RetainZ: Reclaim-Time-Aware Placement for AI Checkpoints on Zoned SSDs

2026-10-04

AI Summary

This study addresses the reclamation bottlenecks and high latency in Zoned Namespace (ZNS) SSDs during AI checkpoint writing, which stem from whole-zone erasure. We propose a retention-aware explicit reclamation epoch mechanism that eliminates the need for data lifetime prediction. By isolating long-lived data and aggregating objects with similar reclamation periods, this approach optimizes space allocation to balance reclamation latency against fragmentation waste. Evaluations conducted on the FEMU emulation platform demonstrate that, compared to ZenFS, our method reduces invalid space overhead by 41.7% and decreases P99 tail latency by 47.5%, while achieving a write amplification factor of 1.003.

Abstract

Solid State Drives (SSDs) are becoming the dominant medium for performance-critical storage. Zoned Namespace (ZNS) SSDs are getting more and more attractive because sequential writes and explicit zone resets reduce address-mapping, over-provisioning, and internal garbage-collection costs. However, reclaiming space requires resetting an entire zone, so deleting one file does not free its space while other files there must be kept. This erase-before-write constraint becomes a bottleneck for write-intensive workloads, particularly AI checkpointing during model training, where frequent saves of model and optimizer state compete for space with long-retained checkpoints. We observe that retention policies inherently encode when data can be retired, avoiding the need for a learned lifetime predictor. We present RetainZ, a storage backend that translates these policies into explicit reclaim epochs. Its allocator isolates long-retained data and groups objects with nearby reclaim epochs, balancing delayed space reclamation against unused space at zone ends. Evaluated on FEMU with multi-tenant streams and checkpoint manifests up to Pythia-1B, RetainZ reduces space occupied by deleted data and unused zone tails by 41.7% and mean per-input checkpoint p99 latency by 47.5% at 84% target load compared with the allocator of ZenFS, an established lifetime-aware ZNS backend, and lowers host write amplification from 1.165 to 1.003.

Problem

• Zoned Namespace SSD
• AI checkpointing
• reclaim-time-aware placement
• erase-before-write
• write amplification

Innovation

• Zoned Namespace SSD
• AI Checkpointing
• Reclaim-Time-Aware Placement
• Write Amplification
• Storage Allocation