"... In PAGE 7: ... The overhead of the poli- cies themselves is in the microseconds range for each speed computation (at most two speed computations per request are necessary for the stochastic scheme). Table4 evaluates the schemes on the Event Extraction and CAF traces described in Tables 1 and 2. To vary the system load for the same trace, the original inter-arrival times between requests are multiplied with a scale factor, as in [7].... In PAGE 7: ...8 results in a overloaded system that cannot keep up with the rate even if running at the max- imum speed at all times. The savings, shown in columns 2-5 of Table4 for each scale factor, are normalized to the no-power-management scheme. The stochastic approach results in the most energy savings, up to 28.... In PAGE 7: ...5x compared to no-power-management and up to 40% less energy compared to the second-best DVS scheme (the application-oblivious prediction). The stochastic scheme also results in the fewest speed changes per second (SC/s) among the DVS policies (see column 6, Table4 ). This is because many requests do not reach the point where they switch to the secondary speed.... ..."

"... In PAGE 7: ... The overhead of the poli- cies themselves is in the microseconds range for each speed computation (at most two speed computations per request are necessary for the stochastic scheme). Table4 evaluates the schemes on the Event Extraction and CAF traces described in Tables 1 and 2. To vary the system load for the same trace, the original inter-arrival times between requests are multiplied with a scale factor, as in [7].... In PAGE 7: ...8 results in a overloaded system that cannot keep up with the rate even if running at the max- imum speed at all times. The savings, shown in columns 2-5 of Table4 for each scale factor, are normalized to the no-power-management scheme. The stochastic approach results in the most energy savings, up to 28.... In PAGE 7: ...5x compared to no-power-management and up to 40% less energy compared to the second-best DVS scheme (the application-oblivious prediction). The stochastic scheme also results in the fewest speed changes per second (SC/s) among the DVS policies (see column 6, Table4 ). This is because many requests do not reach the point where they switch to the secondary speed.... ..."

"... In PAGE 26: ...be updated. Table7 describes the limits imposed and the percentage of URLs or sites whose crawl was stopped by reaching one of the limits. The maximum number of duplicates was overcome by 1.... ..."

"... In PAGE 9: ...elatively high with 0.173239 and 0.34599 respectively. In Table8 , we present the diversity of the downloaded pages. The diversity values of geographically focused col- laborative crawling strategies suggest that most of the geo- graphically focused collaborative crawling strategies tend to favor those pages which are found grouped under the same domain names because of their crawling method.... ..."

"... In PAGE 9: ...elatively high with 0.173239 and 0.34599 respectively. In Table8 , we present the diversity of the downloaded pages. The diversity values of geographically focused col- laborative crawling strategies suggest that most of the geo- graphically focused collaborative crawling strategies tend to favor those pages which are found grouped under the same domain names because of their crawling method.... ..."