Cross-layer protocol optimization for satellite communications networks

INTERNATIONAL JOURNAL OF SATELLITE COMMUNICATIONS AND NETWORKING Int. J. Satell. Commun. Network. 2006; 24:323–341 Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/sat.853

Cross-layer protocol optimization for satellite communications networks: A survey Giovanni Giambene1,*,y and Sastri Kota2,z 1

Dipartimento di Ingegneria dell’Informazione, Universita` degli Studi di Siena, Via Roma, 56, 53100 Siena, Italy 2 Harris Corporation – GCSD, 1134 East Arques Avenue, Sunnyvale, CA 94086, U.S.A.

SUMMARY Satellite links are expected to be one important component of the next-generation Internet. New satellite system architectures are being envisaged to be fully IP based and support digital video broadcasting and return channel protocols (e.g. DVB-S, DVB-S2 and DVB-RCS). To make the upcoming satellite network systems fully realizable, meeting new services and application requirements, a complete system optimization is needed spanning the different layers of the OSI, and TCP/IP protocol stack. This paper deals with the cross-layer approach to be adopted in novel satellite systems and architectures. Different cross-layer techniques will be discussed, addressing the interactions among application, transport, MAC and physical layers. The impacts of these techniques will be investigated and numerical examples dealing with the joint optimization of different transport control schemes and lower layers will be considered referring to a geostationary-based architecture. Our aim is to prove that the interaction of different layers can permit to improve the higher-layer goodput as well as user satisfaction. Copyright # 2006 John Wiley & Sons, Ltd. Accepted 5 July 2006 KEY WORDS:

IP-based satellite systems; radio resource management; transport layer; cross-layer design

1. INTRODUCTION Satellites may indeed provide wide coverage area without the need for deploying complex infrastructures, thus representing a viable choice for scenarios ranging from rural areas to highspeed trains, to emergency/crisis areas. New satellite system architectures are being envisaged to be fully IP based and support digital video broadcasting and return channel protocols, e.g. DVB-S, DVB-S2 and DVB-RCS [1].

*Correspondence to: Giovanni Giambene, Dipartimento di Ingegneria dell’Informazione, Universita` degli Studi di Siena, Via Roma, 56, 53100 Siena, Italy. y Email: [email protected] z Email: [email protected]

Copyright # 2006 John Wiley & Sons, Ltd.

324

G. GIAMBENE AND S. KOTA

The increasing demand for multimedia broadband services and high-speed Internet services via satellite requires the study of new protocols as well as their interactions. Satellite resources are costly and satellite communications impose special constraints with respect to terrestrial systems in terms of attenuation, propagation delays, fading, etc. To make the upcoming satellite network systems fully realizable, meeting new services and application Quality of Service (QoS) requirements, many technical challenges have to be addressed that are constrained by the layered protocol architecture, typical of both the ISO/OSI reference model and the Internet protocol suite. According to the protocol stack paradigm, a protocol solves a specific problem by using the services provided by modules below it and gives a new service to upper layers. The main disadvantages of such approach can be detailed as follows: *

*

*

User service requirements are provided by the communication system at the top level. The hierarchy and the overall system performance is, however, dependent upon the lower-layers. Information is lost during the layer-by-layer conversion, which is particularly critical in the satellite scenario, where the physical layer imposes special constraints (due to packet losses, available net bandwidth, etc.) to higher layers. Layers are independently optimized and this leads to inefficiency and redundancy.

The layered approach is well suited for large network sizes, providing layer management flexibility. However, the current layered approach does not provide sufficient adaptability in the case of hybrid network architectures (e.g. wireless or 3GPP / satellite environment) that need a large-scale adaptability. In particular, a strict modularity and layer independence may lead to non-optimal performance in IP-based satellite communication systems. Since both radio resources and power are strongly constrained, a network optimization is needed. In this framework, a cross-layer air interface design is needed where interactions among even nonadjacent protocol layers are introduced for a system optimization in order to achieve a better adaptation to system dynamics [2]. Without a cross-layer design of the air interface we can expect a loss of system efficiency as detailed by the examples described below: *

*

*

*

IP packets lost due to errors induced by wireless channel are interpreted as signals of congestion at the TCP level, thus lowering the bit-rate (congestion window, cwnd). A long time is needed to recover after a loss event, especially when multiple losses occur that may cause a TCP timeout. This is particularly critical in a satellite scenario where it takes several Round Trip Times (RTTs) before recovering the TCP goodput at the same level as before the loss event. Radio resources can be also allocated to mobile users that have bad channel conditions, thus causing a waste of capacity. Intra- and inter-satellite handoff procedures (with consequent re-routing) in nongeostationary satellite constellations can take a long time that can lead to IP connection interruption with the risk of higher layer timeouts. Packet overflow can be experienced at the application level (e.g. streaming or conversational traffic) during bad channel conditions due to reduced available information bit-rate.

