Skip to content

Latest commit

 

History

History
63 lines (54 loc) · 3.09 KB

running_locally.md

File metadata and controls

63 lines (54 loc) · 3.09 KB
Raw

Running Benchmarks Locally

Most benchmarks can be run locally using docker-compose for testing purposes.

You will need Docker and docker-compose. Note that a few benchmarks which use Knative Eventing can't be without Knative and don't support local execution with docker-compose.

Some benchmarks use XDT, which will require that the GOPRIVATE_KEY environment variable is set to access the private XDT repository. Some benchmarks will also require access to AWS S3 services for file transfer, this requires the AWS_ACCESS_KEY and AWS_SECRET_KEY environment variables. Contact us if you would like to ask for our keys, or see the File Transfer section for more information.

  1. cd to the benchmark's directory, e.g. ./benchmarks/chained-function-serving/.

  2. Make docker images locally with make all-image.

    Note:

    Most benchmarks will require the GOPRIVATE_KEY environment variable to be built. If you do not have access to our key yet you can usually skip this step and docker-compose should pull our pre-built image from docker hub in the next step.

  3. Start the benchmark with docker-compose. Most benchmarks have a number of docker-compose manifests, only one can be chosen. These are used to enable different features, for example docker-compose-s3.yml or docker-compose-inline.yml for choosing between which transfer type to use.

    docker-compose -f <docker-compose manifest filename> up

    Note:

    If s3 is used then the necessary environment variables, AWS_ACCESS_KEY and AWS_SECRET_KEY need to be set. see File Transfer for more details. If you currently don't have access to s3 then you will only be able to run -inline manifests

  4. In a new terminal, invoke the interface function with grpcurl. We have included the grpcurl binary in this repository:

    ../../tools/bin/grpcurl -plaintext localhost:50051 helloworld.Greeter.SayHello

    Depending on the benchmark, this might take a moment to complete. You can track the logs in the first terminal window to monitor the benchmark progress.

    Note:

    All of our docker-compose manifests are configured to expose port 50051, and all interface functions implement the helloworld.Greeter.SayHello rpc service. In rare cases additional data must be passed to the function using the -d flag- this will be stated explicitly in the README of that particular benchmark.

  5. (Optionally) You can also invoke the benchmark using the invoker. golang will be required to build the invoker binary. Find out more about the invoker here.

    # build the invoker binary
cd ../../tools/invoker
make invoker

# Specify the hostname through "endpoints.json"
echo '[ { "hostname": "localhost" } ]' > endpoints.json

# Start the invoker with a chosen RPS rate and time
./invoker -port 50051 -dbg -time 60 -rps 0.0667