Nathan Hurst
Nathan Hurst
Rounding is not a problem - we already round most numbers, as we use 0.1s of a mm which are not exactly representable. I don't think it is sensible to...
I have two systems run exactly as you propose, @eleaner, I'm using {5,3} with a separate CS per drive, > 15 drives. It all works fine, I'm not sure what...
My systems are 16 core somethings with hundreds of GB of ram... but the cpu usage is < 50% of 1 core under normal operation, ram is 3GB. This is...
comments: 1. lizard seems much easier to change than btrfs - every btrfs operation tends to go into a locked kernel state which is unpausable, unmonitorable and can fail catastrophically....
Thanks @eleaner I agree that @onlyjob is a rockstar :) They also made the point that if you want ec{n, 1} you perhaps should use xor, which is perhaps a...
Actually, there is a difference between xor goal and raid5 - raid5 will become corrupted silently because it doesn't have checksums, whereas lizard and btrfs will correctly identify the bad...
hmmm, yes, I think you are right that it is always safer using the maximum factor available for your drives. It would depend on whether lizard prioritises by missing chunk...
look on ebay for xeon server.
I use 12bay 2U supermicros, but I can't say whether they are especially good or bad. Been working solidly for 3 or 4 years.
hmmm, yes, I think you are right that it is always safer using the maximum factor available for your drives. It would depend on whether lizard prioritises by missing chunk...