System efficiency is an important requirement for operators of satellite communications to provide services at competitive costs, allowing a mass-market penetration of satellite Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

325

communications. Whereas, QoS support is mandatory for end users who do not care about resource utilization, but expect a good service level. System optimization and QoS support are typically conflicting needs. For instance, the best QoS condition for delay-intolerant variable bit-rate traffic is to have permanently allocated bandwidth corresponding to the peak traffic, thus causing significant system inefficiency. These conflicting needs can be solved by means of a suitable cross-layer system design and by exploiting the multiplexing effect of packet data traffic. Moreover, the impact of fading due to rain and other environmental effects has not been deeply addressed to assess the end-user performance. Hence, the cross-layer design is an interesting research field posing new challenging target in understanding an end-to-end system level performance and sensitivities to various parameters such as bit error rate, packet error rates, delays, delay variations. The cross-layer design research field is in its infant stage and several challenges should be addressed in system implementation and satisfying end-to-end user service level agreements. The aim of this paper is provide a survey of cross-layer methods and to prove the related gains that can be achieved in performance and efficiency by jointly optimizing the behaviour of diverse protocol layers. This paper is organized as follows: Section 2 presents a brief overview of the cross-layer design approaches for satellite communication networks. A detailed discussion on the interactions between the various layers (e.g. physical layer with MAC, MAC with network layer and interactions of physical layer protocol with TCP and application layer) will be provided in Section 3 where a bibliographic survey and some recent research results will be surveyed. Section 4 provides a simulation model and numerical results for evaluating the impact of lower layer parameters on the TCP goodput performance. We conclude this paper with Section 5.

2. CROSS-LAYER DESIGN OVERVIEW This section provides an overview on cross-layer design approaches and describes some implementation challenges, which require future investigations. 2.1. Satellite protocol reference model The European Telecommunication Standards Institute Technical Committee - Satellites Standard Earth Stations and Systems/Broadband Satellite Multimedia (ETSI TC-SES/BSM) defined an IP-based satellite network architecture including lower layer air interfaces [3]. Figure 1 shows such protocol architecture where lower layers depend on satellite system implementation (satellite-dependent layers) and higher layers are those typical of the Internet protocol suite (satellite-independent layers). These two layers of protocols are interconnected through the Satellite-Independent-Service Access Point (SI-SAP) interface. The standardization foresees only a small number of generic functions that cross the SI-SAP. In particular, these functions are address resolution, resource management, traffic classes QoS. Therefore, it is important that a cross-layer method (linking satellite-dependent and satelliteindependent layers) be implemented taking into account the above protocol structure and the SI-SAP interface. In particular, suitable primitives must be foreseen in the standardization process to support such an extended signalling. Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

326

G. GIAMBENE AND S. KOTA

Applications

External Layers UDP

TCP

Other

IPV4 / IPV6 Satellite Independent

Satellite Independent Adaptation Functions SI-SAP Satellite Dependent Adaptation Functions

Satellite Dependent

Satellite Link Control (SLC) Satellite Medium Access Control (SMAC) Satellite Physical (SPHY)

Figure 1. Satellite protocol stack architecture by ETSI TC-SES/BSM.

2.2. Cross-layer approaches The cross-layer design requires interfaces between non-adjacent layers, thus violating the classical abstraction levels of the Internet protocol stack. Although interfaces between adjacent layers are in general preferable, there is need for direct interaction between non-adjacent layers. In general, a layer should be aware of the other layers of the protocol stack. Cross-layer information can be exchanged from higher to lower layers (top-down approach) or from lower to higher layers (bottom-up approach), as shown in Figure 2. Co-ordination of both exchanges is necessary to avoid loops in the system and oscillating behaviours. In a classical layered approach the exchange of information can be only performed among adjacent layers through ‘send’ and ‘receive’ primitives. Non-adjacent layers can only communicate involving intermediate layers. The novelty of the cross-layer approach is to allow the exchange of control information (signalling) among non-adjacent layers [4]. Let us refer to the example in Figure 3, where a ‘get function’ can be executed by higher-layer protocols to acquire the internal state of lower layer protocols. Moreover, a ‘set function’ can be adopted by higher-layer protocols to change the state of lower layer protocols. Two basic cross-layer design approaches are: *

*

Implicit cross-layer design: there is no exchange of information among different layers during operation, but during the design phase all the layer interactions are taken into consideration to perform a joint protocol optimization. Explicit cross-layer design: signalling interactions among (non-)adjacent protocol levels are employed so that dynamic adaptations can be simultaneously performed at different layers.

