Compressing Rosetta Executables

Message boards : Number crunching : Compressing Rosetta Executables

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Fynjy

Send message
Joined: 18 Sep 06
Posts: 8
Credit: 3,688,731
RAC: 0
Message 31889 - Posted: 1 Dec 2006, 7:34:36 UTC

Hi, Chu!
I had no problems with any cores yet, but i thnk that there is a traffic problem with new cores wich updating every month now. Especially for crunchers who have lots of computers. What do you think about it? I try to copy new cores to the folder with Rosetta and it works. Maybe it would be better to place new cores on your website for free upload? And add to perferences of Rosetta something like ability to update cores manualy or auto. It\'s only one way. Maybe you can give another guess.
Best regards!
Help people! Join TSC!Russia!
ID: 31889 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Robinski

Send message
Joined: 7 Mar 06
Posts: 51
Credit: 45,396
RAC: 0
Message 31891 - Posted: 1 Dec 2006, 9:26:43 UTC - in response to Message 31889.  

Hi, Chu!
I had no problems with any cores yet, but i thnk that there is a traffic problem with new cores wich updating every month now. Especially for crunchers who have lots of computers. What do you think about it? I try to copy new cores to the folder with Rosetta and it works. Maybe it would be better to place new cores on your website for free upload? And add to perferences of Rosetta something like ability to update cores manualy or auto. It\'s only one way. Maybe you can give another guess.
Best regards!


You could, if all clients are in the same network (it also works if it is not, but slower) use a proxy server.
This proxy server is able to cache files downloaded from rosetta (like the core) and transfer the files very quick to the clients.

and you only need to download the file once from rosetta, so the servers are much less stressed, and you get the cores on all machines very fast.
Member of the Dutch Power Cows

Trying to get the world on IPv6, do you have it? check here: IPv6.RHarmsen.nl
ID: 31891 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Fynjy

Send message
Joined: 18 Sep 06
Posts: 8
Credit: 3,688,731
RAC: 0
Message 31898 - Posted: 1 Dec 2006, 13:40:10 UTC
Last modified: 1 Dec 2006, 13:43:03 UTC

Robinski It\'s a pity but there are many people who have no access to proxy.
Chy My team-mate has one more gess about how to reduce traffic with new cores. You can simply compress it! New core with size 9 Mb comresses to 1.9 (!) Mb by 7-Zip. If you plan to update you cores so often I prefer to download 1.9 Mb instead of 9!
Help people! Join TSC!Russia!
ID: 31898 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Fetisoff

Send message
Joined: 30 Sep 05
Posts: 9
Credit: 28,643
RAC: 0
Message 31901 - Posted: 1 Dec 2006, 14:18:20 UTC
Last modified: 1 Dec 2006, 14:18:58 UTC

Why not announce the fact that the new client version is to be released very soon (I know RALPH is an indication, but yet again, not all versions tested in RALPH actually make it over to Rosetta), and place a link to the executable file on this page http://boinc.bakerlab.org/apps.php a day (maybe even few hours in advance) giving traffic-concious crunchers enough time to download it themselves, before BOINC actually has to...??
ID: 31901 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1740
Credit: 3,655,614
RAC: 5
Message 31903 - Posted: 1 Dec 2006, 16:04:36 UTC

All of this is off-topic and a new thread should be opened. But simplest way to get a day advanced \"notice\" of a new Rosetta release is to keep a day of work on deck by increasing your setting for \"connect to network every ... days\" in your general preferences. Then check the current Rosetta version. The new versions ARE available for download, whenever you like, as described in this QA item.
If having a DC project with BOINC is of interest to you, with volunteer or cloud computing resources, but have no time for the BOINC learning curve,
use a hosting service that understands BOINC projects: http://DeepSci.com
ID: 31903 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
FluffyChicken
Avatar

Send message
Joined: 1 Nov 05
Posts: 1260
Credit: 317,557
RAC: 0
Message 31904 - Posted: 1 Dec 2006, 16:12:46 UTC - in response to Message 31898.  
Last modified: 1 Dec 2006, 16:17:09 UTC

Robinski It\'s a pity but there are many people who have no access to proxy.
Chy My team-mate has one more gess about how to reduce traffic with new cores. You can simply compress it! New core with size 9 Mb comresses to 1.9 (!) Mb by 7-Zip. If you plan to update you cores so often I prefer to download 1.9 Mb instead of 9!


