High-throughput Distributed Services
by Dejan Kostic, EPFL, Dejan.Kostic@epfl.ch

Abstract:
With the price of wireless sensor technologies diminishing rapidly we can
expect large numbers of autonomous sensor networks being deployed in the near
future. These sensor networks will typically not remain isolated but will be
interconnected, thus realizing the vision of a global "Sensor
Internet". Important application in this domain will be aggregation of sensor
measurements with a scalable, decentralized data store, such as a DHT. Sensor
measurements are typically small (100bytes) and the sheer volume of generated
measurements in the future can require billions of identifiers to be inserted
into a DHT.  To motivate need for congestion control in DHTs, I will
demonstrate that congestion collapse is possible in a DHT. Moving forward, a
major challenge is to increase the throughput of DHTs to keep up with growing
data insertion rates. Toward this end, I will briefly outline our initial
results on congestion-aware routing to increase DHT throughput. As the
last-mile link capacities are starting to reach tens of Mbps,it becomes
necessary to outline the ways the future Internet applications will leverage
the increased end host bandwidth to provide better service quality and new
functionality. The second application I will discuss, peer-to-peer (p2p)
video-on-demand (VoD), is increasingly popular with Internet users. Currently
deployed pure p2p VoD systems provide poor general performance and they lack
advanced features such as fast forward and seeking to
arbitrary points. Peer-assisted VoD systems can provide such services, but
they require very well provisioned source servers (or server farms) to avoid
overloading the content source. We argue that proactive caching should be 
used to leverage the extra bandwidth to replicate the content within the
overlay. Quickly establishing multiple copies of content can enable advanced
features with out requiring a well provisioned server. Several challenges arise
in this scenario: 
  i)   rapid discovery of peers with required content blocks when peers perform
       seeks, 
  ii)  controlling content block replication under dynamic network conditions,
       and 
  iii) maximizing the use of spare bandwidth without hurting playback
       (multi-overlay bandwidth provisioning). Initial experimental results
       from a VoD prototype application demonstrate that it can both 
       effectively control in-overlay block replication and can efficiently use
       these replicas to perform forward seeks.