Different solutions have been proposed to support the cross-layer exchange of signalling information (explicit cross-layer design). An interesting method has emerged in several papers [4–7]. In particular, a ‘global co-ordinator’ of different layers is considered to acquire status information from different protocols to store it in a shared memory and to set the internal state of the protocols on the basis of suitable events. The co-ordinator can be coincident with a Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

327

Application layer Transport layer Top-down signaling

Bottom-up signaling

Network layer Data link layer Physical layer

Figure 2. Cross-layer interactions.

Application layer ….. Data link layer send

get

receive Physical layer

set status

Figure 3. Example of cross-layer exchange of control commands (signalling).

protocol layer; for instance, it can be at the application layer (so that we have an applicationcentric cross-layer implementation) or at the MAC layer (hence, we have a MAC-centric crosslayer scheme). 2.3. Design challenges Cross-layer design is an attractive new approach to improve the performance of wireless systems. This is particularly important in the satellite scenario where many constraints and costly challenging technical solutions need to be employed to optimize the efficiency of the system. The implementation of cross-layer schemes typically requires ad hoc solutions that have to be tailored to a specific satellite communication scenario. Implicit cross-layer optimization seems relatively simple to be adopted and can be implemented without modifying the current protocol structure. Some implicit cross-layer design methods will be described later in Section 3, and Section 4 will contain some numerical results showing the potential impact of PHY adaptability according to TCP goodput performance. These examples entail a system simulation and/or analytical approach to find optimal settings depending on different operating conditions and system constraints. Even greater adaptivity to system dynamics could be achieved by adopting explicit cross-layer techniques that, however, would entail the modification (or redesign) of current protocol stacks. Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

328

G. GIAMBENE AND S. KOTA

It is also reasonable that both implicit and explicit methods work in tandem. In particular, an interesting possibility is to describe the whole system using mathematical models with operating and conditional constraints; on the basis of this, the implicit optimization can take place in the bounded space and with the policies in place. Once obtained the optimal (normal) operating condition, the model should be ported into the real system and changes in operating parameters will vary optimality on the basis of an explicit technique that is able to track system dynamics. It is important now to focus on the implication that the cross-layer design can have on network management and end-to-end QoS support. In particular, we consider here that the network has to be made aware of cross-layer issues. Policy updates (depending on satellite network congestion conditions, adverse atmospheric situations, emergency situations, etc.) could be elaborated in the control part of the satellite network (for instance, the Network Control Center, NCC). Such information needs to be distributed to the different nodes of the network and, in particular, to user terminals by means of appropriate signalling. The control information exchange can be performed in a static or dynamic way by means of the Common Open Policy Service (COPS) protocol, a signalling scheme defined in IETF RFC 2748 [8]. This signaling process can be used to notify the different Eb/N0 thresholds to change among transmission modes (PHY), the scheduling technique and the priorities (MAC), the user mobility protocol and the prediction of the next cell to be reached, etc. Since network layer vendors could be different from the service operator, it is essential that SLA (Service Level Agreements) negotiations are undertaken between operators and sub-layer vendors to ensure that end-to-end QoS performance levels are meet as well as SLAs are fulfilled. These issues need further concern in order to optimize the implementation approaches.

3. PROTOCOL LAYER INTERACTIONS This section provides a description of cross-layer interactions that could be exploited for the design of satellite communication systems. 3.1. PHY and MAC layer interactions It is important that the resource allocation schemes (i.e. the access protocols and the scheduling techniques) be aware of the physical layer behaviour and the related adoption of Adaptive Coding & Modulation (ACM) to perform optimal choices. These considerations can be well suited to the DVB-S2 scenario [9]. In particular, DVB-S2 benefits from recent developments in channel coding such as LDPC codes combined with a variety of modulation formats, e.g. QPSK, 8PSK, 16APSK and 32APSK. For interactive applications, such as Internet navigation, the adoption of ACM allows optimizing the transmission parameters of each user on a frameby-frame basis, depending on path conditions, under closed-loop control via a return channel (terrestrial link or via satellite). During rain fades, more conservative transmission modes are employed (i.e. a lower-order modulation level and a lower coding rate), thus reducing the information bit-rate available for users. The resource management scheme should take allocation decisions on the basis of the knowledge of the current information bit-rate available for each user. Cross-layer techniques for DVB-S/DVB-S2/DVB-RCS systems have been proposed in References [10–12], where a master station dynamically updates the portion of bandwidth Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

