Advances and Applications in Statistics
Volume 24, Issue 1, Pages 7 - 33
(September 2011)
|
|
ENHANCED VARIANCE-BASED SUSPENDED STREAM DETECTION
Kazumasa Oida
|
Abstract: Audio/video streams may be interrupted permanently or for a long time due to failures, congestion, timeouts, software bugs, malicious attacks, etc. Stream suspensions must be as short as possible, especially when the service is provided for business purposes or when the contents have scientific or economic value. This paper proposes an algorithm that monitors a number of audio/video streams to quickly detect the suspension of one or more of these streams. The algorithm uses only sample variances of the number of arriving packets, with the result that it works in any streaming environment and does not require any stream-related parameter values in advance. The algorithm is based on a new approach, which enables quick detection without using a large amount of traffic data. This is numerically and analytically verified. Experiments are made using captured stream packets. If one of 40 video streams is interrupted, then it takes about ten second on average to detect the interruption. |
Keywords and phrases: suspension detection, streaming media, traffic analysis, variance plot. |
|
Number of Downloads: 352 | Number of Views: 1207 |
|