Missed Signage Points, Why They Matter, How They Cost You XCH

Yesterday a farmer reported VDF’s taking more than 25 seconds to process.

Today another reported missed signage points.

To understand why this is happening and what the consequences are we have to go a bit into the details. The current Chia node is a mix of Python/C++ and is locked to 30%. That means blocks found are only filled up to 30%. This keeps the amount of processing down. The network has come close to freezing before when blocks got to around 25% full. This is now happening again largely due to HPool’s new payout policy of paying everything over 0.001. This results in a lot of transactions on the network that basically spam it. This results in a few problems.

The most obvious one will hit those using worse hardware. Farmers running nodes on hardware that is not high-spec will begin having issues resulting in missed signage points and reduced income. It could even lead to the node itself crashing. Continued 30% full blocks may lead to most farmers even those running nodes on decent hardware suffering crashes. You will notice many NFT pools with bad luck on average because while their farmers may have a lot of space, they are losing a couple % of possible blocks due to missed signage points.

Another thing that may happen lies with pools claiming blocks. The reference pool protocol (It was meant as a reference and its insane that so many pools are using it in production) is slow and issues like this may slow it further. This will result in them taking several blocks to claim, I think a lot of pools are doing it in 20-30 blocks if not more. As most pools have a height of 60-70 to leave for self pooling, if the pool is slow farmers can switch to solo and claim a block they found before the pool can. This may be what happened to Maxiopool. Pools that are slow to claim blocks can have them stolen and I’m sure more farmers will catch on to this. On most Reference NFT pools they only lose a bit by switching to solo right when a block is found by them and in return they have a decent chance of keeping 100%.

Almost all farmers are affected by missed signage points. We see perceptible increases in profitability between those running their own nodes and those connecting to our hosted node. As HPool hosts its own node, this is likely one reason they don’t mind spamming the network as doing so only hurts their competitors. Hosted nodes are obviously running on great hardware and likely have several optimizations over the base one.

To reduce your losses its suggested to run the node on Linux using a good processor.

There are teams building a better node that can handle 100%. Once its released (likely Sept/Oct) this problem should be drastically reduced. Probably the only major problem there is is that a lot of blocks at 100% will likely cause all old nodes to crash which will force everyone to switchover in a hurry. Please note that I am by no means an expert and the above is only my opinion.

1 Like

RIP Raspberry Pi farmers?

Pretty much if your planning to run a full OG node on them.

Do you guys have any evidence that the Chia node slows down at 25% transaction volume? I’ve noticed it get a little busier but if people are slowing down the network isnt it more likely to be network volume than CPU load? From what I’ve observed the network load is pretty staggering when transactions start getting high.

1 Like

Alex will have reports but I don’t have them handy myself

Because being locked to 30% is more likely to slow down transactions than slow down nodes, but my node is doing a consistent 110GB of traffic a week right now, each way. Its noisy.