## Chord with Replication > Project 16 > Moida Praneeth Jain (2022101093) Yajat Rangnekar (2023114008) --- ### Architecture 1. **Chord Node**: Core logic and state 2. **Chord Client**: CLI for get/put 3. **Chord Proto**: Shared gRPC definitions 4. **Monitor**: Live visualization --- ### Core logic - **Consistent Hashing**: SHA-1 (160-bit space) - **Finger tables**: For $O(log(n))$ lookups -v- ### Dynamic Membership (Churn) - **Join**: Connect via existing node $\rightarrow$ Find successor - **Leave**: Graceful exit & key transfer - **Stabilize**: Periodic background task -v- ### Fault Tolerance - **Successor List**: $r$ nearest successors - Periodic pings to detect failures - Ring reconnects automatically -v- ### Replication Strategy - Primary + $k$ successors - Active background maintenance - Prioritize availability --- ### Benchmarks and Analysis -v- #### Scalability <img src="chord_node/benchmark_results/scalability.png" style="width:800px;"/> -v- #### Load Balancing <img src="chord_node/benchmark_results/load_balancing.png" style="width:800px;"/> -v- #### Concurrent Throughput <img src="chord_node/benchmark_results/concurrent_throughput.png" style="width:800px;"/> -v- #### Replication Delay <img src="chord_node/benchmark_results/replication_delay.png" style="width:800px;"/> --- ### Learnings and extensions - Experienced tradeoff between availability and consistency - Handling failures and race conditions is 80% of the complexity - _Virtual nodes_ for better load balancing