Publications
Earlier research publications focused on flash memory systems, storage architecture, and performance tradeoffs in SSD-backed environments.
This paper examines two major Flash NAND problems: random-write latency and limited chip endurance. It explores a Flash-PCM memory design intended to improve write behavior and extend device lifetime while balancing performance and endurance tradeoffs.
WAFTL proposes a workload-adaptive flash translation layer that switches between page-level and block-level mapping based on access patterns. The design reduces mapping-table size, lowers garbage-collection overhead, and improves performance under enterprise workloads through data partitioning and threshold-based migration.
FlashCoop presents a locality-aware cooperative buffer-management approach for SSD-based storage clusters. The work focuses on using locality information across the cluster to improve cache effectiveness, reduce unnecessary SSD traffic, and improve overall storage performance.