Based on a log-structured merge(LSM)tree,the key-value(KV)storage system can provide high reading performance and optimize random writing performance.It is widely used in modern data storage systems like e-commerce,on...Based on a log-structured merge(LSM)tree,the key-value(KV)storage system can provide high reading performance and optimize random writing performance.It is widely used in modern data storage systems like e-commerce,online analytics,and real-time communication.An LSM tree stores new KV data in the memory and flushes to disk in batches.To prevent data loss in memory if there is an unexpected crash,RocksDB appends updating data in the write-ahead log(WAL)before updating the memory.However,synchronous WAL significantly reduces writing performance.In this paper,we present a new WAL mechanism named MyWAL.It directly manages raw devices(or partitions)instead of saving data on a traditional file system.These can avoid useless metadata updating and write data sequentially on disks.Experimental results show that MyWAL can significantly improve the data writing performance of RocksDB compared to the traditional WAL for small KV data on solid-state disks(SSDs),as much as five to eight times faster.On non-volatile memory express soild-state drives(NVMe SSDs)and non-volatile memory(NVM),MyWAL can improve data writing performance by 10%–30%.Furthermore,the results of YCSB(Yahoo!Cloud Serving Benchmark)show that the latency decreased by 50%compared with SpanDB.展开更多
Wearable devices become popular because they can help people observe health condition.The battery life is the critical problem for wearable devices. The non-volatile memory(NVM) attracts attention in recent years beca...Wearable devices become popular because they can help people observe health condition.The battery life is the critical problem for wearable devices. The non-volatile memory(NVM) attracts attention in recent years because of its fast reading and writing speed, high density, persistence, and especially low idle power. With its low idle power consumption,NVM can be applied in wearable devices to prolong the battery lifetime such as smart bracelet. However, NVM has higher write power consumption than dynamic random access memory(DRAM). In this paper, we assume to use hybrid random access memory(RAM)and NVM architecture for the smart bracelet system.This paper presents a data management algorithm named bracelet power-aware data management(BPADM) based on the architecture. The BPADM can estimate the power consumption according to the memory access, such as sampling rate of data, and then determine the data should be stored in NVM or DRAM in order to satisfy low power. The experimental results show BPADM can reduce power consumption effectively for bracelet in normal and sleeping modes.展开更多
Extendible hashing is an effective way to manage increasingly large file system metadata,but it suffers from low concurrency and lack of optimization for non-volatile memory(NVM).In this paper,a multilevel hash direct...Extendible hashing is an effective way to manage increasingly large file system metadata,but it suffers from low concurrency and lack of optimization for non-volatile memory(NVM).In this paper,a multilevel hash directory based on lazy expansion is designed to improve the concurrency and efficiency of extendible hashing,and a hash bucket management algorithm based on groups is presented to improve the efficiency of hash key management by reducing the size of the hash bucket,thereby improving the performance of extendible hashing.Meanwhile,a hierarchical storage strategy of extendible hashing for NVM is given to take advantage of dynamic random access memory(DRAM)and NVM.Furthermore,on the basis of the device driver for Intel Optane DC Persistent Memory,the prototype of high-concurrency extendible hashing named NEHASH is implemented.Yahoo cloud serving benchmark(YCSB)is used to test and compare with CCEH,level hashing,and cuckoo hashing.The results show that NEHASH can improve read throughput by up to 16.5%and write throughput by 19.3%.展开更多
