Yesterday, October 30, a test or attack of some kind was run against the Chia network where tons of 1 mojo transactions (one trillionth of an XCH) were put into the mempool and the transaction volume greatly increased to where blocks where being filled as much as they can be. The Chia transaction spam looked a little bit like this:
The effect of this was interesting, with weaker nodes (those running on Raspberry Pis) getting out of sync with the network as they struggled to verify all the transactions being made. Interestingly this isn’t even the biggest transaction spike in the last week or so, with a very large spike happening on Friday, October 22nd with transactions in excess of 450 000 for the day.
As you can see the volume of transactions is really picking up, not just with spikes but the total average. You can really where the pools started up on this chart with the first real transaction spike and then its been largely increasing ever since. That being said, according to Gene in Keybase yesterday they were seeing transactions getting up to 25 TPS – transactions per second. That is almost 20% more than they have even claimed publicly, so whoever ran this test on the Chia network was most likely testing that claim. If so, the test was successful.
The other test that was run may have simply been a resiliency check. A few weeks ago the Proof of Stake currency Solana underwent a similar stress test but it took the network down for more than 24 hours. The Chia network did not even pause. In fact some farmers are indicating that the stress test increased their luck with extra block wins coming to them.
The one issue I have is that I don’t think Chia Network can maintain their claim that the full node can run on a Raspberry Pi 4. Being capable of running a full node means being capable of running it at max load, not simply at low load. And I think yesterday proved that as the network gets busier nodes will simply have to get faster. Or fewer people will need to run them in order to participate in the network, like FlexFarmer enables. Either way, I think there needs to be a serious discussion internally about what the actual minimum specification is for the Chia Blockchain software based on maximum load, not on normal load.
Edit: To be clear, there is a solution to this and that is transaction fees. Before you add a single mojo as a transaction fee take a look at this from Gene (Thank you to Triv in the Chia Plot Discord). You need to add 0.00005xch minimum by the looks of things.
Update 1: First update to the situation with a post from Chia Network
Update 2: Second update to the situation, with a promise to optimize the full node.
People with Raspberry pi 4 and 8gb of RAM are reporting they have no problem keeping synced. So the problem of keeping up synced with all this high transaction throughput seems to be only with those models with less RAM (2 and 4 gb ram).
That’s really good info. I will see if I can find data on that. Thanks!
Your sample group may not be representative, since my system ([email protected]) has %30+ stale rate during the storm.
There is no problem with the synchronization of some full nodes, but mining is affected and a large number of timeout contributions occur