A performance analysis of context transfer protocols for QoS enabled internet services
Introduction
The introduction of several multimedia services in new generation of wireless networks, brought about the need to develop efficient methods to manage the mobility of users. Nowadays internet services are often session oriented, delay bounded (or real-time) and context sensitive. Just to mention some, VoIP, multimedia streaming, on-line games, on-line transactions and many Content Delivery Networks (CDN) related services are often session oriented, delay bounded and context sensitive. In wired networks, the use of broadband technologies has a significant impact on the user’s perceived Quality of Service (QoS) making Service Level Agreements (SLA) achievable. On the contrary, in wireless networks the introduction of broadband wireless connectivity is not sufficient to guarantee the fulfillment of QoS requirements mostly due to users movement across the network coverage areas managed by different access routers (AR). Handover requests may be issued during critical service phases for which the avoidance of service disruption is mandatory, and the connection must be seamlessly handed off from a point of access to another. The fast handover mechanism, introduced to reduce the packet losses during handovers, needs to be enhanced with proper mechanisms to preserve the service continuity. In context and session based services, the realization of a handover is not only a matter of keeping a connection alive during users movements, but also of transferring the necessary information to avoid the re-establishment of a service session every time the user reaches a new point of access. The re-establishment of a service session causes the repetition of the service protocol message flow from scratch and is necessary if the information to keep the service alive is unavailable when a handover to a new point of access occurs. Thence service continuity and context transfer during handover procedures are very critical for delay sensitive and context dependent applications.
The IETF SeaMoby working group identifies general motivations for Context Transfer [13] and defines a Context Transfer Protocol (CTP) [14]. In Section 2, we consider critical scenarios like the one of Content Delivery Networks (CDN) supporting mobile users, in which context-aware handovers are of significant impact on quality of service. In Section 3, we show the interaction between CTP and Mobile IPv6 protocol, with fast handover mechanisms to reduce packet losses. Since understanding how and when the context transfer can be activated by a mobile node or access router is fundamental to give a performance model and evaluation of the CTP, in Section 4 we describe the CTP message flow in tree different cases: dummy (post-handoff) context transfer, mobile initiated context transfer and access router initiated context transfer.
A performance model of the CTP is given in Section 5, where performance is evaluated in terms of bandwidth occupation, packet loss, percentage of packets that violate the SLA, context transfer time and completion time of the protocol message flow. Section 6 concludes the paper.
Section snippets
Motivation for context transfer
All the information needed to negotiate, establish and manage network services may be considered part of the context to be transferred when a Mobile Node (MN) issues a handover request during an ongoing service.
The context data include:
- •
authentication, authorization, and accounting information [13] needed to permit the re-authentication of the mobile host and the mobile host’s authorization to access the network service from a new subnet;
- •
header compression [13] information that is necessary to
Mobility management mechanisms
Though context transfer may reduce latency in handoff management by reducing the number of messages needed for service re-establishment, mobility management must also be supported by proper mechanisms at the network access level. In order to evaluate the impact of the context transfer protocol on performance, its interaction with the underlying mobility management protocol must be considered and investigated. In the following subsections we discuss about mobility management schemes and we focus
The context transfer protocol
In this section we shortly describe the Context Transfer Protocol (CTP) [14] that has been proposed by the IETF SeaMoby working group. We focus on the possible message flows generated to grant transparent mobility management to a mobile node accessing a CDN service [3], e.g. an e-commerce site or a streaming video, provided by means of a content delivery distributed infrastructure. If CDN services are ongoing while the user moves, context transfer is needed at different layers by different
Performance analysis of CTP
We introduce a performance model to evaluate the cost of CTP in terms of: consumed bandwidth and number of packets that have been lost or erroneously processed according to the default method, without considering the necessary context information.
At least three entities are involved in CTP: the mobile node, the previous access router and one or more new access routers. Thus, as represented in Fig. 2, we distinguish among the amount of data exchanged on the side of the mobile node, BMN, of the
Conclusions and remarks
A considerable number of network services characterized by long lived sessions show a strong need for transparent procedures to transfer context information between network access points. The context transfer must be efficient to support low-latency and real-time applications.
In this paper we made a performance analysis of context transfer protocols, comparing three scenarios differentiated on the basis of the trigger mechanism in use to activate the context transfer procedures. Our analysis
Acknowledgments
The work of Novella Bartolini has been funded by the WEB-MINDS project supported by the Italian MIUR under the FIRB program, and by the POLLENS project, supported by ITEA. The work of Emiliano Casalicchio has been funded by the PERF project supported by the Italian MIUR under the FIRB program.
Novella Bartolini graduated with honors in 1997 and received her Ph.D. in computer engineering in 2001 from the University of Rome, Italy. She is now assistant professor at the University of Rome. She was researcher at the Fondazione Ugo Bordoni in 1997, visiting scholar the University of Texas at Dallas in 1999–2000 and research assistant at the University of Rome ‘Tor Vergata’ in 2000–2002. Her research interests lie in the area of wireless mobile networks and content delivery systems.
References (24)
- G. Agarwal, R. Shah, J. Walrand, Content distribution architecture using network layer anycast, in: Second IEEE...
- et al.
Capacity Planning for Web Performance: Models, Metrics, and Methods
(1998) - N. Bartolini, E. Casalicchio, S. Tucci, Mobility-aware admission control in content delivery networks, in: Proceedings...
- et al.
Performance Tools and Applications to Networked Systems
(2004) - B. Carpenter, C. Huitema, Deprecating site local addresses, Internet Draft, November...
- M. Gritter, D.R. Cheriton, An architecture for content routing support in the internet. Available from:...
- R. Hinder, S. Deering, Ip version 6 addressing architecture, IETF Internet Draft, October...
- R. Hinder, B. Haberman, Unique local ipv6 unicast addresses, Internet Draft, February...
- J. Rosenberg et al., Sip: session initiation protocol, RFC 3261, June...
- D. Johnson, C. Perkins, J. Arkko, Mobility support in ipv6, IETF Mobile IP Working Group Internet-Draft, May...
Cited by (0)
Novella Bartolini graduated with honors in 1997 and received her Ph.D. in computer engineering in 2001 from the University of Rome, Italy. She is now assistant professor at the University of Rome. She was researcher at the Fondazione Ugo Bordoni in 1997, visiting scholar the University of Texas at Dallas in 1999–2000 and research assistant at the University of Rome ‘Tor Vergata’ in 2000–2002. Her research interests lie in the area of wireless mobile networks and content delivery systems.
Emiliano Casalicchio is researcher at the Computer Science Department, University of Roma “Tor Vergata”. He graduated in Computer Engineering in 1998 and received a Ph.D. in Computer Science in 2002, from the University of Roma “Tor Vergata”. His research interests are in the field of modeling and performance evaluation of Computer Networks, Distributed and Web systems. He is currently working also on Quality of Service topics in Grid Computing, and on modeling of complex systems and critical infrastructures.