The future storage systems are expected to contain a wide variety of storage media and layers due to the rapid development of NVM(non-volatile memory)techniques.For NVM-based read caches,many kinds of NVM devices cann...The future storage systems are expected to contain a wide variety of storage media and layers due to the rapid development of NVM(non-volatile memory)techniques.For NVM-based read caches,many kinds of NVM devices cannot stand frequent data updates due to limited write endurance or high energy consumption of writing.However,traditional cache algorithms have to update cached blocks frequently because it is difficult for them to predict long-term popularity according to such limited information about data blocks,such as only a single value or a queue that reflects frequency or recency.In this paper,we propose a new MacroTrend(macroscopic trend)prediction method to discover long-term hot blocks through blocks'macro trends illustrated by their access count histograms.And then a new cache replacement algorithm is designed based on the MacroTrend prediction to greatly reduce the write amount while improving the hit ratio.We conduct extensive experiments driven by a series of real-world traces and find that compared with LRU,MacroTrend can reduce the write amounts of NVM cache devices significantly with similar hit ratios,leading to longer NVM lifetime or less energy consumption.展开更多
Non-Volatile Memory(NVM) offers byte-addressability and persistency. Because NVM can be plugged into memory and provide low latency, it offers a new opportunity to build new database systems with a single-layer storag...Non-Volatile Memory(NVM) offers byte-addressability and persistency. Because NVM can be plugged into memory and provide low latency, it offers a new opportunity to build new database systems with a single-layer storage design. A single-layer NVM-Native DataBase(N2 DB) provides zero copy and log freedom. Hence, all data are stored in NVM and there is no extra data duplication and logging during execution. N2 DB avoids complex data synchronization and logging overhead in the two-layer storage design of disk-oriented databases and in-memory databases. Garbage Collection(GC) is critical in such an NVM-based database because memory leaks on NVM are durable. Moreover, data recovery is equally essential to guarantee atomicity, consistency, isolation, and durability properties. Without logging, it is a great challenge for N2 DB to restore data to a consistent state after crashes and recoveries. This paper presents the GC and data recovery mechanisms for N2 DB. Evaluations show that the overall performance of N2 DB is up to 3:6 higher than that of InnoDB. Enabling GC reduces performance by up to 10%,but saves storage space by up to 67%. Moreover, our data recovery requires only 0:2% of the time and half of the storage space of InnoDB.展开更多
Multi-Clock Snapshot Isolation(MCSI)is a concurrency control mechanism that implements snapshot isolation on a single-layer Non-Volatile Memory(NVM)database.It stores a single copy of data by using multi-version stora...Multi-Clock Snapshot Isolation(MCSI)is a concurrency control mechanism that implements snapshot isolation on a single-layer Non-Volatile Memory(NVM)database.It stores a single copy of data by using multi-version storage to ensure durability and runtime access.With multi-clock transaction timestamp assignment,MCSI can efficiently generate snapshots with vector clocks and use per-thread transaction status arrays to identify uncommitted versions in NVM.For evaluation,we compared MCSI with the PostgreSQL-style concurrency control used in the single-layer NVM database N2DB.The maximum transaction throughput of MCSI is 101%–195%higher than that of N2DB for the YCSB workloads,and 25%–49%higher for the TPC-C workloads.Moreover,the transaction latency of MCSI remains relatively stable as the thread count increases.With 18 worker threads,the average transaction latency of MCSI is 65%–84%lower than that of N2DB for the YCSB workloads and 16%–43%lower for the TPC-C workloads.展开更多
