Dynamic configuration of MAC QoS mechanisms in 802.11 access networks Simona Acanfora1 , Filippo Cacace2 , Giulio Iannello2 , and Luca Vollero1 1
Laboratorio ITeM, CINI, Via Diocleziano 328, Napoli, 80124, Italy
[email protected],
[email protected] 2 Facolt` a di Ingegneria, Universit` a Campus Bio-Medico di Roma Via Emilio Longoni 83, Roma, 00155, Italy {f.cacace,g.iannello}@unicampus.it
Abstract. The purpose of this paper is to explore the dynamic configuration of QoS MAC mechanisms in 802.11 WLANs. First, we discuss how existing mobility protocols and QoS MAC mechanisms can be integrated with minimal change to provide the combined functionalities of continuous connectivity and QoS guarantees. Second, we consider the micromobility protocol Cellular IP integrated with the Acknowledge Skipping mechanism and evaluate through simulation the performance achievable with traffic patterns that force both mechanisms to work in a coordinated manner.
1
Introduction
The notion of Mobile Internet refers to scenarios where mobile users may have access anytime and anywhere to conventional and emerging web applications requiring multimedia and interactive communications. Wireless local area networks (WLANs) based on the IEEE 802.11 standard are a key element to implement these scenarios, representing a typical access network for mobile users. A mobile user moving in an area covered by a WLAN negotiate a contract about the quality of a communication session. Once established the communication session, it should be maintained with the negotiated quality as long as possible. The crucial point is that when a handoff takes place, the contract should be still fulfilled, providing that there are enough resources at the new access point. To meet these requirements both continuous connectivity and resource reservation have to be guaranteed to mobile stations. For what is concerned with the former issue, micromobility protocols have been proposed in recent years to provide low overhead, seamless connectivity in WLANs [1]. As to the resource reservation issue, much work has been done in order to extend the 802.11 MAC protocol with mechanisms supporting traffic differentiation [7], and a new standard has been recently defined [6]. Previous works exist about the integration of mobility and QoS in WLANs [10, 13, 15]. Surprisingly, to the best of our knowledge, none of them considers the integration of mobility and QoS MAC mechanisms proposed in the context of
the 802.11 standard, and in a few cases these mechanisms have been considered in dynamic scenarios. A fundamental point about the integration of 802.11 QoS mechanisms and mobility is the ability to change the configuration of MAC parameters when users move around, i.e. when mobile stations enter or leave wireless cells, and change their traffic patterns. In this paper, we define a general architecture that integrates micromobility protocols and dynamic management of 802.11 QoS mechanisms and evaluate it through simulation in a scenario where a specific micromobility protocol, Cellular IP [16], is integrated with a specific QoS mechanism, ACKnowledgement Skipping [17]. Our results demonstrate that seamless, QoS aware connectivity can be provided to mobile users, and identify directions to fully integrate these functionalities in a comprehensive WLAN management system.
2
Terminology
Recently, a new IETF Working Group, named Control and Provisioning of Wireless Access Points (CAPWAP), started its activity with the goal of finding solutions to problems like configuration, monitoring, control, and management of large scale deployments of IEEE 802.11 networks [19]. The CAPWAP WG has identified a number of functions not included in the 802.11 standard, among which mobility and QoS management are explicitly mentioned, although no further details about their provisioning are given. For this reason we believe that the dynamic management of QoS MAC mechanisms should be considered in the general framework defined by CAPWAP working group. In particular, we will make extensive use of the terminology introduced in [19]. In this paper an 802.11 WLAN consists of one or more wireless cells, called Basic Service Sets (BSSs), consisting of an Access Point (AP) and one or more mobile stations (STAs) associated to the AP. The architectural component used to interconnect BSSs is the Distribution System (DS). The term Access Point (AP) refers to the logical entity that controls a BSS and provides a Distribution System Service (DSS) to the associated STAs. We also introduce the term Wireless Termination Point (WTP) to refer to the physical device that includes the RF antenna and implements the 802.11 physical layer, and the term Access Controller (AC) to refer to a logical entity that may be integrated in the WLAN to provide control and other management functions in a centralized fashion. Traditionally, the AP is a single physical device, including the notion of WTP, and supporting autonomously all the services defined by the IEEE 802.11 standard. Correspondingly the DS is a wired network connecting APs, which may include either L2 (switches), or L3 (routers) devices, or both. This WLAN architecture is called Autonomous and it is characterized by the absence of a clear centralized control point. Alternatively, a number of vendors have recently integrated in their WLAN architecture Access Controllers providing a number of control and management functions. These architectures are called Centralized since they naturally provide a point of coordination and control of the WLAN.
