Serverledge is a Function-as-a-Service (FaaS) framework designed to work in Edge-Cloud environments.

Serverledge allows user to define and execute functions across distributed nodes located in Edge locations or in Cloud data centers. Each Serverledge node serves function invocation requests using a local container pool, so that functions are executed within containers for isolation purposes. When Edge nodes are overloaded, they can to offload computation to neighbor Edge nodes or to the Cloud.

Note that Serverledge is a research prototype and it is not (yet) meant to be used for production use, as it may lack critical security features.

Serverledge has been first described in a paper presented at IEEE PerCom 2023. If you use Serverledge in your own work, please cite it:

@inproceedings{serverledge2023percom,
  author={Russo Russo, Gabriele and Mannucci, Tiziana and Cardellini, Valeria and Lo Presti, Francesco},
  booktitle={2023 IEEE International Conference on Pervasive Computing and Communications (PerCom)}, 
  title={Serverledge: Decentralized Function-as-a-Service for the Edge-Cloud Continuum}, 
  year={2023},
  pages={131-140},
  doi={10.1109/PERCOM56429.2023.10099372}
}

The paper also went through an Artifact Evaluation. An artifact guide, which explains how to obtain and use the software, is available here.

Building from sources

1. Check that Golang 1.20+ is installed.

2. Download a copy of the source code.

3. Build the project:

$ make

You will find executables in ./bin/.

Running (single-node deployment)

As functions are executed within Docker containers, you need Docker to be installed on the host. Furthermore, Serverledge needs permissions to create containers.

You also need an etcd server to run Serverledge. To quickly start a local server, just download a recent release of etcd and just run ./etcd.

Start a local Serverledge node:

$ bin/serverledge

Creating and invoking functions

Register a function func from example code:

$ bin/serverledge-cli create -f func --memory 200 --src examples/hello.py --runtime python310 --handler "hello.handler" 

Invoke func with arguments a=2 and b=3:

$ bin/serverledge-cli invoke -f func -p "a:2" -p "b:3"

For non-trivial inputs, it is recommended to specify function arguments through a JSON file, instead of using the -p flag, as follows:

$ bin/serverledge-cli invoke -f func --params_file input.json

where input.json may contain:

{
	"a": 2,
	"b": 3
}

Asynchronous Invocation

Functions can be also invoked asynchronously using the --async flag:

$ bin/serverledge-cli invoke -f func --async

The server will reply with a requestID, which can be used by the client to poll for the execution result:

$ bin/serverledge-cli poll --request <requestID>

Getting function standard output

You may want to see the content printed by the function to its standard output/error. To do so, add the --return_output flag (-o for short):

$ bin/serverledge-cli invoke -f func -p "a:2" -p "b:3" --return_output

Note that we currently support output capture only for some runtimes (e.g., Python supports it).

Distributed Deployment

This repository provides an Ansible playbook to deploy Serverledge in a distributed configuration.

In this case, you can instruct serverledge-cli to connect to a node other than localhost or use a non-default port by means of environment variables or command-line options:

• Use --host <HOST> (or -H <HOST>) and/or --port <PORT> (or, -P <PORT>) to specify the server host and port on the command line
• Alternatively, you can set the environment variables SERVERLEDGE_HOST and/or SERVERLEDGE_PORT, which are read by the client.

Example:

$ bin/serverledge-cli status -H <host ip-address> -P <port number>

Configuration

You can provide a configuration file using YAML or TOML syntax. Depending on the chosen format, the default file name will be serverledge-conf.yaml or serverledge-conf.toml. The file can be either placed in /etc/serverledge, in the user $HOME directory, or in the working directory where the server is started.

Alternatively, you can indicate a specific configuration file when starting the server:

$ bin/serverledge <config file>

Further information about supported configuration options is available here.

Example

The configuration file may look like this:

container.pool.memory: 4096
etcd.address: "1.2.3.4:2379"
scheduler.queue.capacity: 0
metrics.enabled: true

Additional Documentation

• API reference
• Writing functions
• Serverledge Internals: Executor
• Metrics

License

Serverledge is distributed under the terms of the MIT license.