ECCOS: Event-Driven Computing Continuum Orchestration Solver — Benchmark Simulation Traces and Placement Dataset

This dataset provides high-resolution, telemetry-rich discrete-event simulation traces generated by **ECCOS** (*Event-Driven Computing Continuum Orchestration Solver*). The dataset contains the complete 5,000-iteration benchmark runs evaluated in the research article: > *"Event-Driven Computing Continuum Orchestration Solver (ECCOS): A Discrete-Event Simulation Framework for Optimal Service Placement and ML-Ready Trace Generation"*, Mireia Jaume, Isaac Lera, Carlos Guerrero (Submitted for evaluation). ### Dataset Overview The archive contains 20,000 step-by-step state transition files formatted as Markov Decision Processes (MDP) $(s_t, a_t, r_t, s_{t+1})$, totaling ~3.1 GB of raw JSON telemetry across four canonical benchmark evaluation scenarios: 1. **Electric Storm (`electric_storm_multi_ilp_all_20260910_130039`):** Severe correlated node and link failure cascades across geographical zones, testing multi-objective placement recovery, failover policies, and service migration under degraded capacity. 2. **Crowd Event (`crowd_event_multi_ilp_all_20260916_131352`):** Massive localized influx of mobile users into an ultra-dense area, testing edge access-point congestion, user mobility handoffs, and compute offloading. 3. **Demand Surge (`demand_surge_multi_ilp_all_20260910_131401`):** Sudden viral surge in service request frequency combined with localized link degradations (brownouts), testing dynamic elasticity and SLA compliance. 4. **Normal Conditions (`normal_conditions_multi_ilp_all_20260910_072946`):** Steady-state baseline scenario with standard Poisson request arrivals, continuous uniform mobility, and degree-correlated resource distribution across Cloud, Fog, and Edge tiers. ### File Structure & Data Model Each discrete step is saved as `Simulation{i}.json` capturing: - `users_before` / `users_after`: User spatial coordinates $(x,y)$, mobility vectors, connected access points, and per-application request rates. - `apps_before` / `apps_after`: Microservice chains (SFC/DAG) with compute footprints (CPU, RAM). - `placement_before` / `placement_after`: Active allocation matrix mapping microservices to physical continuum nodes. - `node_before` / `node_after`: Computing capacities, load utilization, and operational/degraded states. - `edge_before` / `edge_after`: Network topology links, latencies, available bandwidth, and failure states. - `action`: The triggering discrete event (e.g., node_drop, user_move, app_request, link_degrade). - `metrics_before` / `metrics_after`: End-to-end latency, energy consumption, migration costs, solver execution time, and global objective values. ### Code Repository & Reproduction The simulation framework, plotting pipeline, and experiment runners are openly available on GitHub: https://github.com/acsicuib/ECCOS