Load Balancing & Price of Anarchy Game
Assign jobs to servers and compare selfish equilibria with centrally optimized load balancing.
Load balancing
How this teaches game theory
Each job is a selfish agent that prefers the server giving it the smallest completion time. A social planner may prefer a different assignment because each new job changes congestion for others. The Price of Anarchy compares the cost of selfish behavior with the centrally optimized assignment.
Complete scenario, mode, and level guide
Cloud Servers
Requests choose among cloud servers. Congestion raises response time as more requests share a machine.
CPU Scheduling
Processes select processors. The experiment emphasizes machine speed and completion time.
Warehouse Allocation
Orders select fulfillment centers, where congestion acts like handling delay.
Modes
Human Job vs Bots
You place one job at a time; remaining jobs can be filled by selfish agents.
Human Assigns All
You control every assignment and can try to beat selfish behavior manually.
Selfish Computer Agents
Each job greedily chooses its individually cheapest server at entry time.
Social Planner
The computer searches all assignments and chooses the one minimizing total completion cost.
Levels
Level 1 — Identical Servers
All servers have the same speed, making congestion effects easy to isolate.
Level 2 — Different-Speed Servers
Servers have heterogeneous speeds, so the fastest machine may become overcrowded.
Level 3 — Weighted Jobs
Jobs have different sizes, creating stronger externalities and more complex optimal assignments.