Citation Context

...video content. Existing video workload generators are unsuitable for our purposes because they generate workloads for analysis and simulation rather than for creating traffic to exercise a web server =-=[1, 18, 31, 34]-=-. Although these workloads could be used as part of a benchmark [18], it would require significant re-engineering of the workload generators. General-purpose workload generators [4, 9, 12] can be used...

Citation Context

...C objects than other categories. 4.4 Category Popularity Distributions Figure 7 shows the Rank-frequency distribution for the 6 categories that showed the most interesting patterns. Previous studies (=-=Abhari and Soraya, 2010-=-; Cheng et al., 2007) showed that although requests for popular YouTube videos follow a Zipf-like distribution, a Weibull distribution fits better because of the heavy tail section, which indicates a ...

Citation Context

... almost impossible. In the literature, one can find many analysis that reflect trends of YouTube traffic from both a local campus network and a global (i.e., Internet wide) perspective [2], [3], [4], =-=[5]-=-, [6] and [7]. 2.1 How does YouTube work? One of the keys to YouTube’s success is its use of Adobe’s Flash Video (FLV) format for video delivery [8]. While users may upload content in a variety of med...

Citation Context

...traffic by QoS parameters based on queuing models and wavelet scaling D-ITG [16] Extensive workload generation framework that can produce traffic for wide range of network scenarios Comparison of original and synthetic traffic by throughput and distributions of IPT and PS values Special Scenario Generators EAR [17] Traffic replay technique for mimic IEEE 802.11protocol behavior Unique metric for measuring wireless traffic replay called Event Reproduction Ratio ParaSynTG [18] Web traffic generator Web specific metrics such as document size andpopularity distributions YouTube Workload Generator [19] Workload generation methods for mimic YouTube video traffic Online video specific metrics like video length, size and rating distributions or cache performance Graph-Based Traffic Generator [20] Flow trace generator based on Traffic Dispersion Graphs templates Graph related metrics such as distribution of degrees or connected edges and verticals validate their methodology by presenting wavelet based analysis [21] of the original and the replayed trace. Even though explanation for visual differences in various time scales are given in the paper, there is no qualitative metric presented to eval...

Citation Context

...c) Average span pattern Fig. 8. Video popularity patterns was proposed by Abhari and Soraya: they investigated the possibility of generating UGC workload that corresponds to statistical user behavior =-=[9]-=- inferred from website crawling. Then, the simulated user simply randomly chooses video files to view based on their popularity, until the expected session time has passed. By contrast, our approach i...

Citation Context

...esented in Section 4. Section 5 concludes the paper and discusses directions for future work. 2 Related Work Many studies have been carried to analyze the users’ behaviors in different media services =-=[1, 18, 22]-=-. Early work in this area is focused on the analysis and characterization of streaming services in the Internet [22]. In [22], the authors explored workload characteristics based on logs from internal...

Citation Context

...we crafted a more detailed Markov model. Another approach was proposed by Abhari and Soraya: they investigated the possibility of generating UGC workload that corresponds to statistical user behavior =-=[9]-=- inferred from website crawling. Then, the simulated user simply randomly chooses video files 60 20 40 60 80 100 120 0 5 10 15 Days #sofss es sio ns (a) Long span pattern 0 20 40 60 80 100 120 0 20 40...