329

assigned to slave stations (running several applications) on the basis of their bandwidth requests and local channel conditions (e.g. weather conditions). In the considered Ka band (20–30 GHz), atmospheric events have a significant impact on the level of the signal. The effect of the physical layer, where fade countermeasures are applied to contrast fading attenuation, is taken into account as a reduction in the net available bandwidth. In the light of the cross-layer approach, the above papers propose an optimization problem for the bandwidth allocation taking into account fade countermeasure (code and rate adaptation) and redundancy coefficients; such approach permits to achieve significant capacity gains at the transport level. Let us now refer to Reference [13], where cross-layer methods involving PHY and MAC layers have been studied for unicast services on DVB-S2. This paper presents a cross-layer technique for the design of the forward link packet scheduler that introduces fairness as a tunable parameter. This approach makes it possible to adapt dynamically the scheduler behaviour depending on the channel conditions in order to guarantee fairness. The proposed algorithm also supports differentiation of services that complies with the requirements for implementing QoS. 3.2. PHY and network layer interactions In the traffic management for an IP-based satellite system, user mobility should be adequately taken into account. For geostationary satellite-based systems providing services to planes or fast trains as well as for non-geostationary satellite constellations, a connection may incur in several handoffs (i.e. change of satellite antenna beam coverage) during its lifetime. We distinguish two cases: (i) handoffs occurring between two beams of the same satellite (intra-satellite handoff); (ii) handoffs occurring between two beams of different satellites (inter-satellite handoff). Intra- and inter-satellite handoffs need to be properly managed to avoid high delay experienced by IP traffic. Hence, layer 2 protocol should provide a prioritized traffic management during user handoff phases. Moreover, efficient mobility protocols have to be employed at layer 3 to prevent excessive delays in redirecting the data flows during handoff phases. Such delays would have a negative impact at the transport level with the risk of timeouts and significant TCP goodput reduction. In the non-geostationary case, mobility protocols could use motion prediction algorithms, mainly based on the known motion characteristics of the satellite constellation, in order to define in advance the cell where user is moving to, which reduces the handoff delays. The cross-layer issues in this field are quite interesting and require further research efforts. 3.3. MAC and network layer interactions In IP-based networks, QoS provisioning can be archived using according to two approaches: Integrated Services (IntServ) and Differentiated Services (DiffServ). It is important that resource allocation scheme at layer 2 manages traffic in a way that is compatible with that adopted at layer 3 using either IntServ or DiffServ [14]. Following IntServ, resources are reserved by the RSVP protocol for each flow through the network. Current implementations of IntServ allow a choice of guaranteed service or controlled-load service. Whereas, the DiffServ approach achieves scalability by aggregating traffic into classes that are conveyed by means of IP-layer packet marking using the type of service (ToS) field in the IPv4 header or the differentiated service (DS) field in the IPv6 header. DiffServ prescribes treatment for aggregated traffic rather than for micro-flows and forces much of the complexity out of the core network into edge devices, which process lower volumes of traffic and lower numbers of flows. Assured Forwarding (AF), Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

