Youngjin Yu, Dongin Shin, Woong Shin, N. Song, Jae-Woo Choi, H. Kim, Hyeonsang Eom, H. Yeom
Fast storage devices are an emerging solution to satisfy data-intensive applications. They provide high transaction rates for DBMS, low response times for Web servers, instant on-demand paging for applications with large memory footprints, and many similar advantages for performance-hungry applications. In spite of the benefits promised by fast hardware, modern operating systems are not yet structured to take advantage of the hardware’s full potential. The software overhead caused by an OS, negligible in the past, adversely impacts application performance, lessening the advantage of using such hardware. Our analysis demonstrates that the overheads from the traditional storage-stack design are significant and cannot easily be overcome without modifying the hardware interface and adding new capabilities to the operating system. In this article, we propose six optimizations that enable an OS to fully exploit the performance characteristics of fast storage devices. With the support of new hardware interfaces, our optimizations minimize per-request latency by streamlining the I/O path and amortize per-request latency by maximizing parallelism inside the device. We demonstrate the impact on application performance through well-known storage benchmarks run against a Linux kernel with a customized SSD. We find that eliminating context switches in the I/O path decreases the software overhead of an I/O request from 20 microseconds to 5 microseconds and a new request merge scheme called Temporal Merge enables the OS to achieve 87% to 100% of peak device performance, regardless of request access patterns or types. Although the performance improvement by these optimizations on a standard SATA-based SSD is marginal (because of its limited interface and relatively high response times), our sensitivity analysis suggests that future SSDs with lower response times will benefit from these changes. The effectiveness of our optimizations encourages discussion between the OS community and storage vendors about future device interfaces for fast storage devices.
{"title":"Optimizing the Block I/O Subsystem for Fast Storage Devices","authors":"Youngjin Yu, Dongin Shin, Woong Shin, N. Song, Jae-Woo Choi, H. Kim, Hyeonsang Eom, H. Yeom","doi":"10.1145/2619092","DOIUrl":"https://doi.org/10.1145/2619092","url":null,"abstract":"Fast storage devices are an emerging solution to satisfy data-intensive applications. They provide high transaction rates for DBMS, low response times for Web servers, instant on-demand paging for applications with large memory footprints, and many similar advantages for performance-hungry applications. In spite of the benefits promised by fast hardware, modern operating systems are not yet structured to take advantage of the hardware’s full potential. The software overhead caused by an OS, negligible in the past, adversely impacts application performance, lessening the advantage of using such hardware. Our analysis demonstrates that the overheads from the traditional storage-stack design are significant and cannot easily be overcome without modifying the hardware interface and adding new capabilities to the operating system. In this article, we propose six optimizations that enable an OS to fully exploit the performance characteristics of fast storage devices. With the support of new hardware interfaces, our optimizations minimize per-request latency by streamlining the I/O path and amortize per-request latency by maximizing parallelism inside the device. We demonstrate the impact on application performance through well-known storage benchmarks run against a Linux kernel with a customized SSD. We find that eliminating context switches in the I/O path decreases the software overhead of an I/O request from 20 microseconds to 5 microseconds and a new request merge scheme called Temporal Merge enables the OS to achieve 87% to 100% of peak device performance, regardless of request access patterns or types. Although the performance improvement by these optimizations on a standard SATA-based SSD is marginal (because of its limited interface and relatively high response times), our sensitivity analysis suggests that future SSDs with lower response times will benefit from these changes. The effectiveness of our optimizations encourages discussion between the OS community and storage vendors about future device interfaces for fast storage devices.","PeriodicalId":318554,"journal":{"name":"ACM Transactions on Computer Systems (TOCS)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125250836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yunsup Lee, Rimas Avizienis, Alex Bishara, R. Xia, Derek Lockhart, C. Batten, K. Asanović
We present a taxonomy and modular implementation approach for data-parallel accelerators, including the MIMD, vector-SIMD, subword-SIMD, SIMT, and vector-thread (VT) architectural design patterns. We introduce Maven, a new VT microarchitecture based on the traditional vector-SIMD microarchitecture, that is considerably simpler to implement and easier to program than previous VT designs. Using an extensive design-space exploration of full VLSI implementations of many accelerator design points, we evaluate the varying tradeoffs between programmability and implementation efficiency among the MIMD, vector-SIMD, and VT patterns on a workload of compiled microbenchmarks and application kernels. We find the vector cores provide greater efficiency than the MIMD cores, even on fairly irregular kernels. Our results suggest that the Maven VT microarchitecture is superior to the traditional vector-SIMD architecture, providing both greater efficiency and easier programmability.
