You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
Go to file
lyuxiaosu 6d5974dc7a
upload scripts
3 years ago
.devcontainer chore: formatting nits 4 years ago
.github/workflows chore: Remove GitHub yaml zombie 4 years ago
.vscode chore: stop hiding wasmception 4 years ago
awsm@cd9b61958f chore: Update awsm submodule 4 years ago
runtime upload scripts 3 years ago
.clang-format chore: align consecutive macros 5 years ago
.editorconfig chore: formatting nits 4 years ago
.env refactor: assorted bash cleanup 4 years ago
.gitignore feat: VSCode build container and visual debugging 4 years ago
.gitmodules change awsm url 4 years ago
Dockerfile.aarch64 chore: formatting nits 4 years ago
Dockerfile.x86_64 chore: Install wasmception after bind mount 4 years ago
LICENSE chore: first rename pass 4 years ago
Makefile add -u 0 for docker exec and set default toolchain for cargo 4 years ago
README.md chore: validation and active as boolean 4 years ago
add_partition.sh upload add_partition.sh and install_docker.sh 3 years ago
devenv.sh add -u 0 for docker exec and set default toolchain for cargo 4 years ago
fix_root.sh chore: improve fix_root script 4 years ago
format.sh chore: Apply shfmt to shell scripts 4 years ago
install.sh chore: Apply shfmt to shell scripts 4 years ago
install_docker.sh upload add_partition.sh and install_docker.sh 3 years ago
install_llvm.sh chore: Install clang-format-11 as minimum 4 years ago
install_perf.sh refactor: assorted bash cleanup 4 years ago
test.sh refactor: assorted bash cleanup 4 years ago

README.md

SLEdge

SLEdge is a lightweight serverless solution suitable for edge computing. It builds on WebAssembly sandboxing provided by the aWsm compiler.

Host Dependencies

Setting up the environment

Note: These steps require Docker. Make sure you've got it installed!

We provide a Docker build environment configured with the dependencies and toolchain needed to build the SLEdge runtime and serverless functions.

To setup this environment, run:

./devenv.sh setup

Using the Docker container to compile your serverless functions

To enter the docker environment, run:

./devenv.sh run

The first time you enter this environment, run the following to copy the sledgert binary to /sledge/runtime/bin.

cd /sledge/runtime
make clean all

There are a set of benchmarking applications in the /sledge/runtime/tests directory. Run the following to compile all benchmarks runtime tests using the aWsm compiler and then copy all resulting <application>_wasm.so files to /sledge/runtime/bin.

cd /sledge/runtime/tests/
make clean all

You now have everything that you need to execute your first serverless function on SLEdge

To exit the container:

exit

To stop the Docker container:

./devenv.sh stop

Running your first serverless function

An SLEdge serverless function consists of a shared library (*.so) and a JSON configuration file that determines how the runtime should execute the serverless function. As an example, here is the configuration file for our sample fibonacci function:

{
  "active": true,
  "name": "fibonacci",
  "path": "fibonacci_wasm.so",
  "port": 10000,
  "expected-execution-us": 600,
  "relative-deadline-us": 2000,
  "argsize": 1,
  "http-req-headers": [],
  "http-req-content-type": "text/plain",
  "http-req-size": 1024,
  "http-resp-headers": [],
  "http-resp-size": 1024,
  "http-resp-content-type": "text/plain"
}

The port and name fields are used to determine the path where our serverless function will be served served.

In our case, we are running the SLEdge runtime on localhost, so our function is available at http://localhost:10000/fibonacci

Our fibonacci function will parse a single argument from the HTTP POST body that we send. The expected Content-Type is "text/plain" and the buffer is sized to 1024 bytes for both the request and response. This is sufficient for our simple Fibonacci function, but this must be changed and sized for other functions, such as image processing.

Now that we understand roughly how the SLEdge runtime interacts with serverless function, let's run Fibonacci!

From the root project directory of the host environment (not the Docker container!), navigate to the binary directory

cd runtime/bin/

Now run the sledgert binary, passing the JSON file of the serverless function we want to serve. Because serverless functions are loaded by SLEdge as shared libraries, we want to add the runtime/tests/ directory to LD_LIBRARY_PATH.

LD_LIBRARY_PATH="$(pwd):$LD_LIBRARY_PATH" ./sledgert ../tests/test_fibonacci.json

While you don't see any output to the console, the runtime is running in the foreground.

Let's now invoke our serverless function to compute the 10th fibonacci number. I'll use HTTPie to send a POST request with a body containing the parameter I want to pass to my serverless function. Feel free to use cURL or whatever network client you prefer!

Open a new terminal session and execute the following

echo "10" | http :10000

You should receive the following in response. The serverless function says that the 10th fibonacci number is 55, which seems to be correct!

HTTP/1.1 200 OK
Content-length: 3
Content-type: text/plain

55

When done, terminal the SLEdge runtime with Ctrl+c

Running Test Workloads

Various synthetic and real-world tests can be found in runtime/experiments. Generally, each experiment can be run be executing the run.sh script.

Removing the SLEdge Runtime

If you are finished working with the SLEdge runtime and wish to remove it, run the following command to delete our Docker build and runtime images.

./devenv.sh rma

And then simply delete this repository.

Problems or Feedback?

If you encountered bugs or have feedback, please let us know in our issue tracker.