Reliable and Fault-Tolerant Storage Systems - Scalable Continuous Media Streaming Systems - page 104

600

400

200

0

0 0.2 0.4 0.6 0.8 1

Server Utilization

Quantum Atlas-III
Quantum Atlas-10K
Seagate Barracuda
Seagate Cheetah
IBM 9ES

Figure 5.10 Rebuild time versus server utilization for track-based rebuild ( Q = 64KB, N D = 5)

well below such a high utilization to minimize blocking, the rebuild time in practice is likely
to be even shorter.

5.7.2 Sensitivity to Server Utilization
Figure 5.11 plots the reduction in rebuild time by track-based rebuild versus server utilization.
We observe that track-based rebuild consistently achieves significant rebuild-time reductions
over a wide range of server utilization. This result demonstrates that performance gain of the
track-based rebuild algorithm is stable with respect to server utilization.

5.7.3 Sensitivity to Media Block Size
In Figure 5.12, we plot the rebuild time versus the media block size at three server utilizations
of 0, 0.25 and 0.5 respectively. We observe that rebuild time for block-based rebuild decreases
with increases in media block size as larger block size increases the overall disk efficiency. By
contrast, rebuild time for track-based rebuild is relatively insensitive to the media block size
used as retrievals are done in whole tracks instead of blocks.
Figure 5.13 plots the reduction in rebuild time versus the media block size. As expected, the
reduction decreases for increases in media block size as rebuild performance for block-based