A content-aware writing mechanism for reducing energy on non-volatile memory based embedded storage systems

Abstract

Emerging byte-addressable non-volatile memory (NVM), such as phase change memory has many advantages in terms of low power consumption, high density, in-place update, and higher write endurance. Reducing the energy consumption of NVM is a fundamental issue when the technology is considered as an alternative to secondary storage in embedded systems. In the design of energy efficiency approaches, one of the most challenging problems is to reduce unnecessary write activities on NVM. In this paper, we present an optimal content-aware NVM memory writing mechanism, CA-NVM, to effectively select a free NVM page for each write request from file system to reduce write activities, since write activities account for most of energy consumption in NVM storage. In order to speed up the procedure of selecting a free page, we introduce a sampling mechanism, by which we only pick a few bits from the free page and the logical page to be written for comparison. Those sample bits of free pages are stored in a special area such that they can be indexed quickly. We conduct a series of experiments based on realistic traces, and the experimental results show that the proposed technique can reduce write activities up to 36.03 %, and reduce energy consumption by 21.3 % on average compared with a random based free-page allocation mechanism.

Non-Volatile Memory File Systems: A Survey

“…Another interesting method to reduce write latency is the read-before-write technique [86]. In this technique, before a page of data is updated, the bitwise difference between the old and the new data is calculated.…”

“…Both main memory and storage are among the main sources of energy consumption and therefore are good candidates for optimization. On the other hand, embedded and mobile systems based on NVM, with limited or no access to long-term power source may also benefit from a smarter more efficient usage of power [86].…”

“…PM is a good candidate to store read-intensive data as reading is cheaper in terms of both energy and time and also because PM do not need periodic refreshes like DRAM. Copy-beforewrite can also be used to change the minimum number of bits in a page [86]. The algorithm is based on search among the pages stored in the device for pages similar to the one being written.…”