Not blackbird by any chance? We had this dicussion almost a year ago and they have still not compressed the executable (ClimatePrediction do it already) BOINC has a built in ZIP capability, and there is of course gzip that they already use as well.

EDIT to add, http://boinc.bakerlab.org/rosetta/forum_thread.php?id=756 Have a good read. Please start a new thread though and do not resurect that one, just link to it :-)

But for your information in future downloads are always at http://boinc.bakerlab.org/download
Team mauisun.org
ID: 31904 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Tymbrimi
Volunteer moderator
Project administrator
Avatar

Send message
Joined: 22 Aug 06
Posts: 148
Credit: 153
RAC: 0
Message 31919 - Posted: 1 Dec 2006, 23:33:56 UTC

Moved this thread out of the Rosetta 5.41 error thread.
Rosetta Moderator: Mod.Tymbrimi
ROSETTA@home FAQ
Moderator Contact
ID: 31919 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1617
Credit: 62,913,403
RAC: 30,128
Message 31920 - Posted: 2 Dec 2006, 0:00:33 UTC

For now, i\'m sure someone could write a batch file or small program that runs on each cruncher and which copies any *.exe or *.gz files from one computers rosetta folder to a central one, and also pick up any that it doesn\'t have already from that share.

It could then be run as a scheduled task every hr or so.
ID: 31920 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Agafonov_KY

Send message
Joined: 18 Jul 06
Posts: 2
Credit: 16,803,114
RAC: 2,187
Message 31951 - Posted: 2 Dec 2006, 17:45:41 UTC - in response to Message 31889.  

I have recently switched almost all of my machines to Rosetta@Home, reason being is that i think this project has the most potential, compared to other DC projects. However, for me personally traffic is somewhat of an issues. I can deal with downloading/uploading of units, however downloading a 9Mb core on 80 machines (720Mb in total) is a problem; especially considering the fact, that it could have easily been compressed to 1.5Mb (600Mb of traffic does make a difference to me). While I\'ll still continue crunching Rosetta, I would really appreciate if the developers would look into this problem, and done something about it
http://stats.free-dc.org/tag.php?name=Agafonov_KY
ID: 31951 · Rating: 13 · rate: Rate + / Rate - Report as offensive    Reply Quote
MattDavis
Avatar

Send message
Joined: 22 Sep 05
Posts: 206
Credit: 1,377,748
RAC: 0
Message 31961 - Posted: 2 Dec 2006, 19:30:31 UTC - in response to Message 31951.  

I have recently switched almost all of my machines to Rosetta@Home, reason being is that i think this project has the most potential, compared to other DC projects. However, for me personally traffic is somewhat of an issues. I can deal with downloading/uploading of units, however downloading a 9Mb core on 80 machines (720Mb in total) is a problem; especially considering the fact, that it could have easily been compressed to 1.5Mb (600Mb of traffic does make a difference to me). While I\'ll still continue crunching Rosetta, I would really appreciate if the developers would look into this problem, and done something about it


I have this image in my head of 80 computers at your house all sharing a single cable modem :)
ID: 31961 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rytis

Send message
Joined: 17 Sep 05
Posts: 8
Credit: 164,018
RAC: 0
Message 31972 - Posted: 2 Dec 2006, 21:08:43 UTC

UPX could be used for exe file compression (http://upx.sourceforge.net/). The compressed files retain .exe extension and suffer no speed loss when starting/during execution. Compression of 5.41 (win) with default settings went from 9261056 bytes to 2633728 bytes.
PrimeGrid
Administrator
ID: 31972 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1740
Credit: 3,655,614
RAC: 5
Message 31977 - Posted: 2 Dec 2006, 23:44:34 UTC - in response to Message 31951.  

Agafonov_KY
...I would really appreciate if the developers would look into this problem, and done something about it


They are indeed working on doing the compression.