3
Related work
Much work has been done on mobility management in WLANs with multiple WTPs. A number of micromobility protocols have been proposed to handle local movements of mobile hosts [1]. Micromobility protocols provide several services, including host localization, paging, routing and fast handoffs. Although they use different strategies to deal with these issues, the operational principles that govern the most popular micromobility protocols, i.e. Cellular IP, Hawaii and Hierarchical Mobile IP (HMIP) are largely similar. A problem related to seamless mobility is the need of an association procedure before any communication can take place between a moving STA and a new WTP. In the 802.11 standard, this procedure includes WTP discovery, authentication and the actual association, and it is performed exchanging special L2 frames between the STA and the network components in charge of these functions. Seamless mobility remarkably depends on this L2 handoff process and much work has been done so far to analyze and improve its performance [8, 14]. In this paper, we do not explicitly consider problems related to L2 handoff performance, but rather we focus our attention primarily on the provisioning of service guarantees after handoffs rather than during handoffs. Many extensions to the standard 802.11 MAC protocol have been proposed to introduce traffic differentiation in 802.11 networks [6, 7]. For space reasons, we give here only a few details about ACKnowledgement Skipping (ACKS), the technique used in the simulation tests reported in section 5, and therefore relevant for the following discussion. ACKS [17] is based on the following behavior of the standard 802.11 DCF: after sending a packet, a station waits for an Acknowledgement (ACK) frame, and, if the frame is not received within an assigned timeout, it assumes a collision and increases its Contention Window (CW). Hence, if the AP skips the ACK reply to STAs with some probability, these stations will “see” a collision rate higher than the actual one, and will contend with larger CWs, resulting in a reduced channel access capability. This skipping probability, called dropping factor , is the parameter controlling differetiation. Some work is then available on QoS management in presence of mobility. In [13], efficient integration of the handoff protocol with the RSVP signaling process in the wired portion of the network is considered. Performance evaluation is focused on the reservation phase, since it is assumed that communication flows experiment the QoS requested once reservation is successfully completed. In [15], a mobility management protocol is extended to deal with handoff requests towards APs that have not enough resources to provide the desired QoS. Again, the proposal considers QoS mechanisms available at routers, providing traffic differentiation only in the wired portion of the WLAN infrastructure. In [4] a hierarchical QoS architecture that extends the DiffServ QoS modelto mobile hosts in an 802.11 environment is presented. The approach provides a quite comprehensive solution to the integration of mobility and QoS. However, the problem of managing QoS requirements inside BSSs is solved through mechanisms working at layer 3 or above. The advantage is that QoS can be managed
in WLANs compliant with the current 802.11 standard, but the issue of how new QoS MAC mechanisms may be coupled with mobility is completely neglected.
