IPFS IPFS þing 2022 - Content Routing 1: Performance, 9 Aug 2022

Previous Meeting

⏯

youtube image

►

From YouTube: Reframe Protocol: delegated routing - @ajnavarro - Content Routing 1: Performance

Description

----------------------------------------
Missed last year's IPFS þhing? Register to attend IPFS þhing 2023 in Brussels, Belgium here: https://2023.ipfs-thing.io/

A

I'm antonio I'm working right now improving on adding more functionality into google, also known as go ipfs, and I'm going to talk about refrain protocol and delegated routing.

A

So this is the agenda for my talk. I will try to make it quick, so we are going to talk about what's reframe, what is refrain for specifications and some of the reframe implementations that are already working, so we we did some previous work to try to.

A

Faced this problem using these two projects, but the main problem was that they were highly coupled with kubo, so you need basically a kubo notes to make them work because you needed the api and implemented a lot of small things to make it work.

A

So after that we thought about a new protocol. That's really simple: um reframe is a request response protocol, so server and client working and talking together is extensible. So you can add more. um You can like change or add more into specifications to use different transports and serialization protocols. Right now we are covering only http and json dax, a serialization and http for transport, but more to come, and you are welcome to add your own specific specs into our specs repository.

A

If you have some ideas or some use cases for different transports and serialization, and it can implement different method types so right now we have find providers, get ipns put a pns because we want to to try to to do some content routing using the frame.

A

But the idea is that you can use the frame for any other use case that you, you have implemented these methods.

A

So right now the main use case, as I said, is uh the content discovery, but it's an open protocol. So you can use this uh for any other use cases, as you can use the frame to point to a bitter and network addresses you can provide some refinable http downloads or you can create a private network with content providers using this this this protocol.

A

So right now we have covered on into specs these methods. They identify it's a way to say from the server to say to the client. Okay. I support this this and that so after that, this is like the main method that, um to put all um in the same page, the client, the server. Then we have the find providers that is able to find provided theirs by a cid, and then we have the get ipns to get some ipns records and they put ipns to put records into the into the server.

A

So we have a specification, so you can go there. You can read them. You can see what's wrong what you can improve, do any pull requests anything that you can implement this. Also, if you want- and we have several frame implementations, the main the first one is a go delegated routing. That is basically a library that is used in in our clients and servers.

A

So this is uh implementing, as I said, the client and the server part of the reframe protocol store the index. It's a as we say before in previous slides.

A

It's a server of the reframe and protocol, so any client can connect directly to it and get information from all the index that is like storing hydra booster is a client for the stored index, so hybridize the this dht node. That is able to improve the dsd performance on the ipfs network and it's connecting to the sturdy index servers using reframe, some guys, an experiment and basically it's a way to teleguide routing to other services using the refrain api. So, basically right now it's implementing dht and store the index under the hood.

A

So you can talk with the server using the reframe protocol and then it's querying all the all these routing systems. Under the hood and kubo we did in the last in the in the last release candidate. We added support from for reframe routing and you can try it now. If you want using this configuration to connect to the stored index servers- and here we have some demos.

A

So here we check the version of ibfs or kubo in this case, and then this is the standard configuration is using the dht for routing system.

A

So here we can see dht working for some testing cids.

A

Another example here.

A

And then we remove the dht routing system from the configuration and then we add the the reframe endpoint here and then we restart the kubo daemon and now this is the final configuration. So we are not using at all the dht right now we are only using the refrain endpoint.

A

So here we can see that we, we are getting some peers like this.

A

And everything is working, so you can test it right now, if you want going to the kubo repository- and here is another another example using some guy, uh that is, this refrain server, but instead of using the dht and the store the index under the hood, I just hardcoded some peers ids to see how simple it is and how easy you can like.

A

Oh, you can create your own provider. So here is the same example as before is basically using the dht. It's a kubo using the dht here I have my custom refrain node that is running locally.

A

Then here is the final configuration, so we are using both we are using dht and our local reframe server. So we use this example cid, and here is my random or hard-coded like peer id and it's like it's all mixes. All results are coming from the dhd and from our uh the frame server.

A

So I think that's all. Okay. Thank you.