...downloading a 9Mb core on 80 machines (720Mb in total) is a problem


I can\'t help make up for the current lack of compression... but I can help you dispurse that network burden across several days, rather then all occuring within hours. This QA item was written with modem users in mind. With 80 machines, I trust you\'re not on a dial-up connection :) ...by using a large work unit run time, and keeping a cache of work on-hand, you connect less frequently, and the downloads are dispursed across a few days. I see now you already HAVE a fairly large WU runtime on the two machines I hit randomly, and that you appear to also have a 2 or 3 day supply of work, so you\'ve already implemented my suggestion on those points.

The other thing you can do is download the new file once, and then drop it on to each machine. You\'d have to find a way to automate. But you can get a copy of the executable as described here.

You might also want to limit BOINC to only use the network at night. That is what I do. This is done in your General Preferences, then update to project, then set BOINC Manager Activity tab to say \"Network activity based on preferences\". Give it at least a few hours during which it can use the network to help assure you keep the work flowing, but it works great. All the downloads happen when noone is around to experience poor network access.

Wow, over a million credits since July!! Keep crunchin\' Rosetta!!!
If having a DC project with BOINC is of interest to you, with volunteer or cloud computing resources, but have no time for the BOINC learning curve,
use a hosting service that understands BOINC projects: http://DeepSci.com
ID: 31977 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
FluffyChicken
Avatar

Send message
Joined: 1 Nov 05
Posts: 1260
Credit: 317,557
RAC: 0
Message 31993 - Posted: 3 Dec 2006, 8:31:55 UTC - in response to Message 31977.  

Agafonov_KY
...I would really appreciate if the developers would look into this problem, and done something about it


They are indeed working on doing the compression.

...downloading a 9Mb core on 80 machines (720Mb in total) is a problem


I can\'t help make up for the current lack of compression... but I can help you dispurse that network burden across several days, rather then all occuring within hours. This QA item was written with modem users in mind. With 80 machines, I trust you\'re not on a dial-up connection :) ...by using a large work unit run time, and keeping a cache of work on-hand, you connect less frequently, and the downloads are dispursed across a few days. I see now you already HAVE a fairly large WU runtime on the two machines I hit randomly, and that you appear to also have a 2 or 3 day supply of work, so you\'ve already implemented my suggestion on those points.

The other thing you can do is download the new file once, and then drop it on to each machine. You\'d have to find a way to automate. But you can get a copy of the executable as described here.

You might also want to limit BOINC to only use the network at night. That is what I do. This is done in your General Preferences, then update to project, then set BOINC Manager Activity tab to say \"Network activity based on preferences\". Give it at least a few hours during which it can use the network to help assure you keep the work flowing, but it works great. All the downloads happen when noone is around to experience poor network access.

Wow, over a million credits since July!! Keep crunchin\' Rosetta!!!


It\'s a shame all the original Roestta@home problems kept TSC! Russia away when we moved from Find-a-Drug. Large and frequent files and no consistent point scoring, I guess Rosetta@Home lost out to Folding@Home on that.
Team mauisun.org
ID: 31993 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile RuDiablo
Avatar

Send message
Joined: 11 Nov 05
Posts: 2
Credit: 463,636
RAC: 0
Message 32000 - Posted: 3 Dec 2006, 10:21:08 UTC - in response to Message 31993.  

FluffyChicken
It\'s a shame all the original Roestta@home problems kept TSC! Russia away when we moved from Find-a-Drug. Large and frequent files and no consistent point scoring, I guess Rosetta@Home lost out to Folding@Home on that.


Regrettably in Russia quick and unlimited Internet until wide-spread. I live in large city, but have only dial-up. :( No ADSL, no cable modem. May be next year...

If possible effectively compress core, why not to do this? This will increase quantity of participants.

Hi Fluffy You understand problems of russian crunchers.
ID: 32000 · Rating: 2 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1617
Credit: 62,913,403
RAC: 30,128
Message 32006 - Posted: 3 Dec 2006, 14:07:30 UTC - in response to Message 31993.  
Last modified: 3 Dec 2006, 14:12:11 UTC

It\'s a shame all the original Roestta@home problems kept TSC! Russia away when we moved from Find-a-Drug. Large and frequent files and no consistent point scoring, I guess Rosetta@Home lost out to Folding@Home on that.