4
Integration of mobility and QoS MAC mechanisms
In this section, we describe an architecture that integrates mobility support with dynamic management of QoS MAC mechanisms to provide service guarantees to mobile users. To make discussion concrete, we basically assume an Autonomous WLAN architecture where mobility is managed by a micromobility protocol, such those mentioned in section 3. Nevertheless, the services we identify can be implemented either in a single physical device or across multiple network elements, and their definition is compatible with either Autonomous or Centralized WLAN architectures. In the following, however, to make discussion simpler, we assume an Autonomous architecture and use the traditional terms of Access Point and Router to refer to the WLAN physical components. As to QoS, we assume that some traffic differentiation mechanisms extending the 802.11 MAC protocol is available, and that service guarantees in the wired part of the WLAN are enforced through overprovisioning, or, alternatively, through admission control and resource reservation based upon solutions similar to those proposed in [4, 13, 15]. According to what required by the 802.11 standard, we assume that beacon packets are periodically broadcast by WTPs, and STAs listen to them and initiate handoff based on signal strength measurements. To perform a handoff, a STA tunes its radio transmitter to the new WTP, and enters the authentication and association procedures envisaged by the standard. After the association completes, the new WTP continues the handoff procedure according to the specific micromobility protocol considered. To implement dynamic QoS management, each AP is controlled by a QoS manager , which carries out at least the following minimal set of functions: (i) acquisition of information about traffic patterns and QoS requirements of the moving station when a handoff initiates; (ii) computation of the MAC parameter configuration that best matches traffic offered by the STAs active in the BSS; (iii) actual setting of MAC parameters of the APs and of all active stations in the BSS; (iv) setting of L3 mechanisms to provide QoS guarantees to downstream flows. In the following, we describe how these functions can be implemented using existing protocols with minimal extensions. QoS information acquisition. We assume that in the WLAN there is a global session directory where a traffic pattern description is maintained for every active communication session. Any STA starting a new session with QoS guarantees has to register to the session directory, providing sufficient information to characterize its traffic patterns. When the new association is started, the MAC layer notifies the QoS manager and informs it about the identity of the moving station. Using this information,
the QoS manager accesses the session directory and get the corresponding traffic information. Note that the proposed scheme is robust, since a default traffic pattern (e.g. best effort traffic) may always be assumed if, for any reasons, the traffic pattern of a moving station is not available (e.g. a STA just entered the WLAN or it is not registered in the session directory). Computing MAC configuration. Once traffic information is acquired, the QoS manager can carry out its second task, i.e. to compute the optimal MAC parameter configuration. The computation can be either model -based, if analytical models are available for wireless channel characterization, or table-based, if optimal parameter values have been derived off-line through simulation or experimentation. Although this problem, in its general form, is still open for most QoS MAC mechanisms, several useful results are available in the literature, and there is much ongoing research trying to extend these results. Using either model or off-line characterization, the QoS manager can determine if QoS requirements of handoff requests can be satisfied by the resources available in the BSS. This means that the manager might perform also access control over handoff requests and notify the micromobility protocol when the requested service guarantees cannot be met. To fully exploit this ability, however, the micromobility protocol should be modified introducing the concept of QoS-conditionalized handoff [15]. In this context, we prefer to follow a simpler and conservative approach. In case the request cannot be satisfied, the QoS manager computes the best possible configuration, and the moving station will experiment a service degradation if the handoff is completed successfully. Setting MAC configuration. After having computed a new configuration of MAC parameters for the stations in the BSS, the QoS manager can force BSS configuration by broadcasting configuration information to all active stations. Note that this step is generally needed since most QoS MAC mechanisms require a coordinated configuration of all stations in a BSS. Setting L3 mechanisms. Finally, since downstream flows may have different QoS requirements, they should receive a differentiated treatment, likewise upstream flows transmitted by the mobile hosts. All downstream flows are transmitted by the same 802.11 device (the AP), so, in most cases, differentiation among these flows cannot be satisfactorily managed at layer 2. A reasonable solution is therefore that the MAC parameters of the AP are set in such a way that enough resources are made available to all downstream traffic as a whole. QoS constraints on individual downstream flows can then be enforced through L3 mechanisms, i.e. through mechanisms similar to those implemented in QoS-aware routers, based on traffic conditioning. The functionalities just described can be implemented by the architecture shown in figure 1, where are reported the functional building blocks and how they interact in the control plane. In the figure functions are mapped onto the two
Fig. 1. Architecture for integration of mobility and QoS.
