Seminar: ceClub: The Technion Computer Engineering Club

Space-efficient FTL for Mobile Storage via Tiny Neural Nets

Date: February,14,2024 Start Time: 11:30 - 12:30
Location: 861, Meyer Building
Add to:
Lecturer: Ron Marcus
With the rapid increase of storage demands and working sets of modern mobile apps, maintaining high I/O performance in mobile SSDs under strict resource constraints is challenging. The Flash Translation Layer (FTL) must increase the capacity of the Logical-To-Physical (L2P) address translation cache to keep up with the new workloads, but it comes at the cost of scaling the on-die SRAM, resulting in higher chip area, power consumption, and costs.

In this talk, I will present RQFTL, a demand-based FTL for mobile storage controllers that boosts the effective cache capacity over state-of-the-art techniques. RQFTL stores a large part of the L2P cache in a compressed form, and employs a learned data structure called RQRMI that leverages tiny neural nets to quickly find the correct translation entry in the cache. RQFTL uses neural network inference for cache lookups, and rapidly retrains the neural nets to efficiently handle L2P cache updates. It is specifically optimized to achieve high coverage for scattered read accesses, making it suitable for popular read-skewed workloads such as mobile gaming.

The talk includes an evaluation of RQFTL on hours-long real-world I/O traces of popular modern mobile apps including games, video editing and social networking apps collected on Google Pixel V6 Phone. It shows that RQFTL outperforms all the state-of-the-art FTLs in these workloads, increasing the effective L2P cache capacity by over an order of magnitude compared to DFTL and up to 5X over the recent LeaFTL. As a result, it achieves 2X and 1.42X higher hit rate compared to DFTL and LeaFTL respectively, under the same SRAM capacity, and allows reduction of the total SRAM capacity of a controller by about a third of that of LeaFTL.

M.Sc. student under the supervision of Prof. Mark Silberstein.


All Seminars
Skip to content