DataSys: Data-Intensive Distributed Systems LaboratoryData-Intensive Distributed Systems Laboratory

Illinois Institute of Technology
Department of Computer Science

ZHT: a zero-hop distributed hashtable

ZHT is a zero-hop distributed hash table, which has been tuned for the specific requirements of high performance computing (e.g. trustworthy/reliable hardware, fast networks, non-existent “churn”, low latencies, and scientific computing data-access patterns). ZHT aims to be a building block for future distributed systems, with the goal of delivering excellent availability, fault tolerance, high throughput, scalability, persistence, and low latencies. ZHT has several important features making it a better candidate than other distributed hash tables and key-value stores, such as being light-weight, dynamically allowing nodes join and leave, fault tolerant through replication and by handling failures gracefully and efficiently propagating events throughout the system, a customizable consistent hashing function, supporting persistence for better recoverability in case of faults, scalable, and supporting unconventional operations such as append (providing lock-free concurrent key/value modifications) in addition to insert/lookup/remove. To provide ZHT a persistent back end, we also created a fast persistent key-value store that could be easily integrated and operated in lightweight Linux OS typically found on today’s supercomputers.

Performance comparison on BlueGene/P

Performance comparison on Amazon EC2