search

MrPig91 / PSChiaPlotter

A repo for powershell module that helps Chia Plotting
MIT License
182 stars 47 forks source link

Increase the number of disks for the TEMP2 temporary directory #121

Open chia-mann opened 3 years ago

chia-mann commented 3 years ago

Hello. Good work. I have a little comment. It would be good to add more options for TEMP2, as it is for the final destination. I use a fast SAS disk for TEMP2. Because Plottman has been storing files on TEMP2 since phase 3, the final overwrite to the destination causes a situation where it is copied from the disk and at the same time another process writes to the same disk with TEMP2. This situation significantly slows down the whole copying process, because it is read and written to the disk at the same time. Thank you.

MrPig91 commented 3 years ago

I did want to add having multiple 2nd temp directories. But I wasn't exactly sure how the PSChiaPlotter would decide which temp drive to chose from if there were multiple 2nd temp directories and ensure that drive has enough space or is not overloaded with other plotting processes. To keep things simple I decided to only add one 2nd temp drive option. Perhaps having a checkbox on the final drives to act 2nd temp drives would be a good option if that is the end goal. I figured the end user could create multiple jobs if they needed to use multiple temp drives until I figured out a more elaborate solution.

chia-mann commented 3 years ago

Thanks for the reply. If I enter several final units, Plotmann places them on them precisely and alternates them. I've never noticed a problem with that. Why should it be a problem to work the same way with TEMP2? If I know that I always have 4 processes in phases 3 and 4, then I will prepare 5 disks with TEMP2 for them and there will never be a collision ... If I have the option to set only one TEMP2, there will be no acceleration. The resulting final copy takes the same as if I had not used TEMP2.

Jacek-ghub commented 3 years ago

@MrPig91,

Having the user create another job defeats the benefit of having PSCP. I truly think that simplifying end-user input (hiding most of those fields) is the way to go.

Also, assuming that one creates one job with 2 queues and 5 jobs per queue, does PSCP starts just two plotters with -n set to 5, or rather is starting a new plotter with -n 1? I think the latter option provides more flexibilities to actively manage those plotters.