| Popis: |
Bandwidth dictates the potential speed at which a channel carries information and hence the services it can support. Speech telephony for example requires only 4kHz bandwidth, while high quality sound requires 15kHz and video 6MHz (Glover & Grant, 2004). When several services share a common channel it is crucial to know the accessible bandwidth such that it can be divided efficiently between customers. Furthermore senders may even have to modify their behaviour in order to make best use of their bandwidth share (Yu et al., 2003). In an IP network bandwidths can be measured directly using the SNMP protocol to interrogate link components for their capacities and traffic loads: However since administrative access is not typically available to customers, numerous “end-to-end” measurement techniques have appeared whereby network traffic and infrastructure can be inferred from “probe” transmissions (Prasad et al., 2003). However, these techniques are only as good as their underlying assumptions and a plethora of misunderstandings and ambiguities have arisen (Jain & Dovrolis, 2004). Some of these problems are technological; wireless components introduce complications not experienced in wired networks (Johnsson et al., 2005) and traffic-shaping mechanisms can make the instantaneous bandwidth different from that experienced by sustained transmissions (Lakshminarayanan et al., 2004). However difficulties still arise under the simplest technological assumptions due to the complex behaviour of packet-streams within bottleneck links. |
