AdaptRoute

A policy-driven middleware framework for runtime message broker selection in Spring Boot microservices.

AdaptRoute sits between Spring Boot producers and two message brokers (Apache Kafka and RabbitMQ), evaluating an asymmetric weighted scoring model at every dispatch to route each message class to the higher-scoring broker.

Architecture

Producers --> Message Dispatcher --> Apache Kafka --> Consumers
                    |           --> RabbitMQ    --> Consumers
                    v
              Scoring Engine
                    ^
            Broker Load Monitor

Modules

Module	Description
common	Shared DTOs, scoring engine (WeightedScoringRouter), broker load monitor
gateway-service	Message dispatcher, circuit breakers, failover buffer, health checks
order-service	ORDER message producer
notification-service	NOTIFICATION message producer
analytics-service	ANALYTICS message producer
experiments	Experiment runner and workload generator
plots	Python scripts for generating figures and statistical tests

Prerequisites

• Java 21 (OpenJDK)
• Docker & Docker Compose
• Maven 3.9+
• Python 3.10+ (for plots: pip install -r plots/requirements.txt)

Quick Start

# Start the testbed (Kafka, RabbitMQ, all services)
docker compose up -d

# Run the full 480-run experiment suite
# (4 strategies x 4 workloads x 30 repetitions)
docker compose run experiments

# Generate figures
cd plots && python generate_figures.py

Configuration

All routing parameters are Spring configuration properties:

Parameter	Default	Description
router.weights.kafka.*	See paper Table 3	Kafka scoring weights (sum = 1.0)
router.weights.rabbit.*	See paper Table 3	RabbitMQ scoring weights (sum = 1.0)
adaptroute.buffer.capacity	10,000	Failover buffer size (messages)
health.check.interval.ms	500	Health probe interval
health.check.timeout.ms	3,000	AdminClient timeout
load.ceiling.msgs	10,000	Throughput ceiling for load factor

Experiment Workloads

Workload	Rate	Composition	Purpose
W1 High-Throughput	500 msg/s	40% ORDER, 50% ANALYTICS, 10% NOTIFICATION	Flash-sale simulation
W2 Low-Latency	300 msg/s	90% NOTIFICATION (70% high-priority)	Real-time alerting
W3 Mixed Traffic	400 msg/s	Equal thirds ORDER/NOTIFICATION/ANALYTICS	Balanced routing
W4 Ordering-Critical	200 msg/s	100% ORDER (EXACTLY_ONCE)	Ordering correctness

Reproducibility

• Seeds: WORKLOAD_SEED=42, JITTER_SEED=137
• Docker images: SHA256 digest-pinned in docker-compose.yml
• Results: Raw CSVs in results/experiment_results.csv
• Figures: Regenerate with python plots/generate_figures.py

Paper

AdaptRoute: A Policy-Driven Middleware Framework for Multi-Protocol Message Broker Selection in Spring Boot Microservices

Umur Inan, Ata Turhan

Submitted to the Journal of Universal Computer Science (JUCS).

License

MIT