330

G. GIAMBENE AND S. KOTA

Expedited Forwarding (EF) and background traffic flows of the DiffServ scheme should have an adequate mapping at layer 2. 3.4. MAC and transport layer interactions TCP is the prominent transport layer protocol, used for assuring a reliable end-to-end data delivery over the Internet. TCP controls at least 85% of all Internet traffic and runs only at the end-hosts. It is designed to utilize the available bandwidth for the source–destination pair in a fair and efficient way. However, the standard TCP congestion control mechanism is known to perform poorly over satellite links, due to both the large RTT value and the possibly high packet error rates [15]. To improve TCP performance in an error-prone channel, we can consider TCP Reno that detects the loss of one segment as a consequence of three received duplicated ACKs. When three duplicated ACKs are received, TCP Reno halves the congestion window (cwnd) value and employs two phases namely Fast Retransmit and Fast Recovery to retransmit the lost data [16]. Different cross-layer mechanisms can be used in a satellite IP network to improve the TCP performance. Some interesting examples are detailed below. Let us refer to a DVB-S/DVB-RCS scenario: when a group of terminals, using TCP as transport protocol, are connected to the system gateway acting as NCC, a capacity allocation strategy which does not take into account the TCP window fluctuating behaviour may lead to an inefficient and unfair sharing of resources. Note that standard Demand-Assignment Multiple Access (DAMA) schemes guarantee an optimal network utilization, but impacting TCP endto-end performance by introducing additional delay contributions to the RTT due to the exchange of capacity request/allocation message signalling. The cross-layer approach proposed in Reference [17], synchronizes the resource requests with the TCP congestion window trend in order assign/remove dynamically capacity to terminals on the basis of their actual needs. This approach calls for a TCP-driven dynamic bandwidth allocation operated at layer 2. Thus, it is capable of reducing the queuing delay at layer 2 and congestion phenomena with consequent RTO expirations. In conclusion, such cross-layer technique allows lowering the average file transfer time and achieving a fair resource sharing among competing flows. In split scenarios [18], the end-to-end TCP semantics is broken. The satellite link is isolated by the terrestrial segment and interconnecting routers called Performance Enhancement Proxies (PEPs) are used to close the TCP flow. PEPs are typically implemented at transport or application layer. Examples of transport layer PEPs are TCP spoofing and TCP connection-split proxies. In both PEP types, the goal is to shield high-latency or lossy satellite network segments from the rest of the network, in a transparent way to applications. A critical issue in PEP is the design of buffers and related management rules and sizes. Interesting proposals envisage the adoption of Active Queue Management (AQM) at the MAC layer for improving the TCP performance. In AQM, when the router determines that the bandwidth is fully utilized, packets are dropped even when the queue is not full so as reduce the buffer congestion and the data injection rate of the TCP sender [19]. Another interesting study on the adoption of PEP is performed in Reference [20] that proposes a novel cross-layer congestion control method, called CSACK, for TCP SACK split-connections applied to a satellite link between two PEPs. The key feature of CSACK is local congestion notification from the MAC layer to the TCP layer in the PEP, based on the current MAC buffer occupancy managed by the RED algorithm. The proposed CSACK Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

331

mechanism offers a 600% improvement in the throughput of a single TCP connection over the satellite link. Finally, as described in Reference [21] resource allocation algorithms must work co-operatively with network and transport layers to optimize system performance and to provide QoS to applications and users. The interest here is on future military packet-switched satellite networks that will dynamically allocate resources on uplink and downlink. This work studies the adoption QoS schedulers to provide service differentiation in the presence of link variations. Moreover, the interaction between TCP and the dynamic resource allocation algorithms is investigated, leading to suggest modifications of either the resource allocation algorithms or the TCP protocol or both. 3.5. Interactions among PHY, MAC and higher-layer protocols Different traffic flows (e.g. real-time traffic and non-real-time traffic) produced by the application layer need to have distinct QoS levels. A monitoring action should be jointly performed by application and MAC layers in order to control adaptively the service priority (top-down approach). Conversely, the adaptation of modulation and coding levels employed at the PHY level should be sent back to the application layer to change dynamically the source generation bit-rate (bottom-up approach). Such source coding adaptation is important to avoid buffer overflow and loss of information at the source during a fading period. An interesting example of PHY-MAC-application interaction is provided in Reference [22]. In particular, according to a cross-layer approach, the control parameters of a MAC protocol, i.e. Packet Reservation Multiple Access with Hindering States (PRMA-HS), are adaptively selected in the case of LEO satellites, depending on the packet error rate (related to the physical layer parameters and channel conditions) and the service type (related to the application, for instance, real-time voice, interactive Web traffic, etc.). PRMA-like protocols are based on contentions either on information slots or on separate signalling resources. The control parameter (i.e. permission probabilities) values can be adapted to modify the rate according to which terminals attempt transmissions. The problem is that a too aggressive protocol (i.e. with high permission probability) may lead to protocol bi-stability, with frequent collisions resulting in successful transmission probability going to zero. This is a critical problem especially when many terminals share the same (access) resources. It is therefore important to adopt a cross-layer scheme that dynamically determines transmission probability values on the basis of the characteristics of each traffic class, radio channel behaviour and traffic load conditions. Analytical studies as those carried out in Reference [22] can provide the appropriate framework for the cross-layer design of access protocol parameters in contention-based systems. Another case of a significant interaction between application and PHY layers is described in Reference [23]. Since real-time streaming traffic may use codecs that are error resilient, it is possible to receive data with (some) bit errors within the packet payload. This suggests that the packet (at the application layer) may be divided into sensitive and insensitive parts with differentiated error protection capabilities. Errors in the sensitive part cause a packet to be discarded whereas errors in the insensitive part, are delivered, leaving the decisions to application codec. A cross-layer technique is proposed that features partial checksum coverage for packet header allowing the application to signal implicitly the link CRC coverage. The sending end-host implicitly signals using a modified transport header, such as UDP-Lite [24]. Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

