3 ms·
How slow can these disks be? Could they be tape-drive slow? Or "1 million blurays in a stack" slow?
by londons_explore 6y ago
How slow can these disks be? Could they be tape-drive slow? Or "1 million blurays in a stack" slow?
- zozbot234 6y agoFrom looking at the Chia wiki, the farming process seems to be very slow. Slow enough that a tape- or optical-media robot library would easily keep up with it.
- londons_explore 6y agoMy reading of the wiki suggests that the time for your disk to read the data really does matter. Chia blocks are every ~20 seconds. Within those 20 seconds, you need to read 7 + 64 different areas of the disk. Some read addresses depend on the last, so in reality you need to do 7 sequential reads, all within the block time to get any reward. So basically, if your storage medium takes over ~1 second to read data, you're going to be throwing away some rewards, and if it takes over ~10 seconds per read, you will barely get any rewards at all. Tape drives are out then, and so is optical media unless it's already in a drive.
- zozbot234 6y agoDo the 7 initial reads access separate data from the remaining 64, or is the whole "plot" fair game for even those initial reads? ETA: from looking at the docs, it seems that the 7 initial reads do access separate "tables". It should be possible to optimize by storing these on lower-latency storage, and even optimize further via parallel farming of many "plots", of which only a few at any given time will ever need to retrieve the 64 blocks denoting a successful proof. ISTM that tape media has real potential here.
- londons_explore 6y agoThe initial 7 reads simply save you from doing the last 64 incase there is no chance of winning. Put those 7 on fast media and you'll very quickly know "I might have won". But without the results from the rest of the 64 reads in a couple of seconds, you won't actually win.