University of Maryland Mike P. Cummings  
Center for Bioinformatics and Computational Biology
HomeResearchPublicationsPersonnel

The Lattice Project
About Lattice
Applications
Client Activity
Create Account
Message Boards
Participant Profiles
Questions & Answers
Research Projects
Rules and Policies
Statistics
Teams
Top Computers
Top Participants
Top Teams
Your Account

BOINC Logo



Forum Thread

BOINC Client Activity
log in

Advanced search

Message boards : News : BOINC Client Activity

Author Message
Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4510 - Posted: 12 Oct 2010, 18:43:44 UTC
Last modified: 12 Oct 2010, 18:45:54 UTC

Check out our new visualization of BOINC client activity!

(powered by Google Maps)

Ironworker16
Avatar
Send message
Joined: 3 Jul 05
Posts: 79
Credit: 1,871,962
RAC: 164
Message 4515 - Posted: 15 Oct 2010, 21:33:44 UTC - in response to Message 4510.

I like it! I looks like someone spent a lot of time on it and it is appreciated.
____________

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4516 - Posted: 15 Oct 2010, 23:56:29 UTC - in response to Message 4515.

Thanks. Jingxin Han and Kareem Johnson were largely responsible for the development.

Professor Ray
Send message
Joined: 15 Feb 08
Posts: 61
Credit: 24,874
RAC: 28
Message 4517 - Posted: 16 Oct 2010, 4:27:45 UTC

Smells like somebody's a coding maniac; wave of the magic wand and the 64-bit Garli app is done and now this.

I'm thankfull that the chains I'm shackled to are long enough to allow me to peer through the grating of my cell so as to view nirvana in some miniscule measure of its glory through the grating where gruel is poured through fairly regularly.

I now have to slink back to the dank dark recesses of my dungeon before they hear my thoughts to work on my escape plot, i.e., implementing a native C# class as a class module for a VB app that'll determine all non-reparse point directories in a selected directory, move them to an alternate location on another volume and create reparse points in the initial directory to those directories so moved; two different source directories are implictated w/explicit 1:1 relationship between source - destination in that regard. However, a requirement is that all *.msi & *.msp files existing in the other source directory likewise be moved to another location on another volume except that the associated {xxx...yyy} registry key REG_SZ values associated with the *.msi & *.msp files be accordingly modified.

The afforementioned shall occur with one button click.

Admittedly, and to my chagrin, not as cool as what you guys are doing, my hope is that it may be worthy enough behavior of a vermiform species, although not all that colorful, sufficiently amusing behavior whereby I can be scooped out of the muck I live in before I wholly succum to nihilism.

I DO mean all of that nicely.

Isn't it interesting how psychotic the long term-unemployed become? Ha ha ha ha ha ha...

P.S. There has got to be a way for the Lattice server to ascertain the memory config of the host-OS such that those that are willing to throw their Wright Flyer computing machines at a Garli WU, they should be afforded that opportunity (despite the very real potential that they may not be able to shut down for several weeks - based on the previous batch where there was a serious issue w/checkpointing); let the chips fall where they may.

Host owners are either reading these boards or they're not. If they're not, then most already detached this summer would be my guess.

My existing system will NOT be able to participate in the speculated Garli WU processing.

I'm arguing on behalf of those who have 4+ GB onboard - despite 32-bit OS - who may be interested in Garli WU crunching. It sounds to me like you're sending 'burn notice' to all those machines.

I can understand if the science absolutely requires a processing power of machines that can be considered 'elite'. If so, then such requirement should be prominently displayed on the web-site somewhere.

Pepo
Avatar
Send message
Joined: 16 Nov 07
Posts: 33
Credit: 13,982
RAC: 0
Message 4522 - Posted: 18 Oct 2010, 16:36:00 UTC - in response to Message 4517.

Professor Ray wrote:
P.S. There has got to be a way for the Lattice server to ascertain the memory config of the host-OS such that those that are willing to throw their Wright Flyer computing machines at a Garli WU, they should be afforded that opportunity [...]

My existing system will NOT be able to participate in the speculated Garli WU processing.

I'm arguing on behalf of those who have 4+ GB onboard - despite 32-bit OS - who may be interested in Garli WU crunching.

According to the messages Adam has posted here around and in the BOINC message lists, and according to your technical post ;-) you surely understand, that "having 4+ GB RAM onboard" is pretty irrelevant in this case - it might also work with as low as 1 GB RAM :-) albeit a bit slowly. What counts is how large user-mode virtual address space is given to a particular user app on particular Win OS version. Whether it is

    a) below 2GB (32-bit system with no 4GT setting or any system with apps without /LARGEADDRESSAWAREness) - the huge Garli apps will most probably fail, or
    b) rather 2-3 GB (depending on the app's /LARGEADDRESSAWAREness and 32-bit system's 4GT settings), with already some chance for the apps to survive, or
    c) even 4-N GB (64-bit system with any /LARGEADDRESSAWARE apps).


Unfortunately as the BOINC client does not propagate this subtle detail to the server, it is just as sure as the Russian roulette, whether the delivered task will fly or crash and explode...

PS: Maybe at this point Adam might initiate a request on BOINC team to let the client find out and be aware of this detail? Other projects might perhaps also profit from this knowledge soon... (Currently, pretty much details about processor, disk, amount of available RAM+swap, etc. are propagated to the project in sched_request's <host_info> tag, but the address space's size is ... simply considered as available.)

Peter

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4523 - Posted: 18 Oct 2010, 17:12:02 UTC - in response to Message 4522.

Actually, David did check in a scheduling change that addresses precisely this issue, when I reported it, and we're using it - which is why we've seen far fewer crashes on 32-bit machines. I feel like this issue is resolved, as much as it needs to be, but time will tell.

Pepo
Avatar
Send message
Joined: 16 Nov 07
Posts: 33
Credit: 13,982
RAC: 0
Message 4524 - Posted: 18 Oct 2010, 17:36:56 UTC - in response to Message 4523.

I know, I've read his response (in "Windows BOINC clients running out of memory" thread), but my feeling was that he (over)simplified it in his description a bit - just taking the OS-bitness and "sufficient RAM" into account and ignoring the available address space and "(in)sufficient swap". (Like noted, there could be a plenty of 32-bit hosts, which would reliably crunch a 2.3 GB job.)

But I'm also aware that he has successfully solved the main goal - that the apps won't crash after being delivered on an inappropriate host.

Peter

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4525 - Posted: 18 Oct 2010, 17:43:08 UTC - in response to Message 4524.

Now I follow you a little better- agree, there is probably room for improvement.


Post to thread

Message boards : News : BOINC Client Activity

.......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... ..........

University of Maryland     UM Home | Directories | Search | Admissions | Calendar
Copyright © 2017 The Lattice Project
Direct questions and comments to Lattice Admin