332

G. GIAMBENE AND S. KOTA

Finally, in an IP-based satellite network, IP datagrams suffer from residual errors provided by the reassembly of layer 2 packets. Note that IP datagrams are discarded in a router if the header checksum verification fails. Once discarded, layer 3 does not attempt any recovery procedure, since this is a task of higher layers. Hence, we can imagine that a receiver of a satellite link experiences an erasure channel at layer 3 or above. Consequently, end-to-end coding suitable for erasure channels can be employed at transport or application layer to recover partly these errors and to make an efficient use of the costly satellite resources. The adoption of such an approach is often the only realistic option for the design of a reliable transport layer, since the capacity and power constraints in wireless environments preclude the use or even the provision of a return channel for user feedback. In such a context we consider References [25,26] where reliable multicast and multimedia broadcasting are, respectively, addressed. Reference [25] proposes a cross-layer technique for geostationary-based systems where most of packet discarding at lower layers is removed and additional protocol header protection is introduced. Moreover, at transport level erasure coding is used in combination with a hybrid-Automatic Repeat reQuest (ARQ) protocol. Such an approach allows applications like massive and reliable file transfers to be less demanding in terms of network resources. It is important to note that when lower layer and higher layer coding are employed a joint optimization (implicit cross-layer approach) has to be performed to maximize system efficiency. Chipeta et al. [26] consider the same approach, for reliable point-to-multipoint distribution of contents in the digital multimedia broadcasting. Radio transmission errors due to the impairments of wireless links and intermittent connectivity during handoffs are the main reasons for data loss. A suitable transport layer encoding is used to achieve partial reliability. Low redundancy may result in high data download times, whereas increasing redundancy beyond the optimum value not only wastes the network capacity, but also increases the data download times. Reference [26] provides an analytical expression for the FEC redundancy levels that result in minimum average content download time.

4. IMPLICIT CROSS-LAYER DESIGN EXAMPLE This section presents an example showing the TCP performance results with interactions from lower layer protocols and, in particular, radio channel conditions, modulation and coding levels. Our aim is to show that it is convenient to select a threshold (PHY) for the selection between two transmission modes (featuring different modulation and coding levels) that allow maximizing the TCP goodput performance as a function of Eb/N0. This is an implicit cross-layer optimization method based on the simulation scenario described in Section 4.1; the optimization results are presented in Section 4.2. A further study is required to perform an explicit cross-layer optimization for this scenario, which requires the adoption of dynamic models for describing the behaviour of the physical channel and detailed assumptions on the link budget. Such an interesting analysis is beyond the scope of this paper and the study results will be reported in the future. In this example, we have considered different TCP versions particularly efficient in supporting segment losses due to channel errors, such as TCP NewReno, SACK, Westwood+, Hybla and scalable-TCP (S-TCP) [27–31]. TCP Reno suffers from multiple segment losses in a window of data. This is particularly important in satellite networks where the high Bandwidth Delay Product (BDP) values entail the use of large windows of data. The sender must wait for RTO to trigger the retransmission of Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

