Trading System - Distributed computing infrastructure engineering

This project is part of Distributed computing infrastructure engineering course at University of Warsaw during the 2025/2026 academic year.

The course is realised in collaboration with Google and one of the employees was a mentor.

Authors: Tomasz Głąb, Hubert Krupniewski. Piotr Wieczorek

How to run on GCP

There are two ways to run the project on GCP:

1. Automatically, using sh scripts/deploy.sh script.
2. Manually, following the instructions below.

Create cluster

gcloud container clusters create trading-cluster \
--zone europe-central2-b \
--num-nodes 1 \
--machine-type e2-standard-2

Get credentials

gcloud container clusters get-credentials trading-cluster --zone europe-central2-b

Install KRaft Kafka with the usage of helm

helm install my-kafka oci://registry-1.docker.io/bitnamicharts/kafka -f k8s/kafka-values.yaml

If you want to update kafka-values.yaml and reload Kafka, you can use:

helm uninstall my-kafka && kubectl delete pvc data-my-kafka-controller-0

Install postgres DB with the usage of helm:

helm install my-postgres oci://registry-1.docker.io/bitnamicharts/postgresql \
      --set auth.postgresPassword=postgres \
      --set auth.database=trading_db \
      --wait

And then, proceed with install once again.

Deploy to kubernetes

Command below does the following:

• Starts the job that creates kafka topics.
• Deploys worker, application-frontend, application-frontend service and executed-trades worker.
• Throws an error that kafka-values.yaml can't be processed - can be ignored.

kubectl apply -f k8s/

Check if it works

In a new terminal tab run the following command, to see the logs from all the components:

kubectl logs -f -l 'app in (afe, worker, executed-trades-worker)' --max-log-requests=10

If everything deployed correctly afe-service should be accessible from outside of the cluster. To obtain the extrernal_ip you have to run

kubectl get svc

Send a few requests to the system with the use of the client:

# Remeber to set AFE_SERVICE_EXTERNAL_IP variable or 
# to replace it with the IP address of the service

# Sell GOOG 100 @ 10
bazel run //client -- \
  -server_address "{AFE_SERVICE_EXTERNAL_IP}:80" \
  -endpoint "SubmitTrade" \
  -body '{"trade": {"side": "SELL", "price": "100", "size": "10", "instrument": {"symbol": "GOOG"}}}'

# Buy GOOG 100 @ 10
bazel run //client -- \
  -server_address "{AFE_SERVICE_EXTERNAL_IP}:80" \
  -endpoint "SubmitTrade" \
  -body '{"trade": {"side": "BUY", "price": "100", "size": "10", "instrument": {"symbol": "GOOG

Verify if outcome is desired.

Alternatively, you can setup port-forwarding and use the default address and port on the client which is localhost:50051.

Clean up

Remember to delete the cluster when you are done with it, to save the reosurces.

Just run:

./scripts/destroy.sh

Bazel cheatsheet

Fixup BUILD files, for example auto add dependencies:

bazel run //:gazelle

Depend on a new tool:

bazel run @rules_go//go -- get -tool your_tool_repo.com/x/useful/tool

To add a new external dependency from Bazel Central Registry, add the dependency to MODULE.bazel:

bazel_dep(name = "my_library", version = "1.0.0")