davidfreina / syssem-artifact-evaluation

Reproducing FaaSFlow

This repository should act as a logbook for the steps I have taken to reproduce the artifact from the paper "FaaSFlow". This is only a loose collection of my thoughts and problems throughout reproducing the artifacts.

Setting up environment

To reproduce the artifact I need 8 VMs. 7 Worker nodes with 8 cores and 32GB RAM each, 1 Master/DB node with 16 cores and 64GB RAM All nodes should be running Ubuntu 18.04. Setup scripts for both node types are available in the GitHub repository of the artifact.

Creating the VMs

I decided to reproduce the artifact on GCE because I am more familiar with the GCP than Alibaba Cloud. In order to setup the VMs I had to request quota increases for the SSDs and CPUs. I had to upgrade my Google Account from the trial version to a full account for this.

ERROR: (gcloud.compute.instances.create) Could not fetch resource:
 - Quota 'CPUS' exceeded.  Limit: 24.0 in region europe-west4.
        metric name = compute.googleapis.com/cpus
        limit name = CPUS-per-project-region
        dimensions = region: europe-west4
Try your request in another zone, or view documentation on how to increase quotas: https://cloud.google.com/compute/quotas.
 - Quota 'SSD_TOTAL_GB' exceeded.  Limit: 500.0 in region europe-west4.
        metric name = compute.googleapis.com/ssd_total_storage
        limit name = SSD-TOTAL-GB-per-project-region
        dimensions = region: europe-west4
Try your request in another zone, or view documentation on how to increase quotas: https://cloud.google.com/compute/quotas.

Setting up the DB Node

I used the db_setup.bash script from the repository. The command failed on the second to last line because I have not cloned the repository.

Then I forked the repository and changed the configurations accordingly. While running the setup script I noticed that pip3 was not found so I installed it manually. Furthermore, the setup script has to be executed in the folder it is located because otherwise a cd command fails. The installation also failed because an old version of PyYAML was installed. I had to update the paths for WORKFLOW_YAML_ADDR in config/config.py. I also had to add a mapping for the custom couchdb config to the docker command.

After these steps I was able to setup the maste/DB node.

Setting up the worker nodes

After fixing all the issues desribed in the previous section and adding them to a fork of the repository I was able to setup the workers. I just executed the worker_setup.bash script with sudo and waited for its completion.

Running the experiments

Starting proxy on workers

Ansible

ansible-galaxy collection install community.docker