EthCC 2: Vanessa Bridge

okay so there's an increased number of empty blocks being mined in the network and so we were trying to identify what was the reason if it's not spy mining that's what's happening and so one of these mining pools are got accused of actually doing spy mining they were saying well no we're not doing this essentially this is a protocol flaw because we know once the block the first block has been sent out before we actually receive the entire information we have to start mining the next block and so since we don't have all the information we're mining empty blocks and that's what they said they were doing and there is some truth to that there is some I guess design flaw it makes it that you have to mine an empty blocks at point in time in the network and so this is kind of what it will look like you're trying to create this block to mine and then you set wait this was yeah and so you're just waiting before you receive the information for the next block and so you just decide to mine an empty block so this is what they're saying that's happening and essentially it it's been seen in the data you actually have more empty blocks but so now we were looking at the uncle rates and so this is some of the data that we looked at I know this table can be a little bit daunting so just bear with me for a little essentially each column represents the data for each of the variables that you see here so starting with the number of transactions to the uncle rate until the gas usage and when you start looking at the rows it's actually the correlation between the variables so for instance the first one 0.76 is the correlation between number of transactions and uncle rates which makes sense the bigger your block the higher your uncle rate so they're positively correlated and so that's how we started looking at this data and looking at what was essentially affecting the uncle rates so yorkshire you also see that hash rate is 0.76 but there's a high positive correlation with that there's a high positive correlation with the block time and with the difficulty with some fees with the uncle mining reward but we separated the data in terms of different time points we're trying to get as much as possible and bias data so when you have the difficulty bomb happening for instance we can't really look at this data because we already know that it's something unusual that's happening so we decided to take it for the month of December of last year and then what we notice is that the correlation that was really high between number of transactions and uncle rates it seems to be non-existent at this point in time and so we noticed that the uncle rates is not really correlated to much other than the mining reward for uncle's which makes sense and the block time but other than that there's like very few things that seem to be affecting the uncle rate and so it really makes you wonder what then what is the relationship within uncle rates and the network held and also uncle rates and the block size so essentially if there's no correlation between uncle rates and the number of transactions it means we can for instance increase the size of the blocks and start mining bigger blocks because it's not going to necessarily impact the network or anything like that and so from this information from all the evaluation that we looked at what we found was that if you can include your uncle's in the first block you're essentially starting to lose revenue so if you're trying to it up you were trying to do something like uncle mining you would really need to be very well connected between the miners and so it's kind of contradicting what we said before that these uncle's were set to incentivize the centralization because now it turns out that you need to be really well connected and who's the best connected in the network is the miners and so it's actually incentivizing centralization at this point also we discovered that mining that's what allows to have this stable time of around 14 15 seconds for each block and there's a number of strategies that can be adopted by miners that will kind of reflect what's happening in the network and you can trace it through the uncle rates and see what's happening and that's why we were able to determine whether or not they're selfish mining or whether or not any other mining strategy that I talked about is taking place and so I would say that when you're designing the systems it's very important that you make sure that the incentives are both users and miners are aligned because for instance another thing that we saw that was happening is that you have not only empty blocks but blocks that are not that big or not that full because it also correlates with the price of aetherium so at points in times like this when the price is low there's an incentive for the miners to not necessarily process transactions because you're not gonna gain much more money from the transaction fee it's probably more beneficial for you to just mine a couple of transactions that have higher gas that it's going to be beneficial for you and not include transactions that you know you don't see as much rewarding and so you're not aligning the incentives of users who are waiting to get their transactions added and miners who want to make money and so it's important that when you're trying to design these systems you make sure that you find a way to align them because otherwise you have funny behavior like that and then it's just not good for the network there's so many transactions I need to be processed and then you have empty blocks blocks that are half filled and that's not good so that's it questions I thank you very presentation a great thing and at the same time I would like to ask about the actual correlations that you are looking at because most of them look at the properties of the blocks themselves like the number of transaction size of the blocks and so on and I'm not sure if I've seen any that are talking about the network like they the latencies on the network number of the nodes because we're talking about the falling number of sorry phone they're following uncle rates and we've also noticed very big fall in the number of nodes and the distinct minors because of the following rewards and now when we say that they the actual falling uncle rate is good for the network and it says the network is getting better it also can suggest that the decentralisation is actually disappearing because for the uncle ray to grow we have to have two different pointers to mine at the same time so it's a possibility to do did you look at this as well so we looked at some of these variables but the assumption is that yes miners are better connected because there's a migration to AWS and systems like this also there's a lot less notes the distinct miners however are constant we look at that it was like 400 distinct miners went etherion started and then it crashed down I guess everybody realized let's just have a pool and then you only have around like 60 mm distinct miners and it's like more or less constant in the last year two years so can you repeat like can you go back few slides yet this one so did you also saw no correlation no strong correlation between the number of transactions sorry between this the block size and the ankle right yeah and do you know can you tell like let's say that I'm I'm increasing the block size by 10 kilobytes how much this would affect the ankle right so we did look at that data and then the reason is that the correlation is lost not because there's necessarily this causation between the two of them but because as I mentioned miners are not filling the blocks to begin with so you have lesser fill blocks and so I guess the ankle rate will depend more on the connectivity of the two they are proposing these blocks and so although yours is smaller it might not get propagated as fast as the other one and so you have this kind of loss of correlation so you don't see here in there in the results so much the network latency because of the size of the block you don't see it directly but because let's see I think there's a slide where here so here we're looking at the number of transactions and the number of uncles and then you can see for instance the in the last two months you have more or less the same number of transactions but then the Ankur rate is significantly lower and so it's a number of factors it's the mining strategies that miners are adopting where they don't include as many transactions but also the connectivity [Applause]