Boosting Streaming Video Delivery with WiseReplica - Transactions on Large-Scale-Data-and Knowledge-Centered Systems XX - page 56

These facts endure our decision of a locality aware solution for infrastructure.
Liu et al. [ 21 ] make a case for a video control plane that can use a global view
of client and network conditions to dynamically optimize the video delivery in
order to provide a high quality viewing experience despite an unreliable delivery
infrastructure. However, the granularity of their server selection mechanism is at
a CDN, ignoring edge resources. WiseReplica addresses this issue by adapting
replication close to the viewers. Thus, WiseReplica can be play an important
role in collaborating with an Internet control plane.
Adaptive replication schemes: Non-collaborative caching remains the sim-
plest approach to provide popularity-aware replication of web content through
cache replacement policies [ 20 ]. However, we showed when we adapt the number
of replicas according to the Internet video popularity properly, cache replacement
policy becomes redundant. EAD [ 27 ] and Skute [ 4 ] adapt the number of replicas
by using a cost-benefit approach over decentralized and structured P2P sys-
tems. EAD creates and deletes replicas throughout the query path with regard
to object hit rate using an exponential moving average technique. Similarly,
Skute provides a replication management scheme that evaluates replicas price
and revenue across different geographic locations. Despite presenting an ecient
framework for replication, they provide an inaccurate bitrate provision, hence
inappropriate for high-quality video delivery. WiseReplica copes with this issue
by analysing the request arrival process, performing accurate predictions about
the ranking of Internet videos, and maintaining replication degree accordingly.
8 Conclusions
In this work, we presented WiseReplica, a SLA-based, adaptive replication scheme
for meeting customers' expectations and enhancing resource allocation in peer-
assisted VoD systems. To adapt replication, WiseReplica relies on a prediction
model for ranking Internet videos in order of demand. Our intuitive model is flex-
ible, and can learn from different sources and big amounts of data, providing a
robust framework for controlling VoD resource allocation. Simulations using You
Tube traces suggest that our ranking predictions of videos are important to enha-
nce video delivery in peer-assisted VoD systems, allowing us to self-adapt replica-
tion degree to video demand properly. WiseReplica increases the average bitrate
provision by roughly 85 %, contributing decisively to enhance viewing experience
of users. Our future work will mainly cover a proof-of-concept prototype for eval-
uating WiseReplica in a real testbed.
References
1. Adhikari, V.K., Jain, S., Chen, Y., Zhang, Z.-L.: Vivisecting youtube: an active
measurement study. In: INFOCOM (2012)
2. Alizadeh, M., Greenberg, A., Maltz, D.A., Padhye, J., Patel, P., Prabhakar, B.,
Sengupta, S., Sridharan, M.: Data center TCP (DCTCP). In: SIGCOMM (2010)