Plotting Recap:
Intro – Bram Cohen, Gene Hoffman, Storage_jm, @kiwihaitch, @xorinox, @farming101
- 6:29 – primary bottlenecks for Chia plotting, CPU, DRAM, SSD temporary storage
- 9:00 – Bram – optimizing the chia plotting algorithms, sorting, bucket sort in memory, new sort functions, cryptographically generated
- 13:00 – different plotting approaches between the experts
- 16:00 – scaling up vs scaling out
- 18:43 – used vs new hardware for plotting, SSD evolution, endurance requirements, used NVMe enterprise
- 20:44 – SATA vs SAS vs NVMe SSDs for plotting
- 22:00 – ratio of CPU cores, to RAM to SSD temp space requirements
- 25:00 – plotting trace, SSD TRIM, temporary files vs buckets
- 27:00 – potential plotting optimizations
- 28:00 – how to effectively multithread the plotting process, staggering, and plotting in parallel
- 29:25 – plotting economics, reference plotting hardware, desktop vs server, used servers vs new
- 33:30 – plotting larger than k=32? Support for the future, io optimization
- 37:48 TB vs TiB
- Asking Bram questions! Why is the final plot file the way it is (a flat .plot file) vs a database?
- What is a good starting place for “Servers for dummies” for someone who doesn’t have a lot of experience building data center grade hardware?
- 55:03 – how do SSDs work? How
- 57:00 – SSD endurance
- 57:10 – what does bitfield plotting do? SSD vs HDD using bitfield (-e) in the plotting algorithm
- 1:05 – do you need to delete plots after winning a block?
- 1:09 – farming101 plotting setup configuration
