CLUSTER RESOURCE MANAGER

This project implements a Cluster Resource Manager in C++ that manages task allocation across a distributed set of nodes. It simulates a simplified real-world cluster system, including dynamic memory-aware task scheduling, load handling, health monitoring, and failover recovery.

---

🧠 Project Objectives

• Efficient Load Balancing
• Dynamic Memory-Aware Task Allocation (Each task specifies its memory requirement; manager schedules tasks dynamically based on node memory)
• Health Monitoring (Periodic node status checks)
• Failover & Recovery (Automatic task reallocation)
• Performance Logging
• Demonstrates core Linux OS concepts like:
  • Memory Management
  • Virtual Memory
  • Processes
  • Scheduling
  • Multi-threading
  • Inter-process Communication
  • Socket Programming
  • File System

---

⚙️ Architecture

• Manager (manager.cpp)
  Accepts node and client connections, tracks node status, assigns tasks (dynamic memory-aware), monitors node health, handles failures.

• Node Agent (node_agent.cpp)
  Connects to the manager, registers itself (with available memory), accepts and executes tasks, sends heartbeat pings.

• Client (client.cpp)
  Sends a batch of tasks to the manager for processing. Each task has a random memory requirement (6-126 MB).

---

📋 Custom Algorithms Used

This project uses a Dynamic Memory-Aware Greedy + FCFS scheduling strategy with health monitoring, reactive failover, and thread-based concurrency.

• Dynamic Memory-Aware Scheduling: Each task specifies its memory requirement; nodes are only assigned tasks if they have enough memory. The manager dynamically adapts as tasks complete and memory is freed.
• FCFS Task Assignment: Tasks assigned in order of arrival.
• Greedy Dispatch: Tasks are assigned immediately to the first node with enough available memory.
• Health Monitoring: Manager checks node status every 10 seconds and reallocates tasks from failed nodes.
• Failover: When a node crashes or disconnects, its tasks are reassigned.
• Multithreading: Each connection is handled in a separate thread.

---

🛠️ Build & Run

Prerequisites

• g++ (C++17)
• make
• bash (for build.sh)

Build

bash build.sh

Run

1. Start the manager:

   ./build/manager

2. Start one or more node agents (in separate terminals):

   ./build/node_agent node1 127.0.0.1 5000 9001
   ./build/node_agent node2 127.0.0.1 5000 9002
   ...

3. Submit tasks from the client:

   ./build/client 127.0.0.1 5000 10

---

Notes

• No CMake is used; the project is built with Makefile and build.sh.
• Tasks are memory-aware and randomly sized (6-126 MB).
• Health monitoring and failover are automatic.
• DAG scheduling is not implemented; all tasks are independent.