Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 7
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 2543 times and has 6 replies Next Thread
ThreadRipper
Veteran Cruncher
Sweden
Joined: Apr 26, 2007
Post Count: 1324
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Checkpointing intervals suggestion

Hi,

I usually monitor my computer using HWInfo64 and after seeing the shear amount of data being written to my SSD, probably due to default 60 second checkpointing setting in BOINC client, I believe there should be a suggestion given to increase the value.

Okay, I run 64 simultaneous threads which probably makes the matter worse faster, but in the course of 15 days or so there was 2000 GB, Yes 2TB written to my SSD (no other write intensive tasks on the PC during this time). In HWInfo64 I could also see that my "Remainning SSD Life" had decreased from 99% to 98% due to all the writes. Now, a few days ago I set BOINC settings to checkpoing at most every 600 seconds instead and now the SSD abuse seems to have dropped off significantly.

Of course, if one often closes starts boinc or reboots the PC, then 600s would mean a lot of lost compute time perhaps. But as PCs now have many cores even on mainstream segment, as well as SSDs becoming very wide-spread there as well, I believe it is time to revise recommendations and recommend either a RAMDisk or increasing the checkpointing/write-to-disk interval in BOINC settings from 60 seconds.

Has anyone else noticed huge disk writes (especially to SSDs since they are so sensitive to writes) with BOINC default settings, especially if you have many threads (>= 16 perhaps)?
----------------------------------------

Join The International Team: https://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=CK9RP1BKX1

AMD TR2990WX @ PBO, 64GB Quad 3200MHz 14-17-17-17-1T, RX6900XT @ Stock
AMD 3800X @ PBO
AMD 2700X @ 4GHz
[Sep 23, 2020 9:24:00 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Falconet
Master Cruncher
Portugal
Joined: Mar 9, 2009
Post Count: 3315
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Checkpointing intervals suggestion

Which projects are you running?

Doesn't ARP do a lot of disk writes?
----------------------------------------


- AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W
- AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W
- AMD Ryzen 7 7730U 8C/16T 3.0 GHz
[Sep 23, 2020 10:14:01 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Checkpointing intervals suggestion

You can set write to disk at most as long as you like, at one time had it to the max of that parm field which at WCG device profiles means 9999, about 2.8 hours in a day, just to cut on the incessant writing. 2.8 hours on my machine is mostly shorter than the checkpoint interval for ARP1. For 64 threads that still regular writing, like every 2.6 minutes one doing a hit.

The default 60 seconds has long been recommended to be upped, just not of this time anymore, pleistocene almost. WCG on it's own could make the default 10x quite quickly, BOINC world as a whole, in the next millennium. And then a higher default would only work for new/re-created devices profiles out of the box.
----------------------------------------
[Edit 1 times, last edit by Former Member at Sep 23, 2020 1:20:51 PM]
[Sep 23, 2020 1:19:01 PM]   Link   Report threatening or abusive post: please login first  Go to top 
KerSamson
Master Cruncher
Switzerland
Joined: Jan 29, 2007
Post Count: 1684
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Checkpointing intervals suggestion

On my side, I set the checkpoint to 600 sec (10 minutes).
Never the less, even if my system is running with a SSD, I have a dedicated HDD for the boinc-data directory if the machine runs 6+ threads simultaneously.
MIP1 and ARP1 are very good candidates for writing a lot of data to the disk.
Cheers,
Yves
----------------------------------------
[Sep 23, 2020 1:49:13 PM]   Link   Report threatening or abusive post: please login first  Go to top 
ThreadRipper
Veteran Cruncher
Sweden
Joined: Apr 26, 2007
Post Count: 1324
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Checkpointing intervals suggestion

Thanks for all the replies!

I run OPN1, MCM1, MIP1 (limited) and I limit ARP1 to 10 max in the cache settings.
Previously I tried to have the BOINC directory on a HDD, but you know...when booting up the machine or BOINC client it would completely put the machine to a halt while reading all that data from 64WUs for several minutes. Boinc would just sit there empty with the "Connecting to client..." message all this time.

As soon as I moved the directory to SSD that particular issue went away.
I will probably go for a RAMDisk as soon as I have tuned my RAM latencies and have it running stable for a week 24/7. Still, I believe that it may be pretty unclear to many that BOINC will be killing off the SSDs if that setting is not changed. That's my suggestion, to make it less abusive to storage per default.
----------------------------------------

Join The International Team: https://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=CK9RP1BKX1

AMD TR2990WX @ PBO, 64GB Quad 3200MHz 14-17-17-17-1T, RX6900XT @ Stock
AMD 3800X @ PBO
AMD 2700X @ 4GHz
[Sep 23, 2020 3:49:01 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Checkpointing intervals suggestion

You can set a flag in cc_config.xml the delay computing start. I got it at 450 seconds, also because onedrive kicks in and truly brings my system responsiveness to it's knees while checking if there were changes that need synching... hundreds of thousands of files. Stupid as when you shut down or boot, that post boot check should not be needed. Things should have been current at shutdown, but not in MS's eyes.
[Sep 23, 2020 5:10:09 PM]   Link   Report threatening or abusive post: please login first  Go to top 
KerSamson
Master Cruncher
Switzerland
Joined: Jan 29, 2007
Post Count: 1684
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Checkpointing intervals suggestion

Two reasons more for me to avoid onedrive and Windows talk to the hand
Yves
----------------------------------------
[Sep 23, 2020 8:06:41 PM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread