Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Memory/Storage

Non-linear cost compression algorithm for MLC memory

Posted: 06 Nov 2013 ?? ?Print Version ?Bookmark and Share

Keywords:MLC memory? compression algorithm? SLC?

A team from Hanyang University has succeeded in developing what they describe as an efficient non-linear cost compression algorithm for multi-level cell (MLC) memory. MLC memory, which is used for high capacity applications, doubles the capacity for the same underlying semiconductor technology by programming two bits in a cell. The main characteristic of MLC memory is its low price, but it has suffered from longer latency, higher energy consumption and lower durability, compared with a single level cell (SLC) memory. For example, the energy consumption of writing '01' or '10', with the cost of 72.64 and 79.01 respectively, is much larger than that of writing '00' or '11', with the cost of 13.37 and 0.00 respectively.

With the aim of solving the non-linear cost compression problem efficiently, Hyun-ok Oh and his research team proposed an encoding symbol frequency based approach using an algorithm. The algorithm was composed of two steps: computing frequencies of encoding symbols to minimise cost function and deploying existing size-decompression algorithms to achieve the computed frequencies of a cost-compressed message.

Using this algorithm, Oh demonstrated an energy/latency minimisation framework for MLC memory that consists of size compression and cost compression. At the size compression stage, any size compression algorithm can be adopted such as Huffman coding or arithmetic coding, regardless of whether it is adaptive or non-adaptive. For the size-compressed message, cost compression algorithm or a size decompression algorithm is applied with computed probability values. In order to produce a decompressed message with a certain probability, Oh adopted an entropy-based compression algorithm.

Oh's experiment was conducted on six different types: text type, a library tape, executable type, PDF type, JPEG type and video type. The results revealed an increase in energy savings and performance compared with the proposed cost compression approach with the size compression algorithm and a digital majority voter (DMV) for 1KB page size. For uncompressed data such as text files, the proposed algorithm can save energy consumption by up to 72 per cent and improve performance by 70 per cent while the size compression algorithm can reduce energy consumption by 43 per cent. For library and executive files, the improvements can be 50-60 per cent while they are less than 17 per cent of the size compression algorithm.

According to Oh, while the cost compression algorithm can improve energy efficiency and performance more than the size compression algorithm and DMV for uncompressed data, DMV is better than the cost compression algorithm for compressed data. Therefore, the best results can be obtained by combining the cost compression and DMV. The cost compression algorithm should be applied first, and then the cost compressed bit patterns should be exchanged if the numbers '01' and '10' are larger than that of '00' and '11' by DMV. The experiment went further to show the improvement of the file of the files by the size compression and the cost compression. For text, library, executable and PDF files, the size compression algorithm can increase its life span by 63 per cent and the cost compression algorithm by 74 per cent on average.

The fact that Oh's research found ways to reduce energy consumption and latency while improving the life of memories can be considered a groundbreaking event. These results can be applied to SSD, Morse code and hard disks, to name a few. Oh's finding has been patented and is expected to be applied to our daily lives. "Cost compression problems have been a mystery since the 1960s. To solve this problem, I tried to detect its fundamental characteristics from within. Using the Huffman encoding tree method, my research team succeeded in discovering the cost effective method," said Oh.

Meanwhile, he has also been endeavoring to develop non-volatile memory in terms of its energy use and intelligent SSD. He is looking forward to engage the field of crypto and security as well. "So far, most of my research can be applied to everyday use and I hope people can find comfort in using them with ease. I will continue to develop my career to contribute to people's happiness in the near future," added Oh.

Article Comments - Non-linear cost compression algorith...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top