To talk about TPS is a very 2017 topic, I know. In the past I had the same question and I got to understand that there is no fixed limit of TPS in the Lightning Network as it depends on the amount of channels.
Each channel increments the total TPS capabilities in the network.
I was wondering if this what I said is 100% right and also which is the technical limit of TPS per channel? Maybe something related with the amount of HTLCs than can be processed by the LN implementation?
There can be a relatively easy calculation based on the number of transaction bytes requirde to open and close channels, the number of channels that thereby can be opened, given some given an average HTLC duration, it's in the LN whitepaper, in fact, and I think they essentially say that for LN to replace all known payment rails in the world, Bitcoin blocks have to be 138mb in size. But again, if the channels are longer lived, their imprint on the chain is reduced proportionally.
As I understand it, several featurse under discussion to be added can drastically reduce the data load. Covenants is one I hear many times.
Bitcoin blocks at 138mb would be practical in about 15 years time if storage capacity and latency continues to improve at even 50% of current rates.
The block size issue will always be about whether you can defer upgrading your monitor and instead get a bigger hard drive to run your node. To use rough comparisons. Bitcoiners will not, and should not allow an increase in block size until, at minimum, LN is filling half the blocks on average. That's not gonna happen for another few years yet.
reply
It has been tested to be around 30 - 100 TPS across implementations on general hardware. https://bottlepay.com/blog/bitcoin-lightning-node-performance/
reply
would be cool to see this ran again, a lot of changes have happened to all the implementations since then
reply
This article really helped me. Thank you Tony.
reply
It is using old implementations though. Pretty sure when I tested it, LND got slower over time to current versions. CLNs currently the fastest one I know of, I can reach 150-180tpspc. There are still optimisations to be done tho
reply
Renè Pickhardt had in the past intensive studies about that aspect and many more. Here is one of his short videos explaining the TPS of LN https://www.youtube.com/watch?v=txcjxSRDvqQ
reply
Max number of open HTLCs is 483. You can probably push more payments per second through a channel between direct peers, but when you consider multihop payments and the need for liquidity and all of that, i think ~500 is a reasonable bound for doing napkin math
reply
Thank you for the answer. The results that @TonyGiorgio posted have 10 times less tps and seems it was tested between two direct peers. I supose that in routed payments it has to be even less, no?
reply
In a follow up post, Joost theorized how to get up to 1000 TPS: https://bottlepay.com/blog/bitcoin-lightning-benchmarking-performance/
reply
ill check it out, thanks!
reply
yep. I'd say the ~500 number is an upper bound. In practice, it'll be an order of magnitude lower because of liquidity requirements, network latency (especially over tor), intermittent failures because of all the crappy hardware in the network, etc.
reply
reply
This tweet was deleted.
reply
Sorry. There should be measurement of TPS between direct channels.
reply