Faasm

Faasm is a high-performance stateful serverless runtime.

Faasm provides multi-tenant isolation, yet allows functions to share regions of memory. These shared memory regions give low-latency concurrent access to data, and are synchronised globally to support large-scale parallelism across multiple hosts.

Faasm combines software fault isolation from WebAssembly with standard Linux tooling, to provide security and resource isolation at low cost. Faasm runs functions side-by-side as threads of a single runtime process, with low overheads and fast boot times.

Faasm defines a custom host interface that extends WASI to include function inputs and outputs, chaining functions, managing state, accessing the distributed filesystem, dynamic linking, pthreads, OpenMP and MPI.

In our paper from USENIX ATC 20 we present the initial Faasm design and evaluate it in the context of stateful serverless functions.

In our paper from USENIX NSDI'25, GRANNY, we present the design of the OpenMP and MPI runtimes, as well as new dynamic resource management policies, all implemented in the faabric library.

Please see the full documentation for more details on the code and architecture.

Quick start

Update submodules and activate the virtual environment:

git submodule update --init --recursive
source ./bin/workon.sh

Start a Faasm cluster locally using docker compose:

faasmctl deploy.compose

To compile, upload and invoke a C++ function using this local cluster you can use the faasm/cpp container:

faasmctl cli.cpp

# Compile the demo function
inv func demo hello

# Upload the demo "hello" function
inv func.upload demo hello

# Invoke the function
inv func.invoke demo hello

For more information on next steps you can look at the getting started docs

Acknowledgements

This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreements No 825184 (CloudButton), No 101086248 (CloudStars), and No 101092646 (CloudSKin), the UK Engineering and Physical Sciences Research Council (EPSRC) award 1973141, and a gift from Intel Corporation under the TFaaS project.