BLOCProxy Performance

The entire experiment can be run on an existing 4 node cluster (look here) by running the run.sh bash script.

A kubernetes cluster on Ubuntu 18 or 20 nodes can be setup by following this guide.

Get All the Repos

The best way of getting the repos is through the submodules command.

git clone --recurse-submodules -j4 https://github.com/MSrvComm/Experiments

If you need to checkout BLOCProxy separately, then use the stable branch (non-default branch).

Citation:

BLOC: Balancing Load with Overload Control In the Microservices Architecture

Build and Deploy

Assumptions

• We assume that the experiments will be running on a Kubernetes cluster able to run 50-60 pods with 2 containers each. Further details on configuration can be found in the deployement yaml files. In our experiments, we used 4 Cloudlab servers, each with four 10 core sockets and 196 GB of RAM.

• The experiments are being run from an Ubuntu machine.

• Python3.9 and the bash shell is available on the system.

• We use matplotlib, numpy and pandas to generate our results.

(Optional - if you want to test with your own images; otherwise start with [Control Plane](#Deploy the Control Plane))

Define Image Names as Environment Variables

# format: hub_name/image_name:version
export WEB_IMG_NAME=ratnadeepb/testapp:latest
export BLOC_PROXY_IMG_NAME=ratnadeepb/micoproxy:latest
export LC_PROXY_IMG_NAME=ratnadeepb/micoproxy:leastconn

Build the Docker Images

cd BLOC/app && ./docker_build.sh $WEB_IMG_NAME && cd $OLDPWD
cd BLOCProxy && ./build_proxy.sh $BLOC_PROXY_IMG_NAME && cd $OLDPWD

Deploy the Control Plane

cd BLOCControl/restEpWatcher && ./deploy.sh && cd $OLDPWD

Remove taint and Label the nodes

kubectl taint nodes --all node-role.kubernetes.io/master-

We labeled the nodes so that 10 backend servers can run on one physical node.

Example command (node names will have to be changed):

kubectl label nodes node0,node1,node2 workload=svc1
kubectl label nodes node3 workload=svc2

Install the hey load generator

wget https://hey-release.s3.us-east-2.amazonaws.com/hey_linux_amd64
chmod +x hey_linux_amd64
sudo mv hey_linux_amd64 /usr/local/bin/hey

Install Python Requirements

pip install -r requirements.txt

Running the experiments

./run.sh

Graphing the results as in Figure 8 of the paper

This will store the results in a file called BLOCProxyVsLeastConn.pdf

./graph.py

Reference Data

We have already provided the CSV files and pdf diagram as a reference of the result. These files have the same names but suffixed with _ref.