I’ve just seen that LTO9 has just been released giving 48Tb on a single tape and they’ve tested upto 500 odd Tb. I’m going to assume a tape drive is horribly expensive making it non - viable but it did make me wonder if this technology could be used to store plots.
I’m going to guess not as it’s a sequential read device and each plot is verified within a certain time or something.
Tape would be way too slow. The seek time would make them worthless in a farm.
Wont work man, tape read and cycle times are too slow. Bram explicitly talked about tape pre launch even. As an owner of a stupid large tape library, I am crying in LTO with ya!
I thought as much. I used to have a tape backup when DAT tape was the latest thing. Cold or archival storage right?
I’m also old enough to have used Mag tape for backup. Hard drives as big as a wardrobe that houses 300Mb and an 80Mb system disk that sounded like a plane taking off when starting up.
Back then an 88Mb Syquest drive was amazing.
Funny to think you would need over 1200 of them to hold a single 32k Chia plot.
Anyhoo, I digress. Just a thought. I bet someone will work out how to index the plots at some point.
Have a good evening.
Wx
Even if that were possible, if you ever got a hit, then the tape would actually need to find that file and read the plot to get a win… again…waaaay to long for 30 sec time window.
AHH I see. So it’s a lottery with the plot ID and then the system checks the validity of the actual plot with some sort of checksum.
When I get 5, which us never, I must have a read up on this for pure interest sake. Ithe system is on the chia wiki imm guessing.
I might have read the same article and thought the same thing.
Honestly, I was just looking for an excuse to justify buying tape drives just to play with them.
That being said, wouldn’t it have been awesome if it were actually viable? What a shame.
Well mine is a 96 tape lto5 based library, a clone of the HP 8096 variant (MSL makes them all or something like that) and I store photos, plate solving (astrophotograhy stuff) and of course, archives and backups on it. It is not realtime in the slightest however there are stubout LTFS software/hardware combos that can do some things pretty fast as a tape is being loaded, but not that fast still. way over 30 seconds to access a specific spot on the reel from random.
A while back, I was doing some back-of-the-envelope calculations if it would be possible to design a tape drive that could be fast enough for chia farming. I think there is room for innovation and it would in principle be feasible. Maybe a startup will attempt it.
However, where tape is already useful today is to temporarily offload plots to tape, use the HDD storage for something else, and then relatively quickly go back to farming without having to replot all those plots.