This study proposes a new generation of floating gate transistors(FGT)with a novel built-in security feature.The new device has applications in guarding the IC chips against the current reverse engineering techniques,...This study proposes a new generation of floating gate transistors(FGT)with a novel built-in security feature.The new device has applications in guarding the IC chips against the current reverse engineering techniques,including scanning capacitance microscopy(SCM).The SCM measures the change in the C–V characteristic of the device as a result of placing a minute amount of charge on the floating gate,even in nano-meter scales.The proposed design only adds a simple processing step to the conventional FGT by adding an oppositely doped implanted layer to the substrate.This new structure was first analyzed theoretically and then a two-dimensional model was extracted to represent its C–V characteristic.Furthermore,this model was verified with a simulation.In addition,the C–V characteristics relevant to the SCM measurement of both conventional and the new designed FGT were compared to discuss the effectiveness of the added layer in masking the state of the transistor.The effect of change in doping concentration of the implanted layer on the C–V characteristics was also investigated.Finally,the feasibility of the proposed design was examined by comparing its I–V characteristics with the traditional FGT.展开更多
新型非易失性存储器(non-volatile memory,NVM)技术日渐成熟,延迟越来越低,带宽越来越高,未来将不仅有可能取代以动态随机存储器(dynamic random access memory,DRAM)为代表的易失型存储设备在主存中的垄断地位,还有可能取代传统Flash...新型非易失性存储器(non-volatile memory,NVM)技术日渐成熟,延迟越来越低,带宽越来越高,未来将不仅有可能取代以动态随机存储器(dynamic random access memory,DRAM)为代表的易失型存储设备在主存中的垄断地位,还有可能取代传统Flash和机械硬盘作为外存服务未来的计算机系统.如何综合各类新型存储的特性,设计高能效的存储架构,实现可应对大数据、云计算所需求的新型主存系统已经成为工业界和学术界的研究热点.提出基于高性能SOC FPGA阵列的NVM验证架构,互联多级FPGA,利用多层次FPGA结构扩展链接多片NVM.依据所提出的验证架构,设计了基于多层次FPGA的主从式NVM控制器,并完成适用于该架构的硬件原型设计.该架构不仅可以实现测试同类型多片NVM协同工作,也可以进行混合NVM存储管理方案验证.展开更多
随着大数据应用的涌现,计算机系统需要更大容量的内存以满足大数据处理的高时效性需求.新型非易失性存储器(non-volatile memory,NVM)结合传统动态随机存储器(dynamic random access memory,DRAM)组成的混合内存系统具有内存容量大、功...随着大数据应用的涌现,计算机系统需要更大容量的内存以满足大数据处理的高时效性需求.新型非易失性存储器(non-volatile memory,NVM)结合传统动态随机存储器(dynamic random access memory,DRAM)组成的混合内存系统具有内存容量大、功耗低的优势,因而得到了广泛关注.大数据应用同时也面临着旁路转换缓冲器(translation lookaside buffer,TLB)缺失率过高的性能瓶颈.大页可以有效降低TLB缺失率,然而,在混合内存中支持大页面临着大页迁移开销过大的问题.因此,设计了一种支持大页和大容量缓存的层次化混合内存系统:DRAM和NVM分别使用4KB和2MB粒度的页面分别进行管理,同时在DRAM和NVM之间实现直接映射.设计了基于访存频率的DRAM缓存数据过滤机制,减轻了带宽压力.提出了基于内存实时信息的动态热度阈值调整策略,灵活适应应用访存特征的变化.实验显示:与使用大页的全NVM内存系统和缓存热页(caching hot page,CHOP)系统相比平均有69.9%和15.2%的性能提升,而与使用大页的全DRAM内存系统相比平均只有8.8%的性能差距.展开更多
为了在数据密集型工作流下有效降低缓存碎片整理开销并提高缓存命中率,提出一种持久性分布式文件系统客户端缓存DFS-Cache(Distributed File System Cache)。DFS-Cache基于非易失性内存(NVM)设计实现,能够保证数据的持久性和崩溃一致性...为了在数据密集型工作流下有效降低缓存碎片整理开销并提高缓存命中率,提出一种持久性分布式文件系统客户端缓存DFS-Cache(Distributed File System Cache)。DFS-Cache基于非易失性内存(NVM)设计实现,能够保证数据的持久性和崩溃一致性,并大幅减少冷启动时间。DFS-Cache包括基于虚拟内存重映射的缓存碎片整理机制和基于生存时间(TTL)的缓存空间管理策略。前者基于NVM可被内存控制器直接寻址的特性,动态修改虚拟地址和物理地址之间的映射关系,实现零拷贝的内存碎片整理;后者是一种冷热分离的分组管理策略,借助重映射的缓存碎片整理机制,提升缓存空间的管理效率。实验采用真实的Intel傲腾持久性内存设备,对比商用的分布式文件系统MooseFS和GlusterFS,采用Fio和Filebench等标准测试程序,DFS-Cache最高能提升5.73倍和1.89倍的系统吞吐量。展开更多
基金Project supported by the National Key Research and Development Project of China(No.2022YFB2702101)the Shaanxi Province Key Industrial Projects,China(Nos.2021ZDLGY03-02 and 2021ZDLGY03-08)the National Natural Science Foundation of China(No.92152301)。
文摘Based on a log-structured merge(LSM)tree,the key-value(KV)storage system can provide high reading performance and optimize random writing performance.It is widely used in modern data storage systems like e-commerce,online analytics,and real-time communication.An LSM tree stores new KV data in the memory and flushes to disk in batches.To prevent data loss in memory if there is an unexpected crash,RocksDB appends updating data in the write-ahead log(WAL)before updating the memory.However,synchronous WAL significantly reduces writing performance.In this paper,we present a new WAL mechanism named MyWAL.It directly manages raw devices(or partitions)instead of saving data on a traditional file system.These can avoid useless metadata updating and write data sequentially on disks.Experimental results show that MyWAL can significantly improve the data writing performance of RocksDB compared to the traditional WAL for small KV data on solid-state disks(SSDs),as much as five to eight times faster.On non-volatile memory express soild-state drives(NVMe SSDs)and non-volatile memory(NVM),MyWAL can improve data writing performance by 10%–30%.Furthermore,the results of YCSB(Yahoo!Cloud Serving Benchmark)show that the latency decreased by 50%compared with SpanDB.
