What makes the >> performance interesting is, how many of these you can compute in >> parallel. If you have the MATLAB Distributed Computing Engine, then you >> can have at least 8 workers available. But you would have to use it: with >> "sched = findResource('scheduler', 'type', 'local');" you are using the >> Distributed Computing Toolbox only, without the Engine. You would need to >> do something like "sched = findResoruce('scheduler', 'type', >> 'jobmanager');" to use the Engine instead of "only" the Toolbox. With the >> toolbox only (local scheduler) you are restricted to the 4 workers >> mentioned before ... >> >> Titus >> >> "Amir Homayoun " <a.h.javadi(a)gmail.com> schrieb im Newsbeitrag >> news:hif1db$j3s$1(a)fred.mathworks.com... >> > Dear Titus >> > >> > Thanks. Well. My MATLAB is 7.4.0.287 (R2007a) and my MATLAB Distributed >> > Computing Engine and Distributed Computing Toolbox are both Version 3.1 >> > and it seems that I don't have PCT installed on my computer, which is >> > weird. >> > Now I can realize one difference between Parallel Jobs and Simple Jobs. >> > It seems that for earlier than 2009 MATLAB editions, it is possible to >> > have more than 4 parallel jobs. What are the other differences between >> > parallel job and simple job? >> > >> > Have a good time >> > Amir >> > >> > >> > "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message >> > <hif09i$75i$1(a)fred.mathworks.com>... >> >> Hi Amir, >> >> >> >> if you are using R2009a or before, you are limited to 4 workers >> >> (that's a limitation of the Parallel Computing Toolbox). If you need >> >> more than this, you will need the MATLAB Distributed Computing Server. >> >> This limit of 4 has been changed to 8 in the current release (R2009b). >> >> >> >> Titus >> >> >> >> >> > >> >
From: Amir Homayoun on 12 Jan 2010 08:48
Hi I have "MATLAB Distributed Computing Engine" installed on my machine. Its version is 3.1 (R2007a). I see it when I type "ver" in MATLAB command window. The file "mdce" is placed in <MATLAB>/toolbox/distcomp/bin. I can see it in "finder" and when I type "ls" in terminal window. Amir "Titus Edelhofer" <titus.edelhofer(a)mathworks.de> wrote in message <hihrot$jf8$1(a)fred.mathworks.com>... > Hi, > > are you sure, you have purchased the MATLAB Distributed Computing Engine? > Because this comes with the mdce script ... > I remember, that the directory where this script is, was slightly different > (maybe subdirectory?). Please search in <MATLAB>\toolbox\distcomp for the > mdce script... > > Titus > |