D-SPTF
D-SPTF is a decentralized storage protocol that coordinates replicated read queues using cache hits and predicted disk positioning times.
A cache miss is forwarded to replica holders. Claims remove competing queued reads; concurrent bids prefer shorter positioning time, then lower brick ID. Pre-scheduling provides a short coordination window. [1]
Delayed or lost claims can still cause duplicate work. A claimant that fails before replying requires client timeout and retry. This is not an exactly-once guarantee. [1]
The ASPLOS 2004 evaluation uses DiskSim, not a deployed fleet. Modern SSD performance is not established. [1]
Relations
Citations
Sources & caveats
- Primary-paper review; no simulator rerun or hardware experiment.