{"title":"Exploring the Tradeoffs between Programmability and Efficiency in Data-Parallel Accelerators","authors":"Yunsup Lee, Rimas Avizienis, Alex Bishara, R. Xia, Derek Lockhart, C. Batten, K. Asanović","doi":"10.1145/2491464","DOIUrl":"https://doi.org/10.1145/2491464","url":null,"abstract":"We present a taxonomy and modular implementation approach for data-parallel accelerators, including the MIMD, vector-SIMD, subword-SIMD, SIMT, and vector-thread (VT) architectural design patterns. We introduce Maven, a new VT microarchitecture based on the traditional vector-SIMD microarchitecture, that is considerably simpler to implement and easier to program than previous VT designs. Using an extensive design-space exploration of full VLSI implementations of many accelerator design points, we evaluate the varying tradeoffs between programmability and implementation efficiency among the MIMD, vector-SIMD, and VT patterns on a workload of compiled microbenchmarks and application kernels. We find the vector cores provide greater efficiency than the MIMD cores, even on fairly irregular kernels. Our results suggest that the Maven VT microarchitecture is superior to the traditional vector-SIMD architecture, providing both greater efficiency and easier programmability.","PeriodicalId":318554,"journal":{"name":"ACM Transactions on Computer Systems (TOCS)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128702334","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
S. Gamage, R. Kompella, Dongyan Xu, Ardalan Kangarlou
Virtualization is a key technology that powers cloud computing platforms such as Amazon EC2. Virtual machine (VM) consolidation, where multiple VMs share a physical host, has seen rapid adoption in practice, with increasingly large numbers of VMs per machine and per CPU core. Our investigations, however, suggest that the increasing degree of VM consolidation has serious negative effects on the VMs’ TCP performance. As multiple VMs share a given CPU, the scheduling latencies, which can be in the order of tens of milliseconds, substantially increase the typically submillisecond round-trip times (RTTs) for TCP connections in a datacenter, causing significant degradation in throughput. In this article, we propose a lightweight solution, called vPRO, that (a) offloads the VM’s TCP congestion control function to the driver domain to improve TCP transmit performance; and (b) offloads TCP acknowledgment functionality to the driver domain to improve the TCP receive performance. Our evaluation of a vPRO prototype on Xen suggests that vPRO substantially improves TCP receive and transmit throughputs with minimal per-packet CPU overhead. We further show that the higher TCP throughput leads to improvement in application-level performance, via experiments with Apache Olio, a Web 2.0 cloud application, and Intel MPI benchmark.
{"title":"Protocol Responsibility Offloading to Improve TCP Throughput in Virtualized Environments","authors":"S. Gamage, R. Kompella, Dongyan Xu, Ardalan Kangarlou","doi":"10.1145/2491463","DOIUrl":"https://doi.org/10.1145/2491463","url":null,"abstract":"Virtualization is a key technology that powers cloud computing platforms such as Amazon EC2. Virtual machine (VM) consolidation, where multiple VMs share a physical host, has seen rapid adoption in practice, with increasingly large numbers of VMs per machine and per CPU core. Our investigations, however, suggest that the increasing degree of VM consolidation has serious negative effects on the VMs’ TCP performance. As multiple VMs share a given CPU, the scheduling latencies, which can be in the order of tens of milliseconds, substantially increase the typically submillisecond round-trip times (RTTs) for TCP connections in a datacenter, causing significant degradation in throughput. In this article, we propose a lightweight solution, called vPRO, that (a) offloads the VM’s TCP congestion control function to the driver domain to improve TCP transmit performance; and (b) offloads TCP acknowledgment functionality to the driver domain to improve the TCP receive performance. Our evaluation of a vPRO prototype on Xen suggests that vPRO substantially improves TCP receive and transmit throughputs with minimal per-packet CPU overhead. We further show that the higher TCP throughput leads to improvement in application-level performance, via experiments with Apache Olio, a Web 2.0 cloud application, and Intel MPI benchmark.","PeriodicalId":318554,"journal":{"name":"ACM Transactions on Computer Systems (TOCS)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122278070","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
J. Corbett, J. Dean, Michael Epstein, Andrew Fikes, Christopher Frost, J. Furman, S. Ghemawat, Andrey Gubarev, Christopher Heiser, P. Hochschild, Wilson C. Hsieh, Sebastian Kanthak, Eugene Kogan, Hongyi Li, Alexander Lloyd, S. Melnik, David Mwaura, D. Nagle, Sean Quinlan, Rajesh Rao, Lindsay Rolig, Yasushi Saito, Michal Szymaniak, Christopher Taylor, Ruth Wang, Dale Woodford
Spanner is Google’s scalable, multiversion, globally distributed, and synchronously replicated database. It is the first system to distribute data at global scale and support externally-consistent distributed transactions. This article describes how Spanner is structured, its feature set, the rationale underlying various design decisions, and a novel time API that exposes clock uncertainty. This API and its implementation are critical to supporting external consistency and a variety of powerful features: nonblocking reads in the past, lock-free snapshot transactions, and atomic schema changes, across all of Spanner.
