8.2 Flash Memories & Numerical Approximation

Printer-friendly version PDF version

Date: Wednesday 11 March 2015
Time: 17:00 - 18:30
Location / Room: Belle Etoile

Philippe Coussy, Universite de Bretagne-Sud, FR

Zili Shao, HongKong Polytechnic University, HK

This session presents papers on flash memories and numerical approximations. The first paper presents a new flash memory management scheme to extend the product lifetime of SSDs. The second paper describes a hardware accelerator approach for solving linear equations. The third paper describes an approach for automatically synthesizing non-linear 2D function approximations without requiring costly, and power-hungry, multipliers. Finally, the session ends with two interactive presentations that address progressive wear-leveling for flash memories, and security threats and countermeasures for solid-state drives.

TimeLabelPresentation Title
Han-Yi Lin1 and Jen-Wei Hsieh2
1National Taiwan University, TW; 2National Taiwan University of Science and Technology, TW
In recent years, flash memory has been widely used in embedded systems, portable devices, and high-performance storage products due to its non-volatility, shock resistance, low power consumption, and high performance natures. To reduce the product cost, multi-level-cell flash memory (MLC) has been proposed; compared with the traditional single-level-cell flash memory (SLC) that only stores one bit of data per cell, each MLC cell can store two or more bits of data. Thus MLC can achieve a larger capacity and reduce the cost per unit. However, MLC also suffers from the degradation in both performance and reliability. In this paper, we try to enhance the reliability and reduce the product cost of flash-memory based storage devices from a totally different perspective. We propose a half-level-cell (HLC) management scheme to manage and reuse the worn-out space in solid-state drives (SSDs); through our management scheme, the system can treat two corrupted pages as a normal page without sacrificing performance and reliability. To the best of our knowledge, this is the first research that reclaims free space by reviving the corrupted pages. The experiment results show that the lifetime of SSD can be extended by 48.54% for the trace of general users applications with our proposed HLC management scheme.

Download Paper (PDF; Only available from the DATE venue WiFi)
Ebrahim M. Songhori, Xuyang Lu and Farinaz Koushanfar, Rice University, US
This paper introduces AHEAD, a novel domain-specific framework for automated (hardware-based) acceleration of massive data analysis applications with a dense (non-sparse) correlation matrix. Due to non-scalability of matrix inversion, often iterative computation is used for converging to a solution. AHEAD addresses two sets of domain-specific matrix computation challenges. First, the I/O and memory bandwidth constraints which limit the performance of hardware accelerators. Second, the hardness of handling large data because of the complexity of the known matrix transformations and the inseparability of non-sparse correlations. The inseparability problem translates to an increased communication cost with the accelerators. To optimize the performance within these limits, AHEAD learns the dependency structure of the domain data and suggests a scalable matrix transformation. The transformation minimizes the memory access required for matrix computing within an error threshold and thus, optimizes the mapping of domain data to the available (bandwidth constrained) accelerator resources. To facilitate automation, AHEAD also provides an Application Programming Interface (API) so users can customize the framework to an arbitrary iterative analysis algorithm and hardware mapping. Proof-of-concept implementation of AHEAD is performed on the widely used compressive sensing and general l1 regularized least squares solvers. On a massive light field imaging data set with 4.6B non-zeros, AHEAD attains up to 320x iteration speed improvement using reconfigurable hardware accelerators compared with the conventional solver and about 4x improvement compared to our transformed matrix solver on a general purpose processor (without hardware acceleration).

Download Paper (PDF; Only available from the DATE venue WiFi)
Jochen Rust and Steffen Paul, University of Bremen, DE
In this paper a novel method for hardware-based realization of two-variable numeric functions is introduced. The main idea is based on the extension of the well-known piecewise linear approximation technique, which is often used for the calculation of one-variable elementary functions. A non-uniform and plane segmentation scheme enables quick segment access at runtime; the use of multiplier-less linear equations causes high performance in terms of throughput. As both the extraction of approximation-related parameters and its mapping to corresponding hardware elements is automated, the design time is also reduced to a minimum. For evaluation, several approximations with varying constraints are generated and compared on the algorithmic level to one another as well as to actual references. In conjunction with the results of logical and physical CMOS synthesis, our work turns out to be highly efficient in terms of throughput, memory requirements and energy consumption.

