Information Technology Reference
In-Depth Information
One of the goals of tree-push protocols is to construct an efficient multi-
cast tree. However, it is observed that even though the data dissemination
paths are efficient, the intermediate nodes tend to overload with increasing
scale. Additionally, most tree-push architectures do not adapt well to node
churns. When nodes join and leave frequently, the service would be inter-
rupted [41,42].
1.4.2 Mesh-Pull Methods
Many popular P2P live streaming systems make use of protocols modified
from P2P file sharing protocols like BitTorrent to deliver media content. The
peers are randomly organized into a mesh structure. The selection of active
peers for data exchange is commonly decided based on the round trip time
(RTT) between peers. Peers exchange information about what video chunks
they have before exchanging media data. This is known as the gossiping
process. After that, in the pulling process, peers request video chunks that
they do not have from other peers who do (see Figure 1.23) [5-7].
To explain the mesh-pull P2P streaming applications, we first briefly
describe the BitTorrent file sharing protocol. There are basically three com-
ponents within the BitTorrent protocol: a web server keeping the torrent files
(the metadata of the sharing files); a tracker keeping track of the information
of all participating peers; and a swarm of seeds and leechers, where seeds are
peers that have finished downloading the file but are still active for uploading
to others and leechers are peers that are still downloading (see Figure 1.24).
When a new leecher tries to join the torrent, it first connects to a web server
to download the torrent file. The torrent file includes the detailed file and
tracker location. After that, the new leecher registers its information with the
tracker and gets an initial peer list. After the leecher has this initial peer list,
it starts connecting to other peers to request exchanges of file chunks.
Figure 1.23
Mesh-pull streaming.