components of the WLAN: the access network, consisting of the the distribution system and the access points, and the mobile stations. This means that in real WLANs there are multiple instances of some blocks. For instance, the MAC and physical layer on both sides are replicated in each AP and in each STA, as well as, the IP layer is replicated in each STA. Other blocks (e.g. the QoS manager) may or may not be replicated, according to performance considerations and to which WLAN architecture (Autonomous or Centralized) is chosen. A distinctive element of the architecture is the presence of a session layer entity in both the access network and in the STAs that implement the global session directory needed to relate mobility events with the update of MAC parameters. It is worth noting that most of the interactions reported in the figure rely on existing or forthcoming protocols. Messages between MAC entities on both sides are the control frames defined by the standard to perform L2 handoffs. Communications between MAC/IP blocks and QoS managers in the access network can be implemented using the forthcoming CAPWAP protocol that is still in a proposal stage, but that explicitly includes messages to notify mobility events and set 802.11e parameters [2, 5]. Finally, session registrations in the global session directory can use existing message formats to specify traffic characteristics (see for instance [12]). Before closing the section, we briefly make some remarks about the actual mapping of the functional elements shown in figure 1 onto physical network devices. A first issue concerns the alternative between Autonomous and Centralized WLAN architectures. In the former case a natural choice is to replicate the QoS manager on each AP, whereas in the latter case it can be mapped on the AC, especially if the WLAN is a variant of the Centralized architecture called Split-MAC. Similar considerations can be done for the IP block that is located at the gateway connecting the DS with the Internet, but that can be replicated if more routers are present in the DS. As to the global session directory, the nat-
ural choice is a centralized implementation, located at the gateway or at the AC, according to the physical topology of the DS. Nevertheless, more sophisticated implementation schemes may be chosen to improve scalability of QoS management functions with the WLAN size. For instance the global session directory can be implemented as a hierarchy of modules to keep local the interactions with STAs and QoS managers.
5
Evaluation
In this section, we present a set of simulation results aiming to provide a first evaluation of our architecture. Simulation environment. To setup the simulation environment we needed to choose a specific micromobility protocol and a specific set of QoS mechanisms extending the standard 802.11 MAC protocol. As a micromobility protocol we chose Cellular IP. Indeed, much previous work on the integration of mobility and QoS support assumes Cellular IP as a micromobility protocol, which makes reasonable here its choice to study the integration of mobility and MAC parameter configuration. To limit the complexity of this first analysis, we then chose ACKnowledgement Skipping (ACKS) as a QoS mechanism because it is very simple to configure (just one parameter at the AP when two traffic classes are considered), and it does not introduce any configuration delay (no communication between the AP and the STAs is required). The availability of a patch to add this mechanism to the simulator we used was also a motivation to use ACKS in this first evaluation. As a simulation environment we chose the Berkeley Network Simulator, ns-2, that has been widely used in the evaluation of both micromobility protocols and QoS MAC mechanisms. In particular we used the implementation of Cellular IP provided by the Columbia IP Micromobility Suite (CIMS) and the patch implementing the ACKS mechanism mentioned above. The choice of using CIMS requires the use of ns-2 ver. 2.1b6, which has a number of minor limitations, including the limit of 2 Mbps for the 802.11 wireless link, since this version of ns-2 does not simulate accurately the 802.11b physical layer. In our simulations we used multiple APs interconnected to a network of Cellular IP routers. As just stated the radio links have a bitrate of 2 Mbps, whereas the wired links have a bitrate of 10 Mbps. In all experiments, the traffic pattern has the following general characteristics. Multiple mobile hosts belonging either to a best effort (BE) class or to a high priority (HP) class are associated to each AP. All hosts communicate with a correspondent node (CN) located somewhere outside the access network. BE STAs generate an upstream flow towards the correspondent node consisting of constant bitrate data stream of 200 Kbps. HP STAs generate an upstream flow towards the CN and, at the same time receive from it a downstream flow. Both flows are constant bitrate data streams of 200 Kbps each. All flows consist of UDP packets and we use as a main performance index the goodput of these flows, i.e. the bitrate corresponding to
successfully received packets. During each simulation, one or more mobile hosts (either BEs or HPs) move through the access network performing handoffs. If not stated otherwise, hosts move at a speed of 5 m/s. It is worth noting that this traffic pattern is quite different than the one usually used to evaluate micromobility protocols or, alternatively, QoS MAC mechanisms. On the one hand, multiple downstream flows are introduced having the same bitrate than the corresponding upstream flows. This roughly models the traffic generated by symmetric multimedia applications like VoIP or videoconferencing. On the other hand, communications of HP hosts (both upstream and downstream) must share the wireless channel with a variable interfering traffic, either because BE hosts leave and join the associated AP during the simulation, or because HP hosts move around and perform handoffs. These rich traffic patterns are needed in our scenarios because we are interested in evaluating how mobility and QoS mechanisms work in a coordinated manner. Indeed, in this scenario, the AP requires much more radio resources than each single STA in the BSS. Hence, the presence of interfering traffic quickly leads to a degradation of the goodput of downstream flows if no differentiation is applied to best effort traffic. Finally, to test how micromobility protocols and QoS MAC parameter configuration integrate in a dynamically changing scenario, we needed a method to compute the new MAC parameters configuration of the AP when there are changes in the workload because STAs move around. Since we use ACKS to differentiate between two traffic classes, we had to determine the dropping factor to be applied to frames transmitted by the BE STAs in every traffic condition of interest. Even though, a model is available to configure ACKS [18], it can be used only when all the STAs are in saturation conditions. Unfortunately, this is not the case of our traffic scenarios where either no differentiation is needed because the workload does not saturate the channel, or only the BE stations are in saturation conditions because base and HP stations transmit at assigned bitrate. For this reason, we used a table-based approach to MAC parameter configuration, performing off-line simulations to determine the dropping factors to be used in each case. Results. For space reasons, as a representative scenario among the many simulations we performed, we consider here a configuration like that depicted in figure 2: a HP STA moves from an AP where it is the only STA to an AP where there are already two HP and three BE STAs. We focus our attention on the goodputs and delays of the moving HP STA, and consider both the case when no dropping factor is applied at the final AP, and the case when it is properly updated according the table approach mentioned above. Figures 3 and 4 show the goodputs and delays of the moving HP STA. Upstream delays are not shown because they are always very low. For goodputs, data refer to mean values taken every 10 s, excluding the range 77.5–90 s, where data refer to mean values taken every 0.5 s to point out the transient behavior. Approximately, at simulation time 79 s the moving STA enters the second BSS, and at time 84.2 s it performs the handoff. For upstream flows (figure 3-a
Fig. 2. Simulation scenario.
and 3-b), the goodput is essentially constant, with a short break in the communication flow during the handoff3 . For downstream flows, the goodput is remarkably less than offered traffic when no dropping factor is applied (figure 4-a), whereas it remains substantially constant when the proper dropping factor is applied (figure 4-b). Note that this behavior is expected, since HP downstream flows are much more affected by the interfering traffic than HP upstream flows. The effectiveness of the dynamic reconfiguration of the MAC parameters is also apparent by looking at the downstream delay (figure 4-c and 4-d), which is dramatically reduced when the right dropping factor is applied. The above data, however, show that there is a short interruption in the communication flows even when the dropping factor is applied, i.e. when resources are reserved to protect HP flows. This is due to the delay induced by the congestion at the target BSS in exchanging the ARP-like messages used by ns-2 to model STA re-association. We therefore repeated the simulation changing the dropping factor at 79 s, i.e. just when the moving STA enters the coverage area of the second AP. Results are reported in figure 3. With respect to figures 3-b, 4-b, and 4-d, both upstream and downstream goodputs are remarkably improved and downstream delay further reduced, leading to an almost seamless handoff with QoS guarantees. Although the L2 handoff process is not precisely modeled, the observed behavior is coherent with available data on L2 handoffs (see for instance [9]). Our results suggest that an anticipated configuration of QoS MAC parameters is desirable. Although a modification to mobility protocols goes beyond the scope of this paper, it is worth noting that such an anticipated configuration could be not too difficult to implement, e.g. by enabling the moving STA to perform some kind of pre-association to the target AP before the actual handoff takes place. 3
This is when the QoS manager of the target AP is notified that the new mobile STA has performed the association.
