Scaling Distributed Erlang
summary
Claim: replace the fully connected mesh with a Distributed Erlang overlay that garbage‑collects idle TCP links and implements global name registration via a Kademlia DHT so lookups take O(log n). Implementation uses consistent hashing, replication to configurable K nodes, iterative/recursive lookups and an extra liveness round‑trip instead of monitors. Current prototype handles hundreds of nodes (tests up to 350 nodes) and shows large speedups versus the old global while trading constant extra lookup work for far fewer connections.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:03Distributed Erlang
- 10:14Distributed Hash Table
- 22:40Iterative Look Up
- 24:38Global Api
- 42:12Nodes Function
related talks
39mContainer Orchestration and Software Defined Network: A Field Report
28mScalable and Multitenant Networking in XDP and Kubernetes Operators
26mTop new CNCF projects to look out for
33mBeyond VLLM: Distributed LLM Inferencing With Llm-d on Kubernetes
31mBreaking the Single-Network Barrier: A Cross-Domain Look at…
13mBuilding Reliable Infrastructure for LLM Workloads on K8s
44mDistributed Locking in Kubernetes
40mMulti-Cloud Global Content Distribution at Cloud Native Speeds