Intel IRP-TR-03-10 warranty No Lookaside, 100 Mb/s

Page 13

Slowdown

Slowdown

Slowdown

Slowdown

1000%

900%

800%

700%

600%

500%

400%

300%

200%

100%

0%

1000%

900%

800%

700%

600%

500%

400%

300%

200%

100%

0%

1000%

900%

800%

700%

600%

500%

400%

300%

200%

100%

0%

1000%

900%

800%

700%

600%

500%

400%

300%

200%

100%

0%

No Lookaside

 

1000%

 

900%

 

800%

 

700%

Slowdown

600%

500%

 

400%

 

300%

200%

100%

0%

Operations Sorted by Slowdown

(a) 100 Mb/s

 

1000%

 

900%

 

800%

 

700%

Slowdown

600%

500%

 

 

400%

 

300%

200%

100%

0%

Operations Sorted by Slowdown

(b) 10 Mb/s

 

1000%

 

900%

 

800%

 

700%

Slowdown

600%

500%

 

 

400%

 

300%

200%

100%

0%

Operations Sorted by Slowdown

(c) 1 Mb/s

 

1000%

 

900%

 

800%

 

700%

Slowdown

600%

500%

 

 

400%

 

300%

200%

100%

0%

Operations Sorted by Slowdown

(d) 100 Kb/s

With Lookaside

Operations Sorted by Slowdown

Operations Sorted by Slowdown

Operations Sorted by Slowdown

Operations Sorted by Slowdown

These figures compares the distribution of slowdown for the operations of the CDA benchmark without lookaside caching to their

slowdowns with lookaside caching to a DVD.

12

 

Figure 12. Impact of Lookaside Caching on Slowdown of CDA Benchmark Operations

Image 13
Contents Integrating Portable Distributed Storage Abstract IntroductionLookaside Caching BackgroundPrototype Implementation Evaluation Experimental SetupKernel Compile Benchmark Description Results Device File ServerInternet Suspend/Resume Benchmark Description How slow is the resume step? Lookaside Win With100 Mb/s 10 Mb/s 69.2% 317 96.2% 100 Kb/s 4301 99.7% With LookasideTime for Trace Replay Broader Uses of Lookaside CachingConclusion ANDERSON, T., DAHLIN, M., NEEFE, J., PATTERSON, D References100 Mb/s No Lookaside