Erynn

Storing all genserver pid in ets table in each node

if i have a thousands of nodes forming cluster where nodes are connected manually forming a ring(node1 to 2, 2 to 3, 3 to 4… last to 1node).

if all nodes have 1 genserver registered with their module. and if any genserver fail then they again restart and in handle_continue, they use Enum.each(Node.list, fn node → GenServer.cast({MODULE,node}, (:presence, {Node.self, self})) and when each node receive in hansle_info they store this data in ets table in each node. then they send their Node name with pid by using Keenel.send. by doing this each table contain all nodes genserver pid?

is it good idea and better in performance?

35 comments

#distributed

42 1104 35

2024-01-11 18:32:21 UTC

Most Liked

josevalim

Creator of Elixir

It is fine to connect 1000 nodes to each other. We benchmarked Phoenix with 1 million websocket connections and those are much heavier than just plain TCP connections.

IMO, focusing solely on the number of connections is the wrong way to approach the problem. The big question is which data will be sent over the connection and how often. So depending on your application and its communication patterns, getting up to 1000 nodes may be very easy or it may be very hard. For example: the trouble with large Erlang clusters is not really the number of connections but using modules like global, which requires all 1000 nodes to agree. I believe this is one of the reasons why the WhatsApp folks rewrote pg to not rely on global and it has much better scalability properties now.

Honestly, I personally find it hard to predict how a system with 1000 nodes will behave, without any detailed information or any real usage. Sometimes there are obvious bottlenecks, but most often they come from unexpected places that you only learn once you get to a certain scale. I’d prefer to first scale the system to 20 nodes and benchmark. Then 100 nodes and benchmark. The 500 nodes and then 1000. After all, you need real data, real use, and real traffic to make sure a system with 1000 nodes is behaving as expected.

Post #34

benwilson512

Author of Craft GraphQL APIs in Elixir with Absinthe

This is a restatement of your question, not an explanation for why a 1000 node cluster was the best way to solve your problem.

To be more direct: I think using 1000 erlang nodes is a bad idea for almost any system, and particularly when structured in a loop. I don’t think there are satisfying answers to the questions you are asking, because the questions you are asking involve a cluster architecture that is fundamentally bad.

If you can talk to us about the larger picture problems you are trying to face then perhaps we could provide some input. Otherwise I think this thread is at an impasse.

See: https://xyproblem.info/

Post #11

benwilson512

Author of Craft GraphQL APIs in Elixir with Absinthe

@Erynn I don’t understand how you expect to get quality answers without asking a quality question. WhatsApp ran a multi billion user chat app on < 50 servers so you’re gonna need to use multiple sentences and expand on your thinking here.

Post #9

sbuttgereit

And yet…

Post #17

Awlexus

It sounds to me like you are reimplementing process groups. Using those should be easier and possibly more performant, since not every process in the group would need to create its own ETS table

Post #2