Optimizing network traffic by reducing metadata in microservices that use JWT and GRPC

Note: This is a modified version of Google's microservices-demo extended for academic research on JWT compression optimization. See ATTRIBUTION.md for details.

Overview

This repository demonstrates JWT compression techniques in a microservices architecture:

• JWT Authentication: Token-based authentication across services
• JWT Compression: Optional compression feature that splits JWT into multiple headers for better HTTP/2 HPACK caching
• Performance Testing: Comprehensive benchmarking tools to measure compression benefits

Thesis

Thesis is publicly available here

Research Focus

This project investigates the performance impact of JWT compression in microservices environments, specifically:

• Network bandwidth reduction through header optimization
• HTTP/2 HPACK caching efficiency with split JWT headers
• Latency improvements under high concurrent load
• Scalability characteristics with compressed vs. uncompressed tokens

Prerequisites

The following tools should already be available in your Github Codespace (at least 4core and 16GB ram):

• kubectl - Kubernetes CLI
• jq - JSON processor
• tshark - Network analysis
• git - Version control
• minikube - Lightweight Kubernetes

Step-by-Step Instructions

Step 1: Enable JWT Compression and Run Test

sudo apt update && sudo apt upgrade

minikube start

minikube ssh "sudo apt-get update && sudo apt-get install -y tcpdump"

Install skaffold for linux x86_64

curl -Lo skaffold https://storage.googleapis.com/skaffold/releases/latest/skaffold-linux-amd64 && sudo install skaffold /usr/local/bin/ && rm skaffold && skaffold version

Install k6 and tcpdump

curl -L https://github.com/grafana/k6/releases/download/v0.51.0/k6-v0.51.0-linux-amd64.tar.gz -o k6.tar.gz && tar -xzf k6.tar.gz && sudo mv k6-v0.51.0-linux-amd64/k6 /usr/local/bin/ && rm -rf k6.tar.gz k6-v0.51.0-linux-amd64 && k6 version

Install tshark

sudo DEBIAN_FRONTEND=noninteractive apt-get installsudo DEBIAN_FRONTEND=noninteractive apt-get install

sudo apt-get update && sudo apt-get install -y tcpdump

Enable JWT compression across all services:

./enable_jwt_compression.sh

This will take a little while.

What this does:

• Updates YAML manifests to set ENABLE_JWT_COMPRESSION=true
• Deploys changes using skaffold run
• Rebuilds and restarts all affected pods

Step 2: Verify Pods are Running

Before running the test, confirm all pods are healthy:

kubectl get pods

If pods are not ready, wait a few more minutes and check again.

Step 3: Run JWT Compression Test (Enabled)

The default settings for the test is for 512kb Dynamic Table size and for 200 Users

Run the load test with JWT compression enabled:

./run-jwt-compression-test.s

What this does:

• Starts network packet capture (tcpdump)
• Runs k6 load test simulating user journeys
• Captures performance metrics
• Saves results to a timestamped directory

Step 4: Disable JWT Compression and Run Test

Now disable JWT compression:

./disable_jwt_compression.sh

Step 5: Verify Pods are Running

Verify all pods restarted successfully:

kubectl get pods

Wait for all pods to be Running with 1/1 ready.

Step 6: Run JWT Compression Test (Disabled)

Run the load test again with JWT compression disabled:

./run-jwt-compression-test.sh

Step 7: Compare Results

Now compare the performance between both tests:

Edit the script first to set your test directories at the top:

# Edit these lines with your actual directory names
ENABLED_DIR="jwt-compression-on-results-20251017_143022"
DISABLED_DIR="jwt-compression-off-results-20251017_144530"

./compare-jwt-compression-enhanced.sh

What the comparison shows:

1. K6 Load Test Results

   • Iterations, request rate
   • Data sent/received
   • Response times (avg, p95, p99)
   • Failed requests and checks

2. Performance Improvements

   • Bandwidth savings (upload/download)
   • Response time differences
   • Network traffic reduction

3. Network Traffic Analysis

   • Total packets and bytes
   • HTTP/2 traffic patterns
   • JWT header usage

4. JWT Header Analysis

   • Compression ON: 4 headers (~744 bytes first request, ~470 bytes cached)
   • Compression OFF: 1 header (~900+ bytes every request)

5. Summary

   • Overall bandwidth savings percentage
   • Performance improvements
   • Key benefits

Troubleshooting

Pods Not Starting

# Check pod status and events
kubectl describe pod <pod-name>

# Check logs
kubectl logs <pod-name>

# Restart deployment
kubectl rollout restart deployment <service-name>

Test Script Fails

# Ensure scripts are executable
chmod +x *.sh

# Check if k6 is installed
k6 version

# Verify network tools
which tcpdump tshark

Skaffold Errors

# Check Skaffold configuration
skaffold diagnose

# Clean and rebuild
skaffold delete
skaffold run

Comparison Script Issues

# Verify test result directories exist
ls -la jwt-compression-*-results-*/

# Check if jq is installed
jq --version

# Manually check results
cat jwt-compression-on-results-*/k6-summary.json | jq .

---

Last updated: December 14, 2025