基金supported by the Research Fund of National Key Laboratory of Computer Architecture under Grant No.CARCH201501the Open Project Program of the State Key Laboratory of Mathematical Engineering and Advanced Computing under Grant No.2016A09
文摘Wearable devices become popular because they can help people observe health condition.The battery life is the critical problem for wearable devices. The non-volatile memory(NVM) attracts attention in recent years because of its fast reading and writing speed, high density, persistence, and especially low idle power. With its low idle power consumption,NVM can be applied in wearable devices to prolong the battery lifetime such as smart bracelet. However, NVM has higher write power consumption than dynamic random access memory(DRAM). In this paper, we assume to use hybrid random access memory(RAM)and NVM architecture for the smart bracelet system.This paper presents a data management algorithm named bracelet power-aware data management(BPADM) based on the architecture. The BPADM can estimate the power consumption according to the memory access, such as sampling rate of data, and then determine the data should be stored in NVM or DRAM in order to satisfy low power. The experimental results show BPADM can reduce power consumption effectively for bracelet in normal and sleeping modes.
基金Project supported by the National Natural Science Foundation of China(No.61806086)the National Key R&D Program of China(No.2018YFB0804204)。
文摘Extendible hashing is an effective way to manage increasingly large file system metadata,but it suffers from low concurrency and lack of optimization for non-volatile memory(NVM).In this paper,a multilevel hash directory based on lazy expansion is designed to improve the concurrency and efficiency of extendible hashing,and a hash bucket management algorithm based on groups is presented to improve the efficiency of hash key management by reducing the size of the hash bucket,thereby improving the performance of extendible hashing.Meanwhile,a hierarchical storage strategy of extendible hashing for NVM is given to take advantage of dynamic random access memory(DRAM)and NVM.Furthermore,on the basis of the device driver for Intel Optane DC Persistent Memory,the prototype of high-concurrency extendible hashing named NEHASH is implemented.Yahoo cloud serving benchmark(YCSB)is used to test and compare with CCEH,level hashing,and cuckoo hashing.The results show that NEHASH can improve read throughput by up to 16.5%and write throughput by 19.3%.
基金supported by the National Key Research and Development Program of China under Grant No.2019YFE0198600the National Natural Science Foundation of China under Grant Nos.61972402,61972275,and 61732014.
文摘The future storage systems are expected to contain a wide variety of storage media and layers due to the rapid development of NVM(non-volatile memory)techniques.For NVM-based read caches,many kinds of NVM devices cannot stand frequent data updates due to limited write endurance or high energy consumption of writing.However,traditional cache algorithms have to update cached blocks frequently because it is difficult for them to predict long-term popularity according to such limited information about data blocks,such as only a single value or a queue that reflects frequency or recency.In this paper,we propose a new MacroTrend(macroscopic trend)prediction method to discover long-term hot blocks through blocks'macro trends illustrated by their access count histograms.And then a new cache replacement algorithm is designed based on the MacroTrend prediction to greatly reduce the write amount while improving the hit ratio.We conduct extensive experiments driven by a series of real-world traces and find that compared with LRU,MacroTrend can reduce the write amounts of NVM cache devices significantly with similar hit ratios,leading to longer NVM lifetime or less energy consumption.
基金supported by the National Key Research & Development Program of China (No. 2016YFB1000504)the National Natural Science Foundation of China (Nos. 61877035, 61433008, 61373145, and 61572280)。
文摘Non-Volatile Memory(NVM) offers byte-addressability and persistency. Because NVM can be plugged into memory and provide low latency, it offers a new opportunity to build new database systems with a single-layer storage design. A single-layer NVM-Native DataBase(N2 DB) provides zero copy and log freedom. Hence, all data are stored in NVM and there is no extra data duplication and logging during execution. N2 DB avoids complex data synchronization and logging overhead in the two-layer storage design of disk-oriented databases and in-memory databases. Garbage Collection(GC) is critical in such an NVM-based database because memory leaks on NVM are durable. Moreover, data recovery is equally essential to guarantee atomicity, consistency, isolation, and durability properties. Without logging, it is a great challenge for N2 DB to restore data to a consistent state after crashes and recoveries. This paper presents the GC and data recovery mechanisms for N2 DB. Evaluations show that the overall performance of N2 DB is up to 3:6 higher than that of InnoDB. Enabling GC reduces performance by up to 10%,but saves storage space by up to 67%. Moreover, our data recovery requires only 0:2% of the time and half of the storage space of InnoDB.
基金supported by the National Key Research&Development Program of China(No.2016YFB1000504)the National Natural Science Foundation of China(Nos.61877035,61433008,61373145,and 61572280).
文摘Multi-Clock Snapshot Isolation(MCSI)is a concurrency control mechanism that implements snapshot isolation on a single-layer Non-Volatile Memory(NVM)database.It stores a single copy of data by using multi-version storage to ensure durability and runtime access.With multi-clock transaction timestamp assignment,MCSI can efficiently generate snapshots with vector clocks and use per-thread transaction status arrays to identify uncommitted versions in NVM.For evaluation,we compared MCSI with the PostgreSQL-style concurrency control used in the single-layer NVM database N2DB.The maximum transaction throughput of MCSI is 101%–195%higher than that of N2DB for the YCSB workloads,and 25%–49%higher for the TPC-C workloads.Moreover,the transaction latency of MCSI remains relatively stable as the thread count increases.With 18 worker threads,the average transaction latency of MCSI is 65%–84%lower than that of N2DB for the YCSB workloads and 16%–43%lower for the TPC-C workloads.
文摘This study proposes a new generation of floating gate transistors(FGT)with a novel built-in security feature.The new device has applications in guarding the IC chips against the current reverse engineering techniques,including scanning capacitance microscopy(SCM).The SCM measures the change in the C–V characteristic of the device as a result of placing a minute amount of charge on the floating gate,even in nano-meter scales.The proposed design only adds a simple processing step to the conventional FGT by adding an oppositely doped implanted layer to the substrate.This new structure was first analyzed theoretically and then a two-dimensional model was extracted to represent its C–V characteristic.Furthermore,this model was verified with a simulation.In addition,the C–V characteristics relevant to the SCM measurement of both conventional and the new designed FGT were compared to discuss the effectiveness of the added layer in masking the state of the transistor.The effect of change in doping concentration of the implanted layer on the C–V characteristics was also investigated.Finally,the feasibility of the proposed design was examined by comparing its I–V characteristics with the traditional FGT.
文摘随着大数据应用的涌现,计算机系统需要更大容量的内存以满足大数据处理的高时效性需求.新型非易失性存储器(non-volatile memory,NVM)结合传统动态随机存储器(dynamic random access memory,DRAM)组成的混合内存系统具有内存容量大、功耗低的优势,因而得到了广泛关注.大数据应用同时也面临着旁路转换缓冲器(translation lookaside buffer,TLB)缺失率过高的性能瓶颈.大页可以有效降低TLB缺失率,然而,在混合内存中支持大页面临着大页迁移开销过大的问题.因此,设计了一种支持大页和大容量缓存的层次化混合内存系统:DRAM和NVM分别使用4KB和2MB粒度的页面分别进行管理,同时在DRAM和NVM之间实现直接映射.设计了基于访存频率的DRAM缓存数据过滤机制,减轻了带宽压力.提出了基于内存实时信息的动态热度阈值调整策略,灵活适应应用访存特征的变化.实验显示:与使用大页的全NVM内存系统和缓存热页(caching hot page,CHOP)系统相比平均有69.9%和15.2%的性能提升,而与使用大页的全DRAM内存系统相比平均只有8.8%的性能差距.
文摘为了在数据密集型工作流下有效降低缓存碎片整理开销并提高缓存命中率,提出一种持久性分布式文件系统客户端缓存DFS-Cache(Distributed File System Cache)。DFS-Cache基于非易失性内存(NVM)设计实现,能够保证数据的持久性和崩溃一致性,并大幅减少冷启动时间。DFS-Cache包括基于虚拟内存重映射的缓存碎片整理机制和基于生存时间(TTL)的缓存空间管理策略。前者基于NVM可被内存控制器直接寻址的特性,动态修改虚拟地址和物理地址之间的映射关系,实现零拷贝的内存碎片整理;后者是一种冷热分离的分组管理策略,借助重映射的缓存碎片整理机制,提升缓存空间的管理效率。实验采用真实的Intel傲腾持久性内存设备,对比商用的分布式文件系统MooseFS和GlusterFS,采用Fio和Filebench等标准测试程序,DFS-Cache最高能提升5.73倍和1.89倍的系统吞吐量。