0.4 0.35
0.3
goodput [Mbps]
goodput [Mbps]
0.4 0.35
0.25 0.2 0.15 0.1
0.2 0.15 0.1 0.05
0.05 0
0.3 0.25
70
60
80
90
100
0
110
60
70
80
time [s]
90
100
110
time [s]
(a) goodput without dropping factor
(b) goodput with variable dropping factor
0.4
0.4
0.35
0.35
0.3
goodput [Mbps]
goodput [Mbps]
Fig. 3. Upstream flows of the moving HP STA.
0.25 0.2 0.15 0.1 0.05 0
0.3 0.25 0.2 0.15 0.1 0.05
60
70
80
90
100
0
110
60
70
80
time [s]
1.5
1.5 delay [s]
2
delay [s]
100
110
(b) goodput with variable dropping factor
2
1
0.5
0
90 time [s]
(a) goodput without dropping factor
1
0.5
60
70
80
90
100
0
110
60
70
80
time [s]
90
100
110
time [s]
(c) delay without dropping factor
(d) delay with variable dropping factor
0.4
0.4
0.35
0.35
0.3
0.3
0.25 0.2 0.15 0.1 0.05 0
2
1.5
0.25
delay [s]
goodput [Mbps]
goodput [Mbps]
Fig. 4. Downstream flows of the moving HP STA.
0.2 0.15
1
0.5
0.1 0.05
60
70
80
90
100
time [s]
(a) goodput upstream
110
0
60
70
80
90
100
110
time [s]
(b) goodput downstream
0
60
70
80
90
100
time [s]
(c) delay downstream
Fig. 5. When the dropping factor change is anticipated at 79 s.
110
These experiments confirm the possibility to guarantee QoS requirements after the handoff when QOS MAC mechanisms are dynamically configured according to the offered traffic patterns. This result clearly holds at steady state (i.e. before and after handoffs). As to the behavior during handoffs, we observed that an anticipated BSS configuration might improve handoff performance. This result however is only qualitative since handoff is only roughly modeled in ns2. Moreover, it cannot be so easily generalized to other scenarios, especially if different QoS MAC mechanisms are considered. Indeed, very few data are available about the dynamic behavior of these mechanisms and more study is needed before drawing general conclusions.
6
Conclusions and future work
In this paper we have considered the problem of how to manage mobility support and QoS MAC parameter configuration in an integrated fashion in 802.11 WLANs. In this respect, we have proposed a cross-layer management architecture characterized by two main functional elements operating at session and MAC layer, respectively. The architecture’s control plane basically relies on existing or forthcoming mechanisms and protocols and it may be integrated with CAPWAP functions for WLAN management. We have also presented a preliminary evaluation of benefits deriving from dynamic management of a simple QoS MAC mechanism through simulation. Our results demonstrate that the integration of these mechanisms may provide almost seamless connectivity and service guarantees to mobile users in traffic scenarios richer than those considered so far in the literature. Full integration of mobility and dynamic QoS management into 802.11 WLANs still requires, however, much work in at least two directions. First, the functions we have identified have to be better refined and experimented on a real testbed. For instance, we have not experimented L3 mechanisms to manage QoS of individual downstream flows (or classes of them). Second, a more sophisticated approach to QoS MAC parameter configuration would be desirable. On the one hand, ACKS should be substituted by the IEEE 802.11e standard, and, on the other hand, model-based configurations should be considered in place of tablebased solutions when possible.
Acknowledgements This work was funded by MIUR with the FIRB 2001 project “WEB-MINDS” and the PRIN 2004 project “QUASAR”.
References 1. A.T. Campbell, J. Gomez, S. Kim, Z.L. Turanyi, A.G. Valko’, and C.-Y. Wan, Internet micromobility, Journal of High Speed Networks, 11, 177–198, 2002.
