The consistent, high-quality delivery of real-time content over IP networks has significant challenges. Buffering, pixilation, stop and starts, long download times - these are all a part of the common experience of using real-time content over the Internet. These negative experiences will only continue as the demand for Internet-based real-time content increases.
Real-time content requires real-time support from the network to provide a consistently high quality experience to the end-user. This entails strict adherence to loss, latency and jitter boundaries, which require the underlying network protocols to react to changes such as congestion in real-time. It is the networks' inability to react to these changes, and utilize the network's resources that poses a challenge, and makes high quality delivery of real-time content over the Internet difficult.
Traditional content distribution network (CDN) solutions have attempted to address this issue by using a distributed set of caching servers that are able to serve content for a client from a nearby location. More recently, a new approach to CDNs called peer-to-peer CDNs (or P2P CDNs) has emerged that organizes end-user machines into a P2P network of distributed caches. Such P2P CDNs may or may not also rely on dedicated caching servers like traditional CDNs.
While P2P CDNs have tremendous promise, exercising them to their fullest capabilities has proven to be unachievable due to two key reasons:
- Unintelligent understanding of the network: Existing CDNs and P2P CDNs do not intelligently characterize the network, and thus, have limited information on which they can react to problems in the underlying network. This is especially amplified when the techniques are deployed in more difficult network environments, such as wireless networks.
- Uncoordinated operation among peers: In existing P2P CDNs, peers transfer data in an uncoordinated manner to the destination. This limits the ability to react to network changes in real-time. Application performance gains are limited, as are the improvements in network parameters such as loss, jitter and latency that applications can benefit from.
The above issues in existing CDNs stem from a disconnect between the use of P2P solutions at the application level, and underlying single-path protocols at the network transport level. This disconnect results in restricted benefits including - (a) limited or no performance improvements (b) no universal applicability to multiple media types, (i.e. typically there is one solution for downloads and another for streaming, and yet another for interactive media) (c) implementation or usage complexity, and (d) an inability to perform in difficult environments such as mobile wireless networks.
parallel networking
multipath transport
multiflow tcp
multicaching
|
 |
 |
 |
 |
 |
|
PDF:
Asankya's Technology
download >
|
|
 |
|