It\'d be a massive boost to the project if TSC! Russia could do here what you did at FaD ;)

Could Blackbird or someone else could produce a small app that copies the .exe and .gz files that as i suggested earlier for those with farms like Agofanov - it\'d mean massively reduced downloads and would work even after compression has been increased.

That way you\'ll only have to download any given file once.

Danny
ID: 32006 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Christoph

Send message
Joined: 10 Dec 05
Posts: 57
Credit: 1,509,845
RAC: 0
Message 32019 - Posted: 3 Dec 2006, 18:39:13 UTC

But when BOINC notices that it uses an old version of the Rosetta app, it wants to download it automatically. The program would have to be run BEFORE BOINC notices the old version.
ID: 32019 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Feet1st
Avatar

Send message
Joined: 30 Dec 05
Posts: 1740
Credit: 3,655,614
RAC: 5
Message 32021 - Posted: 3 Dec 2006, 18:54:57 UTC

I\'ve been dreaming for some time of a BOINC proxy. It would allow control on what projects are allowed. You would point each machine to the proxy as though it were the BOINC project, and it would pass requests on to the true project. It would cache files that can be cached and serve them out locally. I could limit bandwidth used, and allow a number of other controls. You could even scan and/or log all data sent out to get some confidence that no private customer data is leaving your enterprise.

I believe such a thing might help to eliminate some of the barriors that the corporate world has to running BOINC. Such a central proxy would give them control. You\'d only be able to run projects that are allowed. By giving these types of controls, I think the corporate world would be more comfortable running BOINC.

Does anyone know if such a thing has already been done somewhere? There\'s nothing specific to Rosetta required with what I have in mind (other then perhaps the definition of which files can be cached).
If having a DC project with BOINC is of interest to you, with volunteer or cloud computing resources, but have no time for the BOINC learning curve,
use a hosting service that understands BOINC projects: http://DeepSci.com
ID: 32021 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1617
Credit: 62,913,403
RAC: 30,128
Message 32027 - Posted: 3 Dec 2006, 20:44:21 UTC - in response to Message 32019.  

I think that\'s a good idea Feet1st - that way the only computer accessing the internet is the proxy.

But when BOINC notices that it uses an old version of the Rosetta app, it wants to download it automatically. The program would have to be run BEFORE BOINC notices the old version.

My suggestion would result in the sacrifice of some disk space, but would run before BOINC requested any files.

My suggestion is this:

Create a small exe or bat file to run every 1/2 hour or so as a scheduled task and have it do the following:

1. Copy any *.exe or *.gz files from its own rosetta folder to a network share if they don\'t exist there already.
2. Copy any *.exe or *.gz files from that network share to its rosetta folder if they don\'t exist there already.

Then when BOINC checks for the file it\'ll already be there and so won\'t need downloading except by the first computer that requests each file. That\'s got to be pretty easy to implement. I\'m not sure how you would deal with the folder size build up - i guess you can have the program delete files that haven\'t been accessed for x days.

Using a proxy or Feet1st\'s suggestion are more elegant, but I can\'t see why this wouldn\'t be a good workaround where a proxy can\'t be used.
ID: 32027 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Fetisoff

Send message
Joined: 30 Sep 05
Posts: 9
Credit: 28,643
RAC: 0
Message 32031 - Posted: 3 Dec 2006, 22:00:31 UTC

All of these ideas would help, however they take time to implement. All we are really asking for now is to just let us know in advance, when the new version is coming out, so someone like Agafonov can just download the file once, and then manually copy it to all of his computers
ID: 32031 · Rating: 4 · rate: Rate + / Rate - Report as offensive    Reply Quote
TyumenOvers
Avatar

Send message
Joined: 29 Sep 06
Posts: 1
Credit: 820,609
RAC: 0
Message 32066 - Posted: 4 Dec 2006, 15:42:53 UTC

Hello, I\'m a TSC! Russia member, my Rosetta@Home nickname is TyumenOvers. I want to join to Agafonov_KY\'s petition, because the excessive traffic is a serious problem for me too.
ID: 32066 · Rating: 1 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · Next

Message boards : Number crunching : Compressing Rosetta Executables



©2019 University of Washington
http://www.bakerlab.org