333

each lost segment [16]. An improvement to cope with multiple losses in a window of data is proposed by TCP NewReno [27], which modifies the fast recovery phase of TCP Reno. TCP NewReno can recover from multiple lost segments without waiting for the RTO expiration by using the ‘partial ACKs’ received during the fast recovery phase. As soon as a loss is identified, the fast retransmit algorithm is started by resending the lost segment. Then, the (new) fast recovery algorithm manages the ACK as follows: (i) if there has been a single lost segment, the ACK refers to all the segments transmitted up to the fast retransmit phase; (ii) otherwise, the ACK is partial and acknowledges some of the segments sent before the fast retransmit. In this study, we refer to the TCP NewReno Impatient version [27] that resets the retransmit timer only after the first partial ACK allowing RTO to occur soon when many losses occur in a window of data. Another improvement to deal with multiple losses in a window of data is the Selective ACKnowledgement (SACK) [28] in which the receiver informs the sender about the successfully received segments. The sender can retransmit only the lost segments. SACK can be implemented with both fast recovery and fast retransmit algorithms of the TCP NewReno. The SACK scheme is started by a ‘SACK-permitted’ message sent by the sender to the receiver in a SYN segment. Then, the receiver uses the SACK option in the TCP header when it sends an ACK that does not acknowledge the highest sequence number in the received sequence. Each contiguous block of data queued at the receiver is characterized in the SACK by two 4-byte integer numbers. Up to 4 blocks can be specified with SACK due to the TCP option maximum length of 40 bytes. Differently from TCP Reno and NewReno that reduce cwnd to one half after three duplicated ACKs, TCP Westwood (and the recent version Westwood+ [29]) sets cwnd and ssthresh after a congestion episode on the basis of an end-to-end bandwidth estimate, Bwe, made before the loss event was detected; in particular, ssthresh is made equal to Bwe
RTT. Hence, TCP Westwood avoids a conservative cwnd reduction, thus allowing a faster recovery phase. A similar modification is made when RTO expires: ssthresh = Bwe
RTT and cwnd is reset to its initial value. The bandwidth Bwe of a connection is continuously sampled by considering the amount of data sent and the ACK interarrival time. Then, these sample values are averaged by a low-pass filter, since only lower frequencies of the input traffic rate may lead to congestion. TCP Hybla has been recently conceived to address the problems of typical satellite connections, such as high propagation delays and frequent segment losses. In heterogeneous networks, TCP connections characterized by large RTT (i.e. including a satellite segment) present poor performance as compared to the wired connections with shorter RTT. In fact, since the congestion window growth depends on the reception of ACKs, a large RTT value affects the throughput and the channel utilization. To reduce the TCP goodput differences, TCP Hybla [30] proposes a modification to the cwnd update algorithm on the receipt of an ACK in order to accelerate the cwnd increase in both slow start and congestion avoidance phases for connections with large RTT values. Hybla needs to operate in conjunction with SACK in order to recover quickly from losses due to both channel errors and the more aggressive injection of data in the network. S-TCP is a proposal for modifying the TCP congestion control algorithm in order to improve the TCP performance in backbone high-speed networks [31]. S-TCP modifies the congestion control algorithm of TCP in order to take advantage of the large BDP value in such cases. S-TCP gets its name from the fact that the time it takes to recover from a loss occurred in the Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

334

G. GIAMBENE AND S. KOTA

presence of a congestion window value equal to W, i.e. the time it takes to return to W, does not depend on the W value. Whereas, in standard TCP, this recovery time increases linearly with W and this is one of the reasons why standard TCP behaves poorly in networks with large BDP values. S-TCP ‘scales’ well with the congestion window value and, hence, with BDP. S-TCP is based on a Multiplicative Increase Multiplicative Decrease (MIMD) algorithm, according to which the congestion window is increased by a factor a upon an ACK reception and decreased by a factor b upon a packet loss event. In our study, we have used a ¼ 0:01 and b ¼ 0:125:

4.1. Simulation model Figure 4 shows our simulation scenario consisting of a geostationary (GEO) bent-pipe satellite, a client (TCP receiver) connected to a router-Earth station which is in turn connected via satellite to another router-Earth station linked to a remote host (TCP sender) from which files are downloaded. An FTP application is assumed which produces TCP traffic according to an ACK-clocked model. A return channel via satellite is considered that is used both to send lower layer signalling for modulation and coding adaptation and layer 4 ACKs. The GEO satellite link has a raw bandwidth of 2 MHz for both uplink and downlink operating at Ka Band. Earth station to Earth station one-way propagation delay is about 260 ms. Terrestrial links from Earth station and the client and from the other Earth station to the server are at 30 Mbit/s with a one-way propagation delay of 10 ms. Hence, the satellite link is the system bottleneck and the propagation delay contribution to RTT is equal to 560 ms. We study the TCP goodput performance, considering advanced coding and modulations for radio channel conditions. The transmission mode (i.e. combination of modulation and coding) adaptation is performed by the sending Earth station on the basis of the channel quality measure made by the receiving Earth station. We have adopted two modulations: BPSK and QPSK. Moreover, we have employed a convolutional coder/Viterbi decoder over modulated symbols. In particular, we

Figure 4. Reference network architecture. Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

335

assume the standard NASA 1/2 rate convolutional code with constraint length 7 and derived punctured code with rate 3/4 [32]. Two distinct transmission modes are: * *

Mode #1: BPSK with rate 1/2 convolutional encoder, information bit-rate of 1 Mbit/s; Mode #2: QPSK with rate 3/4 convolutional encoder, information bit-rate of 3 Mbit/s.