Download Paper (PDF; Only available from the DATE venue WiFi)
Fu-Hsin Chen1, Ming-Chang Yang2, Yuan-Hao Chang3 and Tei-Wei Kuo4
1Department of Computer Science and Information Engineering, National Taiwan University, TW; 2Graduate Institute of Networking and Multimedia, National Taiwan University, TW; 3Institute of Information Science, Academia Sinica, TW; 4Academia Sinica & National Taiwan University, TW
As the endurance of flash memory keeps deteriorating, exploiting wear leveling techniques to improve the lifetime/endurance of flash memory has become a critical issue in the design of flash storage devices. In contrast to existing wear-leveling techniques that aggressively distributes the erases to all flash blocks by a fixed threshold, we propose a progressive wear leveling design to perform wear leveling in a "progressive" way to prevent any block from being worn out prematurely, and thereby to ultimately minimize the performance overheads caused by the unnecessary data migration. The results reveal that, instead of sacrificing the device lifetime, performing wear leveling in such a progressive way can not only minimize the performance overheads but even have potentials to extend the device lifespan.

Download Paper (PDF; Only available from the DATE venue WiFi)
Xiaotong Cui1, Minhui Zou1, Liang Shi2 and Kaijie Wu1
1Chongqing University, CN; 2College of Computer Science, Chongqing University, CN
In recent years, we have seen an increasing deployment of flash-based storage, such as SSD, in mission-critical applications due to its fast read/write speed, small form factor, strong shock resistance, and etc. SSD uses a host interface and a middle layer called flash translation layer (FTL) to maintain the compatibility with the traditional magnetic-based HDD. Unlike the traditional HDD where the host OS has the full control on where to access the data, SSD uses FTL to translate and implement all operations, and OS has no such control. Even worse, FTL, which is considered as one of most important intellectual property of SSD, is often proprietary. This brings up a security concern on design trustworthiness: what if the manufacturer either accidentally or intentionally implement those operations incorrectly or even maliciously? In this paper we analyze the possible threats and propose a simple yet effective countermeasure.

Download Paper (PDF; Only available from the DATE venue WiFi)
18:30End of session
19:30DATE Party in Museum of Grenoble (Musée de Grenoble, 5 Place de Lavalette, 38000 Grenoble, France)

Musée de Grenoble

As one of the main networking opportunities during the DATE week, the DATE Party states a perfect occasion to meet friends and colleagues in a relaxed atmosphere while enjoying local amenities. It will take place on March 11, 2015, from 19:30 to 23:00 in the renowned "Musée de Grenoble" (Grenoble Museum). This painting museum features a unique collection of ancient, modern and contemporary art including major masterpieces of classical Flemish, Dutch, Italian and Spanish painting and all the great pot-1945 contemporary art-trends, right up to the most recent artwork of the 2000s.

During this evening, you can enjoy the famous French Cuisine and outstanding wines. Discover the region of the French Alps through ist cheese and wine specialties. The dinner will be accompanied by jazz songs and instrumental music from Anna Cruz and her vocal band. Another highlight will be the show waders "THE INSEPARABLES", sweet and ephemeral characters walking through the premises, releasing dreams and laughter. Furthermore, at the very beginning of the evening, from 20h00 to 21h30, you will have the opportunity to visit parts of the permanent collection of the museum (ninetieth and twenties century).

Please kindly note that it is not a seated dinner.

All delegates, exhibitors and their guests are invited to attend the party. Please be aware that entrance is only possible with a valid party ticket. Each full conference registration includes a ticket for the DATE Party (which needs to be booked during the online registration process though). Additional tickets can be purchased on-site at the registration desk (subject to availability of tickets). Price for extra ticket: 60 € per person.

How to get there: The tram B has a stop called "Notre Dame Musee". That stop is next to the Museum. Attendees would take the tram A from Alpexpo and change for Tram B in one of the stations between "Gares" and "Maison du Tourisme" to get to the museum. The trip takes about 30 minutes.