CloudMicroHaskell simplifies distributed Haskell with direct-style graph serialisation

CloudMicroHaskell reimagines distributed programming in Haskell by moving closure handling from the compiler to the runtime. Built on MicroHaskell, which represents all code and data as a combinator graph, the runtime serialises the entire reachable graph when a process or message crosses a node boundary. This eliminates the need for explicit static closures or serialisation constraints, enabling a more direct programming style where developers write remote spawns that look like local calls. The library supports monitors, exit propagation, and higher-level abstractions like generic servers and supervisors.

The authors evaluate CloudMicroHaskell with multiple benchmarks: process and message throughput, a distributed work-pool (showing scalability to microcontrollers via a heterogeneous deployment), and a realistic file-synchronisation case study. The results confirm that runtime graph serialisation makes the programming model substantially more direct, but also introduces tradeoffs. Guarantees that Cloud Haskell enforces at the type level become runtime checks, and programmers must remain aware of Haskell's laziness and runtime-owned resources when moving graphs between nodes. Overall, the paper offers a practical exploration of a different design point for distributed functional programming.