{"title":"Spanner","authors":"J. Corbett, J. Dean, Michael Epstein, Andrew Fikes, Christopher Frost, J. Furman, S. Ghemawat, Andrey Gubarev, Christopher Heiser, P. Hochschild, Wilson C. Hsieh, Sebastian Kanthak, Eugene Kogan, Hongyi Li, Alexander Lloyd, S. Melnik, David Mwaura, D. Nagle, Sean Quinlan, Rajesh Rao, Lindsay Rolig, Yasushi Saito, Michal Szymaniak, Christopher Taylor, Ruth Wang, Dale Woodford","doi":"10.1145/2491245","DOIUrl":"https://doi.org/10.1145/2491245","url":null,"abstract":"Spanner is Google’s scalable, multiversion, globally distributed, and synchronously replicated database. It is the first system to distribute data at global scale and support externally-consistent distributed transactions. This article describes how Spanner is structured, its feature set, the rationale underlying various design decisions, and a novel time API that exposes clock uncertainty. This API and its implementation are critical to supporting external consistency and a variety of powerful features: nonblocking reads in the past, lock-free snapshot transactions, and atomic schema changes, across all of Spanner.","PeriodicalId":318554,"journal":{"name":"ACM Transactions on Computer Systems (TOCS)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129351502","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We present Abstract (ABortable STate mAChine replicaTion), a new abstraction for designing and reconfiguring generalized replicated state machines that are, unlike traditional state machines, allowed to abort executing a client’s request if “something goes wrong.” Abstract can be used to considerably simplify the incremental development of efficient Byzantine fault-tolerant state machine replication (BFT) protocols that are notorious for being difficult to develop. In short, we treat a BFT protocol as a composition of Abstract instances. Each instance is developed and analyzed independently and optimized for specific system conditions. We illustrate the power of Abstract through several interesting examples. We first show how Abstract can yield benefits of a state-of-the-art BFT protocol in a less painful and error-prone manner. Namely, we develop AZyzzyva, a new protocol that mimics the celebrated best-case behavior of Zyzzyva using less than 35% of the Zyzzyva code. To cover worst-case situations, our abstraction enables one to use in AZyzzyva any existing BFT protocol. We then present Aliph, a new BFT protocol that outperforms previous BFT protocols in terms of both latency (by up to 360%) and throughput (by up to 30%). Finally, we present R-Aliph, an implementation of Aliph that is robust, that is, whose performance degrades gracefully in the presence of Byzantine replicas and Byzantine clients.
{"title":"The Next 700 BFT Protocols","authors":"R. Guerraoui","doi":"10.1145/2658994","DOIUrl":"https://doi.org/10.1145/2658994","url":null,"abstract":"We present Abstract (ABortable STate mAChine replicaTion), a new abstraction for designing and reconfiguring generalized replicated state machines that are, unlike traditional state machines, allowed to abort executing a client’s request if “something goes wrong.” Abstract can be used to considerably simplify the incremental development of efficient Byzantine fault-tolerant state machine replication (BFT) protocols that are notorious for being difficult to develop. In short, we treat a BFT protocol as a composition of Abstract instances. Each instance is developed and analyzed independently and optimized for specific system conditions. We illustrate the power of Abstract through several interesting examples. We first show how Abstract can yield benefits of a state-of-the-art BFT protocol in a less painful and error-prone manner. Namely, we develop AZyzzyva, a new protocol that mimics the celebrated best-case behavior of Zyzzyva using less than 35% of the Zyzzyva code. To cover worst-case situations, our abstraction enables one to use in AZyzzyva any existing BFT protocol. We then present Aliph, a new BFT protocol that outperforms previous BFT protocols in terms of both latency (by up to 360%) and throughput (by up to 30%). Finally, we present R-Aliph, an implementation of Aliph that is robust, that is, whose performance degrades gracefully in the presence of Byzantine replicas and Byzantine clients.","PeriodicalId":318554,"journal":{"name":"ACM Transactions on Computer Systems (TOCS)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2008-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134424630","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：