2. P. Calhoun, B. O’Hara, R. Suri, N. Cam Winget, S. Kelly, M. Williams, S. Hares, Light Weight Access Point Protocol, Internet-Draft, draft-ohara-capwap-lwapp03.txt, June 2005. 3. H. Duong, A. Dadej, S. Gordon, Proactive Context Transfer and Forced Handover in IEEE 802.11 Wireless LAN Based Access Networks, Mobile Computing and Communications Review , Vol. 9, No. 3, 2005. 4. J.A. Garcia-Macias, F. Rousseau, G. Berger-Sabbatel, L. Toumi, and A. Duda, Quality of Service and Mobility for the Wireless Internet, Wireless Networks, 9, 341–352, 2003. 5. S. Govindan (Editor), ZH. Yao, WH. Zhou, L. Yang, H. Cheng, Objectives for Control and Provisioning of Wireless Access Points (CAPWAP), Internet-Draft, draft-ietf-capwap-objectives-04.txt, Sept. 2005. 6. IEEE 802.11e/D6.0. Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: Medium Access Control (MAC) Enhancements for Quality of Service (QoS). Draft Supplement to IEEE 802.11 Standard, October 2003. 7. A. Lindgren, A. Almquist, and O. Schel´en, Quality of Service Schemes for IEEE 802.11 Wireless LANs: An Evaluation. Special Issue of the Journal on Special Topics in Mobile Networking and Applications (MONET) on Performance Evaluation of Qos Architectures in Mobile Networks, 8:223–235, June 2003. 8. A. Mishra, An Empirical Analysis of the IEEE 802.11 MAC Layer Handoff Process, ACM Computer Comm. Review, 33, 2, pp. 93–102, 2003. 9. N. Montavont and T. Noel, Analysis and Evaluation of Mobile IPv6 Handovers over Wireless LAN, Mobile Networks and Applications, 8, pp. 643–653, 2003. 10. B. Moon, A.H. Aghvami, Quality-of-Service Mechanisms in All-IP Wireless Access Networks, IEEE Journal on Selected Areas in Communications, vol 22, no. 5, June 2004. 11. S. Pack, H, Jung, T. Kwon, Y. Choi, SNC: A Selective Neighbor Caching Scheme for Fast Handoff in IEEE 802.11 Wireless Networks, Mobile Computing and Communications Review , Vol. 9, No. 4, 2005. 12. Partridge, C., A Proposed Flow Specification, Internet RFC 1363, July 1992 13. S. Paskalis, A. Kaloxylos, E. Zervas, and L. Merakos, An Efficient RSVP-Mobile Internetworking Scheme, Mobile Networks and Applications, 8, 197–207, 2003. 14. S. Shin, A.G. Forte, A.S. Rawat, H. Schulzrinne, Reducing MAC Layer Handoff Latency in IEEE 802.11 Wireless LANs, Procs. of MobiWac’04 , Oct. 1, Philadelphia, Pennsylvania, USA, 2004. 15. S. Sroka and H. Karl, Performance Evaluation of a QoS-Aware Handover Mechanism, Procs of the 8th IEEE Int. Symp. on Computers and Communication (ISCC’03), 2003. 16. A.G. Valk` o. Cellular IP: a new approach to Internet host mobility. ACM SIGCOMM Computer Communications Review, 29(1), 50–65, 1999. 17. L. Vollero, G. Iannello, ACK Skipping: enabling QoS for multimedia communications in WiFi hot spots, Int. Journal of High Performance Computing and Networking, 2005. 18. L. Vollero, Providing throughput guarantees in WLANs using ACKS, Procs of the 6th IEEE International Symposium on a World of Wireless Mobile and Multimedia Networks, Taormina, Italy, 13-16 June 2005. 19. L. Yang, P. Zerfosm, E. Sadot, Architecture Taxonomy for Control and Provisioning of Wireless Access Points (CAPWAP), Internet RFC 4118, June 2005.
