2000
DOI: 10.21236/ada382318
|View full text |Cite
|
Sign up to set email alerts
|

Towards Higher Disk Head Utilization: Extracting Free" Bandwidth From Busy Disk Drives"

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

1
52
0

Year Published

2003
2003
2018
2018

Publication Types

Select...
3
3
2

Relationship

1
7

Authors

Journals

citations
Cited by 48 publications
(53 citation statements)
references
References 0 publications
1
52
0
Order By: Relevance
“…Since I/O requests usually fall into several patterns in parallel file systems, it is possible to reorganize the data layout to reduce the number of disk seeks [19][20] [21]. These data reorganization techniques improve the overall I/O performance due to a prior knowledge of application I/O behaviors.…”
Section: B Layout Optimizationmentioning
confidence: 99%
“…Since I/O requests usually fall into several patterns in parallel file systems, it is possible to reorganize the data layout to reduce the number of disk seeks [19][20] [21]. These data reorganization techniques improve the overall I/O performance due to a prior knowledge of application I/O behaviors.…”
Section: B Layout Optimizationmentioning
confidence: 99%
“…Both impacts result in shorter seek time in terms of equation (3). Though the seek time has been decreased significantly due to the increasing BPI, the Lumb et al[2000] investigated the impact of seek time, rotational latency and data transfer time that add up to 100% of the disk head utilization for five modern disk drives which were sold on the market from 1996 to 1999. The investigation indicated that the faster seek of the Cheetah 18LP (average seek time 5.2ms), relative to the Cheetah 9LP and Cheetah 4LP which have average seek time of 5.4ms and 7.7ms respectively, resulted in lower seek components.…”
Section: Seek Timementioning
confidence: 99%
“…In addition, when supervisors decide to move data from one worker to another, it should be a direct transfer (via a worker-to-worker COPY request) rather than passing through an intermediary. Further, maintenance functions such as these should happen with minimal effect on real user workloads, requiring careful scheduling of network usage [5,33] and disk activity [20,35].…”
Section: Workersmentioning
confidence: 99%
“…Disk bandwidth for maintenance functions will be provided by freeblock scheduling [35], which we have implemented [34] and refined into a working system. For most access patterns, it gives background disk maintenance activities 15-40% of the disk head's time with near-zero impact on foreground activity (i.e., real user requests).…”
Section: Building Blocksmentioning
confidence: 99%