Protocol Berg v2: Ekaterina Riazantseva - Standardizing Ethereum metrics PeerDAS and FOCIL retrosp.

Yeah, thank you. So, my name is Kaja. I'm Ethereum protocol contributor. Um, and um I I made an initiative to standardize metrics in Ethereum protocol. And today I would like to talk about this initiative and how it can help uh to build Ethereum to test Ethereum.

So um this talk is mostly for DevOps and core developers and but maybe if you're a staking operator and you run many uh nodes many different clients at a time you probably would be also interested. So quick agenda for this talk. Uh I would share my EPF journey. Um how I built metrics specs uh standardization process itself. Uh what challenges it had uh what's going to be next and uh a quick overview of this talk.

So I joined Ethereum protocol fellowship last year. It was cohort five. Um for those of you who are not familiar what is EPF? It's the Ethereum Foundation program for those who want to contribute into the protocol mostly developers and researchers and this is where you can start contributing um and at that time pure was it's in early stage and uh I was interested how can I help in this process and um I I I had a conversation with Eastbandop's team and I found out that uh that would be great to have the same metrics for all the clients because it was early testing phase. Uh there was an initiative in 2020 but um it's still like it wasn't adopted uh because it takes a lot of resources to do this.

So yeah I started building u metric specs uh beacon metric specs specifically. So for consensus clients um I had initial uh list of metrics uh which was taken from the uh first implementation of pureas. Um then I started to build them all across the consensus uh clients. I also built graphana dashboards for curtosis. Corttosis is the tool for running uh blockchains locally and uh cordvs u mostly use it for testing Ethereum and I also built the dashboard for ETH bundles.

This is an example how the dashboard can look like with the standardized metrics. So we can see all the clients on one dashboard and uh this is super helpful because uh before this dashboard um DevOps just need to monitor many dashboards with different clients and uh compare them or make different queries uh but they are not sure if this metric is the same on different across the different clients. So uh and here you have just one dashboard and you can easily apply filters and um just to explore to monitor the clients and to explore the issues. Um yeah just discover any u problems they may have. Uh and this is super helpful during the test nets because currently um we are testing pureas which is foraka next upgrade um the major like the flagship EIP for Faka which is uh scaling Ethereum.

So uh I also started to standardize fossil which is another EIP probably for Glamsterdam not sure yet. Uh but um from these two uh EIPs pure and fossil what I brought here is like I tried to make some steps how standardization can look like. So the first step is when EIP is in progress. uh this is the first this is the good time to start standardizing. So uh we can make an initial list of metrics uh what is needed uh what would like to monitor um and uh during the breakout rooms we can discuss the initial list.

So the next step is uh prototyping. So normally after EIP is more or less ready um the builders uh build the first prototype and here we can already apply the initial list of metrics uh and and test it in curtosis. So cortis has graphana dashboard. So we can already see the result of the metrics then sorry uh other clients join and uh we can have like a small interop between clients. Um as you know Ethereum is not one client it's many clients client diversity.

So um uh we can also already if if it's the devet phase we can already see all the clients with the same metrics on one dashboard. After that definitely there will be corrections because specs evolve and metrics also will evolve. Uh so we we will have discussions on uh do we need uh any new metrics or um should we correct uh the current ones and the the fast uh the late the last step is production. So all these metrics standardized metrics will go to production. Um and this is where for example if you're staking operator so you you will see uh the same metrics for clients on on one dashboard.

the challenges I have. So it's hard to plan uh the list of metrics ahead because definitely you need at least one implementation just to see the metrics on the dashboard because it's not really clear um so uh what what buckets you need or what what metric what metric types you need. So you can definitely correct them after you have first implementation. So you just can't give the list uh the pre-built list of metrics to the builders and say yes please build this uh you definitely need to correct them. The client diversity is also a challenge uh because uh clients are written in different languages and though metrics it for metrics it doesn't matter but anyway uh clients have their own style on um how to build metrics u the granularity of metrics uh the naming etc.

So and even if one person tries to build the metrics all across the clients, it takes time and a lot of efforts to to do this. Uh naming in units is also a challenge because as I've previously said clients has their own style what units they use, what namings they use. So uh to standardize metrics which are already in production is definitely difficult. So um this is why I've started from the EIP which is not yet in production. So what's next uh next uh so currently we are standardizing metrics for Fusaka and the latest ones are get blobs v2 and custody group count metrics um almost done and also uh the metrics for execution client.

So execution client joined for saka I mean pureas specifically a little bit later so we're still in the process of standardizing it. The next goes glamsterdam which is currently in discussion. So which EIPs join Glamsterdam? Uh and there will be metrics for execution and consensus clients and the next age upgrade we definitely need to standardize the P2P and gossip submetrics because the uh bandwidth is a bottleneck for for Ethereum and uh it's definitely important. So a quick overview uh why we need standardized metrics.

We definitely need them for testing. So uh when you test a lot of different clients, you definitely want to see them one at a time and one one dashboard and that makes uh uh the proc the testing process much much easier uh because you can easily discover any issues. Second uh monitoring uh if you want to monitor many clients uh after the metrics went to in into production. So it will be easier to do. Uh research teams also may need uh standardized metrics uh because they working on analysis and they build benchmarks.

Uh so and uh of course um uh staking operators may need them because they run different uh different consensus clients and execution as well. So uh that's why they might want to have the same metrics and one dashboard for all the clients just for for devops. Thank you. Um yeah this is my experience and uh if if you have some cases so why you need standardized metrics on Ethereum specifically please let me know after this talk maybe yeah you are interested u yeah thank [Applause]