A basic requirement for every P2P system is fault-tolerance. Since the primary objective is resource location and sharing, we require that this basic operation takes place in a reliable manner. In a variety of situations with skewed data accesses (e.g., [1], etc) the demand for content can become overwhelming for certain serving peers, forcing them to reject connections. In many cases, these skewed distributions take extreme forms: Flash crowds, regularly documented surges in the popularity of certain content, are also known to cause severe congestion and degradation of service [2]. Data replication techniques is one commonly utilized solution to remedy these situations. Nevertheless, there are cases in which the requested resources cannot be arbitrarily replicated. Distributed data-structures that support range-queries is such an example: The keys are stored in the network nodes so that a natural order is preserved. These structures can be very useful in a variety of situations: On-line games, web servers, data-warehousing, etc. In such cases, adaptive and on-line load-balancing schemes must be employed in order to avoid resource unavailability and performance in a variety of workloads[3,4].
[1]
Pablo Rodriguez,et al.
I tube, you tube, everybody tubes: analyzing the world's largest user generated content video system
,
2007,
IMC '07.
[2]
Balachander Krishnamurthy,et al.
Flash crowds and denial of service attacks: characterization and implications for CDNs and web sites
,
2002,
WWW.
[3]
Hector Garcia-Molina,et al.
Online Balancing of Range-Partitioned Data with Applications to Peer-to-Peer Systems
,
2004,
VLDB.
[4]
David R. Karger,et al.
Simple Efficient Load-Balancing Algorithms for Peer-to-Peer Systems
,
2004,
SPAA '04.