From Nanoseconds to Decades: The Timescales of Ethereum by Jannik Luhn | Devcon SEA

[Music] being here uh my name is Yanik and yeah so one way to think about ethereum is this this very big machine that consists of uh different gears uh that connect to each other that drive each other uh and that function together to do what ethereum does and these gears have different sizes some of them are very small and therefore they spin very fast and some are very big and therefore they spin spin slowly but therefore they um they're all the more powerful and in this talk for actually no particular reason other than I think it's it could be fun and interesting exercise I want to go through a bunch of these gears these processes that make ethereum run and have a look at how fast they are very quickly starting with the uh slot time 12 seconds um every 12 seconds we get a new slot every 12 seconds we get a new Block in ethereum kind of the defining time of of ethereum I would say and now from there we can go lower um these blocks need to be prop propagated through the network that takes about two seconds another thing we propagate through the network is transactions they're smaller they're easier to verify um so they propagate much faster takes about 250 milliseconds to do that now we reach a boundary 100 milliseconds is about the network ping time two nodes talking to each other um they will always have a latency of this order of magnitude so if we go lower than that then we're in this purely local do domain of um um yeah that only happens on uh on a single node everything above that can be Network wide 100 milliseconds is also the time roughly takes to execute a block of course some blocks are bigger they take longer some are shorter um some are smaller they are quicker to execute two orders of magnitude smaller attestation validation takes about 1 Mill MC attestations are these votes that valid dat sent to agree on what the state of the chain is and now we're getting into a range of numbers that are very difficult to imagine for humans one millisecond is not a very human time scale so I would like to compare this to length which might be easy to think about if we uh if we go back to a slot and imagine that this is the size of a human then at the station validated validation would be about the size or the thickness of a human hair so quite small already but we will go even smaller 100 micros um roughly time of an average transaction to be executed big part of that is the signature check of course uh we want to make sure that transactions cannot be forged and that's why we have these cryptographic signatures and they take a long time compared to a transaction execution to check um but the most probably the most important cryptographic function in ethereum in blockchains as hashes we hash all the time whenever we check a signature whenever we change State whenever we build a block we do a hash and therefore these hashes have to be very fast and they are 350 nond and lastly kind of the smallest we can go a single unit of gas gas is the unit of measure we uh we use to um measure um um resources uh and one of these resources is computational time CPU time and one unit of gas is corresponds to 10 NCS um of CPU time um 10 NCS again very small to if we go to um length scales that's about the size of a glucose molecule the most important energy carrier in biological systems such as humans we can't go lower than this but now we can go higher um starting again nine orders of magnitude above um this 10 NC with slot time 32 slots in a row 6.4 minutes an Epoch the time during which each validator gets to submit their vote on on the state of the chain two EO in a row 12.8 minutes finally period and that's um in this presentation at least the longest um protocol um defined time scale above that uh we're talking about humans we talking about human coordination human humans are um are um slower so they need more time uh to uh to coordinate the fastest that I could think about it was Emer emergency response times say etherum breaks for some reason how quickly do we get developers um online to their machines to look into what's going on and 30 minutes um is kind of the number I came up with based on a lot little bit of data of course there not a lot fortunately because nothing major has happened so far uh two days about the time it takes to sing a note hopefully we can get this down um at some point maybe using light c as we saw before two weeks um the time between uh the all cord Des meetings the meetings where um the core developers talk about the state of the chain and um the uh changes they want to make to it uh to improve it one year the time between hard Fox when we actually Implement those changes in in these intervals two years the time to include an EIP so the time between an Pro proposal to implement the chain uh is uh proposed and uh until it's included in the hard for is about two years uh in the past two years also the time between deathcon so may read again in uh 2026 and lastly ethereum is 10 years old now and the most long-term features of ethereum like proof of stake and charting have been discussed since the very beginning uh and these long-term features can take 10 years to implement and now I originally I wanted to stop there but for for fun I just wanted to add another number maybe 100 years maybe that could be the lifetime of ethereum if we think about if we compare to other um protocols like email or social systems then 100 years is kind of the time scale we have to think about here and if we look at these numbers we should uh be very careful what what we actually Implement here because it will hopefully affect a lot of generations Afters to come so yeah maybe that's the only uh outcome of this talk as well to maybe make you think a little bit about these longer time time scales thank you thank you Yanik um we have questions for you so guys do we have question for Yanik oh okay cool uh you mentioned that the 10 nond for unit of gas and computation also the 12 seconds for a block I don't know if this is a beginner question or question but I ask um when in the future the CPU power uh becomes more powerful will do you think will this uh 12 seconds change to 6 second um yes or no I think the 12 seconds are um kind of constrained by um um Hardware of of um computers but not necessarily by CPU time I think there's other resources like Network latency that are more important here but of course if Hardware improves and physical capabilities of computers improve then these 12 seconds can go lower and also the amount of gas we can spend in a single block can be improved right now the target is I think 15 million gas we can if CP F go get more powerful we can maybe increase this but also there's other resources that are measured in gas and they also have to be taken into account yeah uh thanks for the talk it was really cute um you you mentioned lots of different metrics um I wonder which one is the most important to be optimized oh interesting um uh I think not all of these are metrics that should be optimized I think sometimes are fine if they take longer uh I think some time scales probably can't be optimized a lot like um for example I think um the all C meetings it would be very annoying if if people would have to go to a meeting every day or something like this so um but in general I think the two years Improvement proposal that sounds like a a very long time um um on the other hand if we compare it to the 100 Years of lifetime of ethereum then maybe that's not too big uh too bad so don't not really have don't have a clear answer to that uh I guess the easiest answer would be like the technical time scales like in uh reducing um uh block propagation times or things like this but um I think none of these are very crucial to be honest all right thank you so much Yan for this amazing