東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Beyond Storage Interfaces: Finding a...

He, Jun.

FindBook

Google Book

Amazon

博客來

Beyond Storage Interfaces: Finding and Exploiting Unwritten Contracts In Storage Devices.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Beyond Storage Interfaces: Finding and Exploiting Unwritten Contracts In Storage Devices./
作者:	He, Jun.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2019,
面頁冊數:	168 p.
附註:	Source: Dissertations Abstracts International, Volume: 80-09, Section: B.
Contained By:	Dissertations Abstracts International80-09B.
標題:	Computer science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=13805871
ISBN:	9780438899179

Beyond Storage Interfaces: Finding and Exploiting Unwritten Contracts In Storage Devices.
He, Jun.

Beyond Storage Interfaces: Finding and Exploiting Unwritten Contracts In Storage Devices. - Ann Arbor : ProQuest Dissertations & Theses, 2019 - 168 p.

Source: Dissertations Abstracts International, Volume: 80-09, Section: B.

Thesis (Ph.D.)--The University of Wisconsin - Madison, 2019.

This item must not be sold to any third party vendors.

Efficient interactions between the software and storage devices are essential for achieving high performance in data processing systems. However, efficient interacting is challenging because such interactions are beyond the storage interfaces: complying with the specifications of the interfaces, which we call "written contracts", does not guarantee high performance. To achieve high performance, clients of storage devices must follow an additional set of implicit rules, which we call "unwritten performance contracts". The specific unwritten performance contract of a device depends on its internal architecture. Violating the contract lowers the performance of the data processing system. As a result, we must find violations of the unwritten contract and exploit the contract to achieve high performance. We start by finding violations of the unwritten contracts for HDDs and SSDs, respectively. First, we study the methodologies for finding violations of the HDD unwritten contract by using statistical methods (e.g., Latin hypercube sampling and Sensitivity Analysis). Our study finds and fixes four design issues in a popular file system, Linux ext4. Second, we find violations of the SSD unwritten contract by full-stack analysis of modern applications, file systems, and SSD FTLs. We begin by formalizing the rules of the SSD unwritten contract; we then conduct analysis to uncover application and file system designs that violate the contract. As a result, we make numerous interesting observations that could shed light upon the future storage system designing. Finally, we exploit the unwritten contract of SSDs to build a cost-effective and high-performance data processing system. To this end, we propose Tiny-Memory Processing Systems (TMPS), which exploit fast and cheap SSDs to reduce expensive memory. To study the TMPS approach, we build a complete search engine that employs careful designs in data layout, early pruning, prefetching and space/traffic tradeoffs. When compared with a state-of-the-art search engine in a tiny-memory configuration, our search engine reduces I/O amplification by up to 3.2 times, query latency by up to 16 times, and increases the query throughput by up to 2.9 times.

ISBN: 9780438899179Subjects--Topical Terms:

523869
Computer science.

Beyond Storage Interfaces: Finding and Exploiting Unwritten Contracts In Storage Devices.
LDR:03324nmm a2200313 4500 001 2206300
005 20190829083236.5
008 201008s2019 ||||||||||||||||| ||eng d
020 $a 9780438899179
035 $a (MiAaPQ)AAI13805871
035 $a (MiAaPQ)wisc:16021
035 $a AAI13805871
040 $a MiAaPQ $c MiAaPQ
100 1 $a He, Jun. $3 1030425
245 1 0 $a Beyond Storage Interfaces: Finding and Exploiting Unwritten Contracts In Storage Devices.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2019
300 $a 168 p.
500 $a Source: Dissertations Abstracts International, Volume: 80-09, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Arpaci-Dusseau, Andrea C.;Arpaci-Dusseau, Remzi H.
502 $a Thesis (Ph.D.)--The University of Wisconsin - Madison, 2019.
506 $a This item must not be sold to any third party vendors.
520 $a Efficient interactions between the software and storage devices are essential for achieving high performance in data processing systems. However, efficient interacting is challenging because such interactions are beyond the storage interfaces: complying with the specifications of the interfaces, which we call "written contracts", does not guarantee high performance. To achieve high performance, clients of storage devices must follow an additional set of implicit rules, which we call "unwritten performance contracts". The specific unwritten performance contract of a device depends on its internal architecture. Violating the contract lowers the performance of the data processing system. As a result, we must find violations of the unwritten contract and exploit the contract to achieve high performance. We start by finding violations of the unwritten contracts for HDDs and SSDs, respectively. First, we study the methodologies for finding violations of the HDD unwritten contract by using statistical methods (e.g., Latin hypercube sampling and Sensitivity Analysis). Our study finds and fixes four design issues in a popular file system, Linux ext4. Second, we find violations of the SSD unwritten contract by full-stack analysis of modern applications, file systems, and SSD FTLs. We begin by formalizing the rules of the SSD unwritten contract; we then conduct analysis to uncover application and file system designs that violate the contract. As a result, we make numerous interesting observations that could shed light upon the future storage system designing. Finally, we exploit the unwritten contract of SSDs to build a cost-effective and high-performance data processing system. To this end, we propose Tiny-Memory Processing Systems (TMPS), which exploit fast and cheap SSDs to reduce expensive memory. To study the TMPS approach, we build a complete search engine that employs careful designs in data layout, early pruning, prefetching and space/traffic tradeoffs. When compared with a state-of-the-art search engine in a tiny-memory configuration, our search engine reduces I/O amplification by up to 3.2 times, query latency by up to 16 times, and increases the query throughput by up to 2.9 times.
590 $a School code: 0262.
650 4 $a Computer science. $3 523869
690 $a 0984
710 2 $a The University of Wisconsin - Madison. $b Computer Sciences. $3 2099760
773 0 $t Dissertations Abstracts International $g 80-09B.
790 $a 0262
791 $a Ph.D.
792 $a 2019
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=13805871