DocumentCode :
3217916
Title :
Admission control and loss management for an application-level statistical service
Author :
Xie, Geoffrey G. ; Lam, Simon S.
Author_Institution :
Dept. of Comput. Sci., Naval Postgraduate Sch., Monterey, CA, USA
fYear :
1997
fDate :
28-31 Oct 1997
Firstpage :
142
Lastpage :
151
Abstract :
We present an admission control framework and loss management techniques in support of a guaranteed statistical service. The service is characterized by (i) the loss rate of application data units (ADUs) bounded below a specified value, and (ii) ADU losses distributed evenly among flows subscribing to the service and uniformly over the duration of each flow. Specifically, a flow is modeled as a sequence of bursts, each of which is a sequence of packets that carry the bits of an ADU. The first packet of each burst carries information on the ADU (e.g., its bandwidth requirement). This traffic model enables admission control at the burst level as well as at the flow level. Such a two level admission control approach is very effective in bounding end-to-end ADU loss rates of flows while maintaining high channel utilization in the network. The traffic model also enables simple techniques that can be used at a network channel to distribute ADU losses evenly among flows subscribing to the same statistical service, and to protect high priority ADUs (e.g., 1 frames of MPEG applications)
Keywords :
bandwidth compression; packet switching; transport protocols; ADU losses; MPEG applications; admission control; application data units; application-level statistical service; bandwidth requirement; guaranteed statistical service; loss management; loss rate; Admission control; Application software; Communication system traffic control; Delay; Loss measurement; Performance loss; Quality of service; Time measurement; Traffic control; Web and internet services;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Network Protocols, 1997. Proceedings., 1997 International Conference on
Conference_Location :
Atlanta, GA
ISSN :
1092-1648
Print_ISBN :
0-8186-8061-X
Type :
conf
DOI :
10.1109/ICNP.1997.643707
Filename :
643707
Link To Document :
بازگشت