-
I've seen this benchmark being referenced in one of replies https://github.com/bottlepay/lightning-benchmark and that your team had not tried running eclair on raspberry pi. Could you please share the minimum and recommended hardware specifications required to operate Eclair effectively? Additionally, if available, I would greatly appreciate any performance metrics related to the number of channels opened, payments processed, and forwarded transactions. Such information would be invaluable for estimating the costs associated with deploying Eclair as a routing or LSP node. Thanks |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 3 replies
-
There is no good answer to this: hardware requirements directly depend on the number of active channels you expect to have and number of payments you expect to relay, as resource usage scales based on those metrics. You should probably ask other node operators what they use and the size of their node to get a rough idea of what to use (e.g. on the t.me/eclair_lightning telegram group). Note that eclair is mostly RAM-intensive, and you should probably tweak the default JVM configuration as detailed here. You should also carefully read the configuration file and tweak what makes sense for your node.
Such performance metrics don't make sense at all today: the payment volume is low enough that there are simply no performance issues related to payments. The only thing that's important is to scale your machine based on the number of active channels you expect to have. |
Beta Was this translation helpful? Give feedback.
-
I run my eclair on this setup: Application Server: Primary DB Server: Secondary DB Server: I have only 50-60 channels, but at my peak I did 8 payments per second over several days. (rebalancing) Probably not your use case. But I think it nicely showcases, what kind of throughput that kind of setup can bring. |
Beta Was this translation helpful? Give feedback.
With this hardware you should be able to comfortably handle 25,000 channels. Setting
-Xmx 12G
should be a good starting point. For better security, and to handle a large number of concurrent connections, consider the cluster mode.