search

πŸ”ŽAnalysis of the Bridge Node

This guide will analyze the load of servers on which Bridge Node Celestia and Consensus Node will be installed. In order to run a bridge node, you need to use public RPCs, or run Consensus Node yourself to receive RPCs. It’s worth mentioning right away that the team recommended using pruning = "nothing" for Consensus Node , but we decided to test various configurations

For our test, we took 4 VPS on Hetzner in the same location (Finland) - CPX41 8/16GB/240GB Accordingly, we will ru 4 Bridge nodes and 4 Consensus Nodes on 4 servers. Basic settings and installation are described in this article. The settings of each server will be slightly different and in our case have the following config:

circle-info

Server1:

  • pruning = "default"

  • indexer = "null"

Server2:

  • pruning = "custom" 1000/10

  • indexer = "null"

Server3:

  • pruning = "nothing"

  • indexer = "null"

Server4:

  • pruning = "default"

  • indexer = "kv"

We will measure the following parameters:

  • Uptime at https://tiascan.com/bridge-nodes

  • Used hard disk space with df -m /dev/sda1

  • Network traffic

  • Load average with htop

Measurements will be taken immediately after synchronization, 2 hours after synchronization, 24 hours after synchronization, and 48 hours after synchronization. All measurements will be reflected in this Google spreadsheetarrow-up-right and also in the screenshots

Main results

Network traffic results during Bridge run

Results of used hard disk space

Size comparison of .celestia-app and .celestia-bridge-blockspacerace-0 directories

hashtag
Conclusions

  • The network load on all 4 servers is approximately the same. And the difference that we see does not give us reason to pay attention to it

  • Uptime on all 4 nodes also does not vary much and we see that all 4 nodes with different configurations stay within 99.34% - 99.36%, which is quite normal for these VPS

  • Load average showed us the main load during the node synchronization period. But with a synchronized node, we see almost the same load on the servers. Only the server with Pruning=custom 1000/10 has a higher load compared to the other 3 servers

  • The main difference is when measuring the occupied space on hard drives. Server3 with Pruning=nothing showed great results in saving hard disk space, although before the test I thought it would be the other way around. So, the difference between server3 occupied space was more than 2 times compared to other servers on which Pruning was configured differently. So, from the screenshots above, we see that the .celestia-bridge-blockspacerace-0 directory on server 3 takes only 6+GB instead of 80+ on other servers

Summing up, we can conclude that during installation we do not have to pay attention to indexing and network consumption - and most importantly, set Pruning=nothing

PreviousState Sync [consensus node]NextCelestia-testnet

Last updated