You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After adding the NUPD column I often think about using it to rank the channels in order of their actual usefulness. NUPD represents the amount of attempts, not just successful forwards and it actually consumes disk space every time it increments. Though it's less of an issue in 0.15.0 after some great optimizations.
The main issue here is all the rebalancing or probing attempts, you can easily see them in the log as the amounts are very round (50/100/400k etc.) and 99.999% (use any amount of 9s lol) of them fail. So all they do is wasting your disk space. I don't think one should raise fees to prevent these from even going through the node, after all some of them succeed and earn you fees and you can't know beforehand which those are. But it's an interesting metric nevertheless as it represents how busy that particular channel is.
So I'm thinking about a formula that can represent the useful traffic share from the forwarded amounts and update count. The simplest would be (sent+received)/nupd in sats/update but it could be misleading compared to busier channels if nupd is low. The idea is to quickly locate inefficient channels that should be closed. If you only have a few dozens of channels it can be done manually just by sorting by NUPD and looking at the bottom but if there are many channels it becomes complicated. There are other tools like faraday (by Lightning Labs) but they don't have an UI and just spew a big JSON.
Another metric to consider might be the channel lifetime in blocks. So the result could be measured in sats/update/block or something like that. Maybe make two metrics, sats/block and sats/update. Some channels might have a lot of updates but only because they were opened long ago.
Ideas are welcome!
The text was updated successfully, but these errors were encountered:
After adding the
NUPD
column I often think about using it to rank the channels in order of their actual usefulness. NUPD represents the amount of attempts, not just successful forwards and it actually consumes disk space every time it increments. Though it's less of an issue in 0.15.0 after some great optimizations.The main issue here is all the rebalancing or probing attempts, you can easily see them in the log as the amounts are very round (50/100/400k etc.) and 99.999% (use any amount of 9s lol) of them fail. So all they do is wasting your disk space. I don't think one should raise fees to prevent these from even going through the node, after all some of them succeed and earn you fees and you can't know beforehand which those are. But it's an interesting metric nevertheless as it represents how busy that particular channel is.
So I'm thinking about a formula that can represent the useful traffic share from the forwarded amounts and update count. The simplest would be
(sent+received)/nupd
insats/update
but it could be misleading compared to busier channels ifnupd
is low. The idea is to quickly locate inefficient channels that should be closed. If you only have a few dozens of channels it can be done manually just by sorting byNUPD
and looking at the bottom but if there are many channels it becomes complicated. There are other tools like faraday (by Lightning Labs) but they don't have an UI and just spew a big JSON.Another metric to consider might be the channel lifetime in blocks. So the result could be measured in
sats/update/block
or something like that. Maybe make two metrics,sats/block
andsats/update
. Some channels might have a lot of updates but only because they were opened long ago.Ideas are welcome!
The text was updated successfully, but these errors were encountered: