MuLe-RAID:面向大容量高性能SSD的层次化RAID

利用RAID机制构建基于闪存的SSD可以提高SSD的性能和可靠性,但是随着SSD容量的增大,传统的单层RAID结构容易造成控制器的瓶颈.针对这个问题,提出了一种新的RAID策略——MuLe-RAID(Multiple Level RAID),给出了详细架构和设计细节.其采用层次化RAID结构,设置多个控制器,在每一层的控制器中实现不同粒度的损耗均衡管理,在保证SSD损耗均衡的同时,通过减轻控制器的瓶颈提高了SSD的性能.为了提高可靠性,MuLe-RAID采用了冗余机制,能够达到RAID5的容错级别.通过模拟实验测试表明,在大容量SSD中,相比传统RAID,MuLe-RAID在保证损耗均衡的基础上有大约30%的性能提升.     

一种用磁盘备份SSD的高性能可靠存储系统

固态盘(solid state drive,SSD)因为其优越的性能已被大量部署于当前的存储系统中.但是,由于寿命有限,SSD的可靠性受到广泛的质疑.磁盘阵列(redundant arrays of inexpensive disk,RAID)是一种传统的用来提高可靠性的手段,但并不适用于SSD.这项工作提出一种基于SSD和磁盘的混合存储系统,构建该系统的主要思想是SSD响应所有I/O请求,从而获得较高的性能;磁盘备份所有数据,从而保证系统的可靠性.但是,磁盘的I/O性能显著低于SSD,构建该系统的问题在于磁盘能否及时地备份SSD上的数据.为了解决这一问题,从两方面提出优化:在延迟方面,采用非易失主存弥补磁盘与SSD的延迟差距;在带宽方面,采用两种措施:1)在单块磁盘内部重组I/O请求,使磁盘尽可能的顺序读写;2)采用多块磁盘备份多块SSD,通过将一块SSD上的写请求分散到多块磁盘上,有效应对单块SSD上出现的突发写请求.通过原型系统实现表明,该混合系统是可行的:磁盘能够为SSD提供实时的数据备份;与其他系统相比,该混合系统取得较高的性价比.     

OpenCL计算软件栈评估

随着智能计算和大数据应用的发展,人们对GPU等加速部件的需求不断增长.计算软件栈比如CUDA、OpenCL软件栈是能充分发挥GPU硬件性能的关键.考虑计算软件栈未来在国产基础软硬件平台(比如飞腾CPU和麒麟操作系统)上的可移植性和适配性,重点研究OpenCL开源计算软件栈.测试分析OpenCL应用在不同平台上的表现,评估应用在不同OpenCL软件栈上(比如Mesa、ROCm等)进行GPU计算的表现,评估软件栈中驱动、内核等对GPU计算的影响,并且整个测试涵盖了编译、数据传输和内核执行等OpenCL计算各个阶段的时间开销.经过测试评估发现,国产平台更迫切也更适合使用GPU进行加速计算,ROCm是比较理想的OpenCL开源软件栈,有较好的性能和稳定性,并且与闭源软件栈相比存在一定的优化空间.     

Reorder Write Sequence by Hetero-Buffer to Extend SSD’s Lifespan

The limited lifespan is the Achilles’ heel of solid state drives(SSDs) based on NAND flash.NAND flash has two drawbacks that degrade SSDs’ lifespan.One is the out-of-place update.Another is the sequential write constraint within a block.SSDs usually employ write buffer to extend their lifetime.However,existing write buffer schemes only pay attention to the first drawback,while neglect the second one.We propose a hetero-buffer architecture covering both aspects simultaneously.The hetero-buffer consists of two components,dynamic random access memory(DRAM) and the reorder area.DRAM endeavors to reduce write traffic as much as possible by pursuing a higher hit ratio(overcome the first drawback).The reorder area focuses on reordering write sequence(overcome the second drawback).Our hetero-buffer outperforms traditional write buffers because of two reasons.First,the DRAM can adopt existing superior cache replacement policy,thus achieves higher hit ratio.Second,the hetero-buffer reorders the write sequence,which has not been exploited by traditional write buffers.Besides the optimizations mentioned above,our hetero-buffer considers the work environment of write buffer,which is also neglected by traditional write buffers.By this way,the hetero-buffer is further improved.The performance is evaluated via trace-driven simulations.Experimental results show that,SSDs employing the hetero-buffer survive longer lifespan on most workloads.     

基于决策知识学习的多无人机航迹协同规划

考虑无人机群体行为决策与状态变化的内在驱动, 从信息处理角度提出基于决策知识学习的多无人机航迹协同规划方法. 首先, 基于马尔科夫决策过程对无人机的行为状态进行知识表示, 形成关于连续动作空间的决策知识; 然后, 提出基于知识决策学习的深度确定性策略梯度算法, 实现无人机在决策知识层次上的协同规划. 实验结果表明: 在研发设计演示系统的基础上, 所提方法通过强化学习能够得到一个最优航迹规划策略, 同时使航迹综合评价和平均奖励收敛稳定, 为无人机任务执行提供了决策支持.     

CSWL: Cross-SSD Wear-Leveling Method in SSD-Based RAID Systems for System Endurance and Performance

Flash memory has limited erasure/program cycles.Hence,to meet their advertised capacity all the time,flashbased solid state drives(SSDs) must prolong their life span through a wear-leveling mechanism.As a very important part of flash translation layer(FTL),wear leveling is usually implemented in SSD controllers,which is called internal wear leveling.However,there is no wear leveling among SSDs in SSD-based redundant array of independent disks(RAIDs) systems,making some SSDs wear out faster than others.Once an SSD fails,reconstruction must be triggered immediately,but the cost of this process is so high that both system reliability and availability are affected seriously.We therefore propose cross-SSD wear leveling(CSWL) to enhance the endurance of entire SSD-based RAID systems.Under the workload of random access pattern,parity stripes suffer from much more updates because updating to a data stripe will cause the modification of other all related parity stripes.Based on this principle,we introduce an age-driven parity distribution scheme to guarantee wear leveling among flash SSDs and thereby prolong the endurance of RAID systems.Furthermore,age-driven parity distribution benefits performance by maintaining better load balance.With insignificant overhead,CSWL can significantly improve both the life span and performance of SSD-based RAID.