Service Discovery Network Simulation

Logos
1

Hello!

Our implementation of service discovery has reach a higher level of maturity and it’s now time to look into what the DST team can do to help.

The first scenario would be a set of services Zipf distributed in popularity in a large network.

In this network we will try to find nodes supporting the least and most popular services. Measuring; median latency, closer peer count, time to first result, etc…

We could then complexify a bit by having nodes support multiple services. In this case, comparing the speed of search for node that support many vs few services might be interesting.

We will surely find more scenario but for now, I just wanted to get the ball rolling on this topic. :folded_hands:

1 Like
2

One quirk of the protocol is that when triggering `lookup`, using `startDiscovering` before hand should speed up the search because to search, a node need to maintain a table centered on the searched for services.

It would be interesting to answer questions like, what is the performance difference between a “primed“ search VS not doing the `startDiscovering` before or how long before a `lookup` is `startDiscovering` beneficial (if it is).

1 Like