{"title":"Tree-based Read-only Data Chunks for NVRAM Programming","authors":"Kumud Bhandari, Vivek Sarkar","doi":"10.1145/3292533.3292535","DOIUrl":null,"url":null,"abstract":"As the DRAM technology is fast reaching a scaling threshold, emerging non-volatile, byte-addressable memory (NVRAM) is expected to supplement and eventually replace DRAM. Future computing systems are anticipated to have a large amount of NVRAM, possibly spanning across more than one coherence domain. Furthermore, taking advantage of in-place persistence provided by the NVRAM in future systems requires a strategy to prevent tolerated failures (e.g. power failure) from leaving persistent data in an incoherent state. A fresh look at memory management approaches across the system stack is required to fully utilize future NVRAM. In this paper, we carefully assess the NVRAM-related memory access and management challenges, its implication to application level programming, and examine the suitability of tree-based read-only data chunks to NVRAM programming.","PeriodicalId":195082,"journal":{"name":"Proceedings of the Sixth Workshop on Data-Flow Execution Models for Extreme Scale Computing","volume":"35 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2016-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the Sixth Workshop on Data-Flow Execution Models for Extreme Scale Computing","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3292533.3292535","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
As the DRAM technology is fast reaching a scaling threshold, emerging non-volatile, byte-addressable memory (NVRAM) is expected to supplement and eventually replace DRAM. Future computing systems are anticipated to have a large amount of NVRAM, possibly spanning across more than one coherence domain. Furthermore, taking advantage of in-place persistence provided by the NVRAM in future systems requires a strategy to prevent tolerated failures (e.g. power failure) from leaving persistent data in an incoherent state. A fresh look at memory management approaches across the system stack is required to fully utilize future NVRAM. In this paper, we carefully assess the NVRAM-related memory access and management challenges, its implication to application level programming, and examine the suitability of tree-based read-only data chunks to NVRAM programming.