|
Title: Parameters for heavy duty full node Post by: rico666 on February 23, 2016, 03:34:16 PM Hi,
I've been running a full node for quite some time and am pretty high in the https://bitnodes.21.co/nodes/leaderboard/ - it's fun. However - in the metrics of the leaderboard computation, there is a NI = Nodes index NI = (p ∩ N) / N p = number of peers returned in addr responses N = number of reachable nodes which is constantly low with my node and it seems no one gets really high numbers there. What parameters would I need to feed to bitcoind, so that number gets higher? Also this brings me to the more generic question "What parameters could I use to run a really really heavy duty full node?" By heavy duty I mean a full node "serving" a lot of peers. The machine I run it on could probably pull a lot, so I would like to test the limits. Thanks for the hints. Rico Title: Re: Parameters for heavy duty full node Post by: unamis76 on February 23, 2016, 03:39:31 PM A "heavy duty" node is a node with lots of RAM and nice bandwidth... There's not much else to it :) A fast SSD and a nice processor also help.
Title: Re: Parameters for heavy duty full node Post by: rico666 on February 23, 2016, 03:52:10 PM A "heavy duty" node is a node with lots of RAM and nice bandwidth... There's not much else to it :) A fast SSD and a nice processor also help. Thanks, but obviously my node with its 8 CPUs, 64GB RAM, 2.2GB/s IO and dedicated 1Gps right on the ISP switch is nowhere near utilizing the hardware with the default parameters. That's why I am asking how to tweak these to make use of the hardware (and be useful to the network). Rico Title: Re: Parameters for heavy duty full node Post by: shorena on February 23, 2016, 04:08:51 PM A "heavy duty" node is a node with lots of RAM and nice bandwidth... There's not much else to it :) A fast SSD and a nice processor also help. Thanks, but obviously my node with its 8 CPUs, 64GB RAM, 2.2GB/s IO and dedicated 1Gps right on the ISP switch is nowhere near utilizing the hardware with the default parameters. That's why I am asking how to tweak these to make use of the hardware (and be useful to the network). Rico How long has it been online and how many connections does it currently have? Title: Re: Parameters for heavy duty full node Post by: rico666 on February 23, 2016, 04:48:50 PM # uptime
17:47:49 up 35 days, 19:42, 1 user, load average: 0.14, 0.21, 0.29 # bitcoin-cli getconnectioncount 97 Title: Re: Parameters for heavy duty full node Post by: shorena on February 23, 2016, 05:24:35 PM # uptime 17:47:49 up 35 days, 19:42, 1 user, load average: 0.14, 0.21, 0.29 # bitcoin-cli getconnectioncount 97 Sounds good to me as you are pretty close to the default max (125). From personal experience the NI value will mostly improve over time. The longer your node runs the more connections to other nodes than are constantly online. This should in turn improve their reachability, because they tend to have open ports. Title: Re: Parameters for heavy duty full node Post by: unamis76 on February 25, 2016, 02:04:37 PM A "heavy duty" node is a node with lots of RAM and nice bandwidth... There's not much else to it :) A fast SSD and a nice processor also help. Thanks, but obviously my node with its 8 CPUs, 64GB RAM, 2.2GB/s IO and dedicated 1Gps right on the ISP switch is nowhere near utilizing the hardware with the default parameters. That's why I am asking how to tweak these to make use of the hardware (and be useful to the network). Rico If that's so, then maybe we found an overkill node :) How much RAM is bitcoind using? |