Forecasting lifetime and performance of a novel NVM last-level cache with compression

TL;DR

This paper introduces a novel NVM last-level cache organization using data compression and wear-leveling, enabling accurate lifetime and performance forecasting through detailed simulation and prediction methods.

Contribution

It proposes a new NV-LLC architecture with compression and wear-leveling, along with a forecasting procedure for detailed temporal analysis of cache degradation.

Findings

Compression reduces cache block size and delays wear-out.

The proposed method accurately predicts cache lifetime and performance metrics.

The new NV-LLC design has minimal additional cost but significantly extends effective capacity over time.

Abstract

Non-volatile memory (NVM) technologies are interesting alternatives for building the on-chip Last-Level Cache (LLC). Their advantages, compared to SRAM memory, are higher density and lower static power, but each write operation slightly wears out the bitcell, to the point of losing its storage capacity. In this context, this paper proposes a novel NV-LLC organization leveraging data compression. Data compression reduces the size of the blocks and, together with wear-leveling mechanisms, can defer the degradation of such a NV-LLC. Moreover, as capacity is reduced by write wear, data compression enables degraded cache frames to allocate blocks whose compressed size is adequate. From a methodological point of view, although different approaches are used in the literature to analyze the degradation of a NV-LLC, none of them allows to study in detail its temporal evolution. In this sense, this work proposes a forecasting procedure that combines detailed simulation and prediction, enabling an accurate analysis of different cache content mechanisms (replacement, wear leveling, compression, etc.) on the temporal evolution of the indices of interest, such as the effective capacity of the NV-LLC or the system IPC. The proposed NV-LLC organization has a small added cost compared to that of a baseline NV-LLC without compression in terms of area, latency and energy consumption, and increases more than 6 times the time a NV-LLC takes to reach 50% effective capacity.