| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 6
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have been using BOINC for a while now, and recently i have noticed that the hdc app seems to have problems when it is running on my machine. I have a lenovo T60, with a core duo 2.0ghz processor, and 1 gb of ram. I am running Ubuntu, and every time i run hdc, every couple of minutes i see that all of the sudden, the processor stops recieving work, disk usage goes through the roof, and the ram starts going through the roof as well. During these times, i am finding that my computer runs slow, especially right after resuming my computer from sleep mode. These issues only seem to happen when running hdc, and faah seems to run great, with no noticable system usage. The peek memory usage that hdc has ever used so far has been a little over 1.8gb, which is GIGANTIC, especially since i have 1 g of ram, and 1 g of swap, which means my computer was probably pretty close to crashing. Why is hdc having so many problems?
|
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
The memory footprint of all but HDC are small. The disk IO which is the part that stops the actual crunching, because it happens at a higher priority, is when running very large models of 1gb+ and Virtual Memory or Checkpoint saves take place. Defragmenting the disk and using the PageDefrag utility to be run at boot time (once or every time), can reduce much of the slow down. If still suffering, you can de-select HDC.
----------------------------------------A New option in the BOINC 5.8.8, just released is the 'use memory while busy and 'use memory when idle'. The default is 50% busy / 90% idle. It would simply put the HDC job in a 'waiting to run' state and start the next job. With a mix of FAAH and HDC the next one is very likely a FAAH job. The above noted options are not yet all accessible thru the website or thru BOINC preference controls, so have to be put manually in an text file called global_prefs_override.xml
Lines explained in brief: Batteries: 0 means not to run on battery. Leave in memory: 1 makes that project switching goes without crunching loss, though checkpoint switching is better in 5.8.8 Max CPUS: 1 on a dual core machine 1 would only use half the crunch capacity. Don't know what a T60 has. Disk interval: 999 seconds keeps the model for 18 minutes without writing to disk unless a checkpoint is found. ram.... busy: 60 the important one is my choice on a 1.5gb ram machine ram.... idle: 90 is the default and leaves some for other processes. Will post if i find a file explaining them all in detail. When saving the file to the BOINC dir, make sure to save it with the xml extension and not txt. Else just rename it an file manager. 5.8.8 for Linux is not gold at time of this writing, but for MAC and Windows it is!
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 4 times, last edit by Sekerob at Feb 4, 2007 12:34:55 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello onlngenius,
HDC is our biggest project ( http://www.worldcommunitygrid.org/help/viewTopic.do?shortName=minimumreq ). At intervals the working set of memory pages jumps up to (from memory, I am uncertain) about 545 MB, then after a period drops back down more than 300 MB. Are you running 2 threads of HDC at the same time? 2 simultaneous threads of HDC, both in the large array section of the algorithm, would eat up 1 GB of memory and thrash a bit as working parts of memory were forced to use the swap file of your hard disk. With only 1 GB of swap space, you are probably pushing the limits and coming close to crashing your computer. Even when the program is demanding 545 MB of working RAM, it is also using several hundred megabytes of virtual memory to hold arrays that it is not currently working on. Just for safety, you might consider increasing the swap file space. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I've had to give up on HDC. It's just too big and too painful to run in background on my machine (T43, 1GB Memory 2GB swap file minimum).
I become VERY conscious of when HDC is running and have had to remove it from my possible projects as, even when it's running at idle priority, it pushes me up well beyond physical memory and EVERYTHING swaps too painfully to get work done. Sorry HDC developers, but, unless you make the footprint similar to other projects, I can't help participate in the research. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello randyg,
Lenovo Thinkpad T60 - here is a link into a thread discussing slowdowns on it and how to solve it: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=1804#81379 I forgot about this problem. We have a query in at Lenovo asking them to look at the situation. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
i have a t41, no problems at all.
|
||
|
|
|