Ethereum Protocol overhead looks excessive


I'we made a lab:
2 VirtualBox VMs, fresh, connected via internal network, disconnected from the Internet.
Installed Wireshark and Ethereum POC3 in both VMs and connected to each other as peers.
Now what I see in Wireshark is that the Ethernet+IP+TCP overhead is very big compared to the payload.
The actual data (payload) is 10 bytes x3 per second. (10 bytes of data compared to 64 bytes per frame with overhead)
(it does not matter if I run mining or not... just have connected peers)

1. Why do we need to generate 3 packets per second between peers, if the block generation rate is 60 sec?
2. Why not consolidate those small packets with huge overhead into something bigger?

Ping me if you need Wireshark traffic capture.
It could cause issues with slow and high-delay links (satellite) ?



  • jTAjTA Member Posts: 50
    Curious... What's your full lab set-up?
  • TechnologovTechnologov Member Posts: 102 ✭✭
    In this lab - only 2 VMs
  • ThreeIfByAirThreeIfByAir Member Posts: 4
    At the risk of reactivating an old thread... I ran some Wireshark traces (I may even go implement a dissector assuming nobody else has done so yet). The packets you're seeing are ping-pong packets used to verify that peers are still present and operating. As such they don't need much TCP content. In AlethZero you can see the result of those packets, because you see the latency to your peers.

    As a noob to this stuff, I don't know quite why the protocol pings so often (every 5-10 seconds would seem perfectly adequate) but there's nothing wrong with the existence of the packets themselves.
  • TechnologovTechnologov Member Posts: 102 ✭✭
    That's good that you re-activated the topic, as I also wonder about "pings so often".
Sign In or Register to comment.