Staggering on a HP DL380 G7 plotman or swar

Hi. i will be adding 72tb jbod to my farming rig tonight and im planning on installing linux and plotman or swar on my HP DL380 G7 server that i use for plotting, i have used windows and powershell to do plotting up untill now, a have done 11 plots simultant in 22h all at once but im thinking that i migt get more out of it if i use linux and a manager, the server have dual xeon x5650 cpuer in it and 64gb ram and i have more to put in, the temp drive is a total of 4tb ( 8 sas drive raid 0 )

anyone here that can give some input on a good starting config to plot on it ? like staggering time and how many plots, with windows i did 11 witout problem but there they all was runnig so what i understand with stagering i might can do more ?

1 Like

Timings are hard and specific system dependent.

You’re good starting with one hour and observing tmp space filling up.

You can do the arithmetic knowing:

  • It’s possible to fit in 4 or 5 processes by TB with careful staggering as phases 1 and 2 are filling while 3 and 4 freeing the space. The precise number and staggering time depends on how fast processor turns out to be.
  • You’re limited in parallel processes by something 1.5× the number of cores as a rule of thumb.
  • A process takes 3.6 GB of RAM.

If you get clogged you can suspend a process with kill -STOP <pid>, move out its files (hash in the name is in the log file) and the rest will resume automatically after short time; then move back in and kill -CONT <pid>.

I’d also try with 8 separate tmp mount points instead of a raid0 as plotman can marshal between them and there’s lot of random seeks that kill performance, especially with parallel plots. Limiting to 1 plot per mechanical drive could turn out faster in TB/day than trying 18 parallel to a raid0.

Great info, then i think i Will do 8 temp drives for it to use