Call for participants
in [Cryptography]
|
Prev: Permutation Extrapolation Function (PXF) (you might find theend of the post useful)
Next: Collecting true randomness/entropy from decimal part of the values calculated in stock exchange market
From: Mok-Kong Shen on 6 Apr 2010 10:46 Chip Eastham wrote: > What operating system are you using? The FAQ at > the BOINC website (a utility for sharing workload > used by the NFS(a)HOME project and others) suggests > that if CPU limits are needed, these must be set > by an external utility (external to the application). I use Windows 7 and XP. I don't think that there are readily available mechanisms to control CPU usage. (Once I sought such a utility for my own purposes, but failed). My main point is that I believe quite some people may have the same thinking like me, namely one desires to "retain" one's control over one's own computer (at all time). If that condition could be satisfied, then it's likely that more people would like to donate their machine power than in case it is not satisfied. The ideal situation in my view is the following: There is an icon for the project. With it one can download a new task. Each time the computer goes up, the process automatically resumes running at a (user-set, default) CPU utilization limit. With the icon one can at any time change the CPU utilization (e.g. when one goes to lunch or when one has some computing intensive job of one's own). When the task ends, one gets a message telling one to upload the result via the icon. If one forgets to do so, one gets a reminder at some regular intervals (and anyway each time when the computer is switched on). An internet connection should not be necessary when the task is running, particularly so because there may be computers that are not connected to the internet at all. M. K. Shen
From: Chip Eastham on 6 Apr 2010 13:31 On Apr 6, 10:46 am, Mok-Kong Shen <mok-kong.s...(a)t-online.de> wrote: > Chip Eastham wrote: > > What operating system are you using? The FAQ at > > the BOINC website (a utility for sharing workload > > used by the NFS(a)HOME project and others) suggests > > that if CPU limits are needed, these must be set > > by an external utility (external to the application). > > I use Windows 7 and XP. I don't think that there are > readily available mechanisms to control CPU usage. (Once > I sought such a utility for my own purposes, but failed). > > My main point is that I believe quite some people may > have the same thinking like me, namely one desires to > "retain" one's control over one's own computer (at all > time). If that condition could be satisfied, then it's > likely that more people would like to donate their machine > power than in case it is not satisfied. > > The ideal situation in my view is the following: There is > an icon for the project. With it one can download a new task. > Each time the computer goes up, the process automatically > resumes running at a (user-set, default) CPU utilization limit. > With the icon one can at any time change the CPU utilization > (e.g. when one goes to lunch or when one has some computing > intensive job of one's own). When the task ends, one gets a > message telling one to upload the result via the icon. If one > forgets to do so, one gets a reminder at some regular intervals > (and anyway each time when the computer is switched on). An > internet connection should not be necessary when the task is > running, particularly so because there may be computers that > are not connected to the internet at all. > > M. K. Shen I think your points are good ones. In Windows one can specify the "priority" at which a process should run, which generally allows the desired effect of making the computer stay responsive to the user while allocating "spare" cycles to the background process. See the "Set Priority" item on the right-click menu of the Processes tab of Task Manager. Looking through BOINC's "defect tracking", I see that lack of an internet connection can cause "hangs": [BOINC daemon hangs when internet not available] http://boinc.berkeley.edu/trac/ticket/938 Probably the BOINC approach is better for desktop machines that stay connected to the internet on a more or less consistent basis. It appears from the Wiki documentation that some thought has been given to managing internet disconnects in a more graceful way, but this isn't a top concern for me. If I disconnect from the internet temporarily, eg. to reboot my router, I consider it necessary interference with the BOINC client's progress. regards, chip
From: Mok-Kong Shen on 6 Apr 2010 16:02 Chip Eastham wrote: > Mok-Kong Shen wrote: [snip] >> The ideal situation in my view is the following: There is >> an icon for the project. With it one can download a new task. >> Each time the computer goes up, the process automatically >> resumes running at a (user-set, default) CPU utilization limit. >> With the icon one can at any time change the CPU utilization >> (e.g. when one goes to lunch or when one has some computing >> intensive job of one's own). When the task ends, one gets a >> message telling one to upload the result via the icon. If one >> forgets to do so, one gets a reminder at some regular intervals >> (and anyway each time when the computer is switched on). An >> internet connection should not be necessary when the task is >> running, particularly so because there may be computers that >> are not connected to the internet at all. > > I think your points are good ones. In Windows > one can specify the "priority" at which a > process should run, which generally allows the > desired effect of making the computer stay > responsive to the user while allocating "spare" > cycles to the background process. See the "Set > Priority" item on the right-click menu of the > Processes tab of Task Manager. > > Looking through BOINC's "defect tracking", I see > that lack of an internet connection can cause > "hangs": > > [BOINC daemon hangs when internet not available] > http://boinc.berkeley.edu/trac/ticket/938 > > Probably the BOINC approach is better for desktop > machines that stay connected to the internet on > a more or less consistent basis. It appears from > the Wiki documentation that some thought has been > given to managing internet disconnects in a more > graceful way, but this isn't a top concern for me. > If I disconnect from the internet temporarily, > eg. to reboot my router, I consider it necessary > interference with the BOINC client's progress. I can't see any reason why the task can't run without constant connection to the server for purposes of NFS and those of some similar researches. It would be fine, if someone could transmit that viewpoint to BOINC and similar software producers. After that problem is solved, the CPU-load issue could be solved to some more or less agreeable extent for most potential participants, I surmise. M. K. Shen
From: Chip Eastham on 6 Apr 2010 16:35 On Apr 6, 4:02 pm, Mok-Kong Shen <mok-kong.s...(a)t-online.de> wrote: > Chip Eastham wrote: > > Mok-Kong Shen wrote: > [snip] > >> The ideal situation in my view is the following: There is > >> an icon for the project. With it one can download a new task. > >> Each time the computer goes up, the process automatically > >> resumes running at a (user-set, default) CPU utilization limit. > >> With the icon one can at any time change the CPU utilization > >> (e.g. when one goes to lunch or when one has some computing > >> intensive job of one's own). When the task ends, one gets a > >> message telling one to upload the result via the icon. If one > >> forgets to do so, one gets a reminder at some regular intervals > >> (and anyway each time when the computer is switched on). An > >> internet connection should not be necessary when the task is > >> running, particularly so because there may be computers that > >> are not connected to the internet at all. > > > I think your points are good ones. In Windows > > one can specify the "priority" at which a > > process should run, which generally allows the > > desired effect of making the computer stay > > responsive to the user while allocating "spare" > > cycles to the background process. See the "Set > > Priority" item on the right-click menu of the > > Processes tab of Task Manager. > > > Looking through BOINC's "defect tracking", I see > > that lack of an internet connection can cause > > "hangs": > > > [BOINC daemon hangs when internet not available] > >http://boinc.berkeley.edu/trac/ticket/938 > > > Probably the BOINC approach is better for desktop > > machines that stay connected to the internet on > > a more or less consistent basis. It appears from > > the Wiki documentation that some thought has been > > given to managing internet disconnects in a more > > graceful way, but this isn't a top concern for me. > > If I disconnect from the internet temporarily, > > eg. to reboot my router, I consider it necessary > > interference with the BOINC client's progress. > > I can't see any reason why the task can't run without > constant connection to the server for purposes of NFS > and those of some similar researches. It would be fine, > if someone could transmit that viewpoint to BOINC and > similar software producers. After that problem is > solved, the CPU-load issue could be solved to some > more or less agreeable extent for most potential > participants, I surmise. > > M. K. Shen I suspect that particular problem occurs only under some odd conditions, not the general case. The BOINC manager allows you to limit the computer's resources used in terms of when it connects to the internet, how much CPU is used (in a very coarse grained manner), how much memory and disk, and how often it switches between projects (remember it's designed to allow participation in more than just NFS(a)HOME). See the BOINC wiki about this Preferences dialog here: http://boinc.berkeley.edu/wiki/Preferences regards, chip
From: Mok-Kong Shen on 7 Apr 2010 03:15
Chip Eastham wrote: > I suspect that particular problem occurs only under some > odd conditions, not the general case. The BOINC manager > allows you to limit the computer's resources used in > terms of when it connects to the internet, how much CPU > is used (in a very coarse grained manner), how much > memory and disk, and how often it switches between projects > (remember it's designed to allow participation in more > than just NFS(a)HOME). > > See the BOINC wiki about this Preferences dialog here: > > http://boinc.berkeley.edu/wiki/Preferences I didn't mean that's a current "operational" problem for the already paarticipating users which is inconvenient etc., but that's a "principal" problem in the sense that it tends to render a participation to such internet collective research projects less attractive to many people (because one doesn't want to loose "whole" control of one's computer) or impossible (because the machine power to be donated is on computers that have no connection to the internet). As to CPU-usage, one shouldn't in my humble view let the burden of figuring out how to best limit the usage (see my previous post for a desired situation) to be on the side of a potential participant. Such a facilitiy should be provided from the very beginning. (For, it is the projects that want to have something from the potential participants, not the other way round.) So I believe, if people initiating such projects desire to "maximize" the computing power they want to get, some thoughts in the direction I indicated should perhaps be done. M. K. Shen |