Stochastic Modeling of Large-Scale Solid-State Storage Systems: Analysis, Design Tradeoffs and Optimization

Solid state drives (SSDs) have seen wide deployment in mobiles, desktops, and data centers due to their high I/O performance and low energy consumption. As SSDs write data out-of-place, garbage collection (GC) is required to erase and reclaim space with invalid data. However, GC poses additional writes that hinder the I/O performance, while SSD blocks can only endure a finite number of erasures. Thus, there is a performance-durability tradeoff on the design space of GC. To characterize the optimal tradeoff, this paper formulates an analytical model that explores the full optimal design space of any GC algorithm. We first present a stochastic Markov chain model that captures the I/O dynamics of large-scale SSDs, and adapt the mean-field approach to derive the asymptotic steady-state performance. We further prove the model convergence and generalize the model for all types of workload. Inspired by this model, we propose a randomized greedy algorithm (RGA) that can operate along the optimal tradeoff curve with a tunable parameter. Using trace-driven simulation on DiskSim with SSD add-ons, we demonstrate how RGA can be parameterized to realize the performance-durability tradeoff.

💡 Research Summary

The paper addresses the fundamental performance‑durability trade‑off inherent in solid‑state drive (SSD) garbage collection (GC). Because SSDs write out‑of‑place, invalid pages must be reclaimed by erasing whole blocks, which incurs extra writes (reducing performance) and consumes a limited number of block erasures (affecting durability). The authors develop a rigorous analytical framework that captures the full design space of any GC algorithm and identifies the optimal trade‑off curve.

System Modeling – An SSD is abstracted as N blocks, each containing k pages. A block’s state is defined by the number i (0 ≤ i ≤ k) of valid pages it holds; thus there are k + 1 possible states. The dynamics of program (write) and invalidate operations are modeled as a continuous‑time Markov chain (CTMC) where a block in state i moves to i + 1 with rate λ·(k − i)/k (program) and to i − 1 with rate λ·i/k (invalidate). Reads and GC itself do not change the block‑type counts. The global state of the SSD is the vector n(t) = (n₀(t), …, n_k(t)), where n_i(t) is the number of blocks of type i. The state space size (k + 1)^N is astronomically large for realistic SSDs (N≈10⁶).

Mean‑Field Approximation – To make the model tractable, the authors apply a mean‑field technique. As N → ∞, the empirical occupancy measure M_N(t) = (n₀/N, …, n_k/N) converges to a deterministic trajectory s(t) governed by a set of ordinary differential equations (ODEs). The ODEs (Eq. 5) describe the net flow into each state i based on the program/invalidate rates and the current fractions s_i(t). This reduction yields a low‑dimensional deterministic system that accurately approximates the stochastic dynamics for large SSDs.

Fixed‑Point Analysis – Setting the time derivatives to zero yields a unique steady‑state solution π_i = (k choose i)/2^k, i.e., a binomial distribution over the number of valid pages per block. This closed‑form fixed point enables analytical expressions for two key GC metrics:

1. Cleaning Cost – the expected number of valid pages that must be copied during a GC operation (proportional to the average i under π).
2. Wear‑Leveling – the variance of erase counts across blocks, which can be expressed as a function of π and the GC selection policy.

The authors identify the extremal points that minimize cleaning cost (favoring blocks with few valid pages) and maximize wear‑leveling (favoring uniform block selection). By introducing weighting parameters for the two objectives, they derive the full Pareto frontier – the optimal trade‑off curve between performance and durability.

Randomized Greedy Algorithm (RGA) – Guided by the analytical results, the paper proposes RGA, a simple yet tunable GC algorithm. RGA first builds a candidate set consisting of the α % of blocks with the smallest number of valid pages. It then selects a block from this set uniformly at random. A single tunable parameter θ ∈