An Energy Management For Hybrid Memory Based On Write Frequency Of Pages | Posted on:2016-04-13 | Degree:Master | Type:Thesis | Country:China | Candidate:J B Zhang | Full Text:PDF | GTID:2348330479453430 | Subject:Computer application technology | Abstract/Summary: | PDF Full Text Request | The rapid development of computer science and advent of big data era have given rise to a series of data intensive applications. These applications raise a huge requirement on the performance and capacity of memory. However, an obvious increase of memory makes the corresponding energy consumption unacceptable in practice. To address this question, any effort only to reduce the energy consumption of dynamic random access memory(DRAM) is ineffective. An in-memory computing architecture based on hybrid memory emerged and the hybrid memory consists of DRAM and phase change memory(PCM). This architecture intends to benefit the advantages of DRAM and PCM, increase the capacity of memory and reduce the power of memory. It has become the focus of industry and academia in recent years.But stacking two memory simply cannot achieve the goal. Writing on PCM frequently will cost much power and reduce the performance of main memory. Therefore an effective memory management is needed to reduce the write times of PCM and the energy consumption of memory. Previous methods to reduce the energy consumption of hybrid memory are based on little information of pages during a short time. These methods are not accurate and reasonable, and the effect on reducing energy consumption is limited as well.Energy management for hybrid memory based on write frequency of pages(EMHMBWFP) makes a comprehensive mechanism according to global and local access information so that pages can be placed to the reasonable storage and the energy consumption of main memory can be reduced. EMHMBWFP records global and local access information and calculates the write frequency according to these information. Then EMHMBWFP classifies pages according to the write frequency and estimates the saving energy through the energy consumption model. At last the best group of migration pages will be found and the whole page migration strategy is finished.Experimental results show that EMHMBWFP can achieve 9.4% of energy saving and 9.6% of performance improvement at most compared with APG and PDRAM. | Keywords/Search Tags: | Hybrid Memory, Phase Change Memory, PCM, Energy, Migration | PDF Full Text Request | Related items |
| |
|