Wednesday, March 21, 2012

system reduces Norway VPN

savings on the access network over a 48-hour period. In the48 hour period, 102 hours of movies are consumed in total,and each house watches 8 to 18 hours of content.1) Bandwidth savings: Figure 6 shows the broadband bandwidth use over the 48 hour period. We refer to our system asNaN and w/ NaN shows the performance of our system. W/oNaN shows the amount of bandwidth that would have used ifall the content was coming from the ISP through broadband.The average second-mile link bandwidth with NaN andwithout NaN is 10.1Mbps and 21.2Mbps respectively. Our system reduces Norway VPN the average bandwidth by 52%. The peak bandwidth usage is reduced by 16% (from 60Mbps to 50.3Mbps),while the 95th percentile bandwidth usage is reduced by 40%(from 50Mbps to 29.9Mbps).The peak bandwidth savings in the nine node testbed wassmall because of its scale. Although the top 1111 moviesaccount for 50% of demand, the hit rate is not always 50%in this small neighborhood. For example, at around the 13thhour, five out of the six movies requested (over 80%) were notstored in the neighborhood and had to be requested from thevideo server. In practice, however, a single second-mile linkserves hundreds of homes, and the probability that 80% ofthe nodes simultaneously request unpopular movies not storedin the neighborhood would be much smaller. Thus, the peakbandwidth in this case will be much closer to the average (dueto the law of large numbers).To get a better understanding of the ISP's bandwidth savings, we simulated a larger neighborhood of 500 homes withdifferent placement policies. Figure 7 shows the aggregatebandwidth demand for a range of different placement policiesfor content in the neighborhood. The w/o NaN line representsa system where neighborhood nodes store nothing. The no

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.