580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How do consistency protocols impact the availability and performance of replicated files in distributed systems?
This area focuses on the design, evaluation, and comparison of consistency protocols that maintain data consistency in replicated file systems under site failures and network conditions. It matters because the choice of protocol directly affects data availability, reliability, fault tolerance, and system overhead in distributed environments.
Key finding: Using Markov models, this study demonstrates that available copy protocols, including naive and optimistic variants that do not require instantaneous failure detection, achieve nearly the same high availability and... Read more
Key finding: Simulation analysis comparing Static Majority Consensus Voting, Dynamic Voting (DV), and Lexicographic Dynamic Voting (LDV) under realistic network parameters finds that both DV and LDV improve data availability over majority... Read more
Key finding: This research develops a method for detecting the last site to fail in available copy protocols to enhance recovery performance after total failure. By tracking which sites received the most recent updates without requiring... Read more
Key finding: By analyzing replica control (RC) protocols, this paper highlights a shift from focusing on failure unavailability to conflict unavailability (unavailability caused by conflicting operations) in replica control. It... Read more
2. What strategies and algorithms optimize data replication and placement for performance and resource efficiency in distributed and cloud file systems?
This theme explores techniques and models that predict, select, or dynamically adjust replication factors, replica placement, and replication policies to improve access performance, improve data locality, lower network overhead, and balance storage costs in distributed and cloud storage systems. This is essential given large-scale data volumes and the importance of efficient data access under resource constraints.
Key finding: This paper introduces an algorithm using a binary tree structure to estimate file demand potential by tracking temporal locality and demand trends over time intervals. By considering file popularity dynamics, scope, and type,... Read more
Key finding: Proposes a hybrid data replication strategy based on voting structures that balances trade-offs between read and write availability, operation costs, and scalability. The approach recognizes that existing methods cannot... Read more
Key finding: Introduces hybrid replication which partitions storage volumes into extents assigned to either continuous or snapshot replication based on overwrite frequency. This technique achieves bandwidth efficiency close to snapshot... Read more
Key finding: Experimental study on a 20-node Hadoop cluster demonstrates that increasing the replication factor for 'hot' (frequently accessed) data significantly improves data availability and locality, reducing job execution times. This... Read more
Key finding: Develops a dynamic replication policy for HDFS by clustering files using unsupervised machine learning to classify them based on usage importance. Different replication strategies are applied per cluster, optimizing storage... Read more
3. How can emerging technologies like blockchain and advanced security mechanisms improve file synchronization and security in distributed and cloud file replication systems?
This research theme investigates novel applications of blockchain technology and fragmentation-replication combined security schemes aiming to enhance immutability, auditability, and integrity for file synchronization and distributed storage, addressing challenges in trust, data provenance, multiuser access, and resilience against data attacks in cloud systems.
Key finding: Proposes a resilient cloud storage solution using blockchain to store differential file updates as linked blocks, embedding version and user signature information to enable immutable, verifiable file versions in a hybrid... Read more
Key finding: Introduces a security mechanism where files are fragmented and hashed, with each fragment stored and replicated in separate cloud locations. The approach mitigates risks from attacks by ensuring no single fragment reveals... Read more