Eliminating the Hidden Cost of Zone Management in ZNS SSDs

TL;DR

The paper addresses device-level write amplification and wear in ZNS SSDs caused by fixed physical zones and full-zone management. It proposes SilentZNS, a flexible zone allocation framework that maps logical zones to on-demand storage elements and applies dummy writes only to partially written elements, with chunk and stripe modes to balance DLWA, SA, and parallelism. The authors implement SilentZNS in the ConfZNS++ emulator and show substantial reductions in DLWA (up to ~95%), reduced wear (up to ~77%), and faster workloads (up to ~3.7x), while keeping SA modest (≈0.42). The work demonstrates that decoupling logical zone lifecycles from fixed physical mappings dramatically improves ZNS efficiency and highlights the need for more flexible host interfaces to enable dynamic zone creation.

Abstract

Zoned Namespace (ZNS) SSDs offer a promising interface for stable throughput and low-latency storage by eliminating device-side garbage collection. They expose storage as append-only zones that give the host applications direct control over data placement. However, current ZNS implementations suffer from (a) device-level write amplification (DLWA), (b) increased wear, and (c) interference with host I/O due to zone mapping and management. We identify two primary design decisions as the main cause: (i) fixed physical zones and (ii) full-zone operations that lead to excessive physical writes. We propose SilentZNS, a new zone mapping and management approach that addresses the aforementioned limitations by on-the-fly allocating available resources to zones, while minimizing wear, maintaining parallelism, and avoiding unnecessary writes at the device-level. SilentZNS is a flexible zone allocation scheme that departs from the traditional logical-to-physical zone mapping and allows for arbitrary collections of blocks to be assigned to a zone. We add the necessary constraints to ensure wear-leveling and state-of-the-art read performance, and use only the required blocks to avoid dummy writes during zone reset. We implement SilentZNS using the state-of-the-art ConfZNS++ emulator and show that it eliminates the undue burden of dummy writes by up to 20x, leading to lower DLWA (86% less at 10% zone occupancy), less overall wear (up to 76.9%), and up to 3.7x faster workload execution.

Figures (4)

• Figure 1: SilentZNS allows a ZNS SSD to offer reduced DLWA without any penalty regarding space amplification. In contrast, state-of-the-art ZNS approaches can reduce DLWA at the expense of increased space amplification. The percentage associated with each point is the finish threshold of a zone.
• Figure 2: (a) SilentZNS constructs a zone from erase block as chunk size is 1 ($c_s = 1$). (b) SilentZNS constructs a zone from chunks where each chunk has 2 erase blocks ($c_s = 2$). (c) SilentZNS constructs a zone from stripes that contain a fixed set of erase blocks from every LUN. (d) Baseline approaches represent a zone as a fixed set of erase blocks.
• Figure 3: (a) SilentZNS's DLWA benefits are highest at low zone occupancy. (b) A higher FINISH threshold reduces SA. (c) SilentZNS reduces zone management writes as the FINISH threshold increases. (d) SilentZNS distributes wear more evenly.
• Figure 4: (a) SilentZNS maintains stable throughput with concurrent FINISH. (b) SilentZNS maintains competitive read/write throughput. (c) SilentZNS improves the workload latency. (d) stripe configuration is the most efficient in zone allocation.