We have assumed a segment length of 1500 bytes (Ethernet payload capacity). The BDP product is 560 000 bits (about 47 packets) for mode #1 and 1 68 0000 bits (140 packets) for mode #2. Each router has a buffer capacity in packets (i.e. segments) equal to BDP for mode #2 (i.e. the maximum BDP value between mode #1 and mode #2). We also assume: (i) Earth stations with a Line-Of-Sight (LOS) path to the GEO satellite; (ii) a memoryless channel with uncorrelated losses; (iii) attenuation fluctuation only due to slow tropospheric events (long-term variations of the received signal strength due to cloud attenuation and rain fades, no shadowing) so that the channel can be considered of the AWGN type; (iv) residual segment losses after the decoding process are uncorrelated and occurring according to a given Segment Error Rate (SER) at the transport level; (v) channel variations are very slow compared to the delay of the feedback signal to inform the sending Earth station to modify its transmission mode; and (vi) no ARQ technique is used since it would cause a high delay to recover packet losses that would probably cause an RTO expiration. According to Reference [33], it is possible to obtain the SER behaviour as a function of Eb/N0 (at the level of coded bits for the BPSK case, i.e. the reference Eb/N0 value for short) for both modes #1 and #2. We are interested in defining an optimal criterion for the selection between mode #1 and mode #2 in order to maximize the TCP goodput performance for the different Eb/N0 values. Note that we expect that the switching point between different modes (i.e. Eb/N0 threshold value) also depends on the TCP version (i.e. TCP NewReno, SACK, Westwood+, Hybla, STCP), since different TCP versions have a different ‘resilience’ to packet losses, as detailed in the following Section. For instance, note that TCP NewReno is able to exploit fully the available bandwidth for a SER 4 SER* = 1.5/(BDP2) that corresponds to 6.8
104 for mode #1 (i.e. Eb/N0 5 3 dB) and to 7.65
105 for mode #2 (i.e. Eb/N0 5 8 dB) [34]. 4.2. Simulation results We have implemented the network simulation model described in Figure 4 under the ns-2 environment. Figure 5 compares the TCP goodput performance of different TCP versions as a function of SER for transmission mode #2, a similar behaviour could be obtained in the case of mode #1. We note that for low SER values, Westwood+ achieves the highest goodput performance, whereas as SER increases TCP Hybla turns out to have a better performance. The results in Figure 5 clearly highlight that different TCP versions have a different ‘resilience’ to the presence of channel errors. Figure 6 shows the TCP goodput for TCP NewReno, SACK, Westwood+, Hybla, and STCP as a function of the reference Eb/N0 for both the transmission modes. Figure 6 shows that for high Eb/N0 values, mode #1 does not allow to exploit adequately system capacity so that mode #2 is more convenient. Whereas, for low Eb/N0 values mode #2 entails too frequent errors that drastically affect the TCP goodput; therefore, the more protected mode #1 has to be used. The Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

336

G. GIAMBENE AND S. KOTA

TCP goodput [bit/ s]

107

106

10

10

Hybla Westwood+ SACK NewReno STCP

5

4 −5

10

10

−4

−3

10 10 Segment Error Rate, SER

−2

10

−1

Figure 5. Simulation results for the different TCP versions as a function of SER for mode #2. 3 Hybla NewReno SACK Westwood+ STCP

TCP goodput [Mbit / s]

2.5 2 1.5

Transmission mode # 2

Transmission mode # 1

1 0.5 0

0

2

4

6 Eb/N0 [dB]

8

10

12

Figure 6. Simulation results for the different TCP versions as a function of the Eb/N0 (reference value for BPSK) for fixed modes 1 and 2; the maximum 95% confidence interval amplitude for all the curves is  7%.

results in Figure 6 suggest that there should be switching points between mode #1 and mode #2 for the five different TCP variants in the range from 5.5 to 7.5 dB. Figure 7 shows a graph derived from that in Figure 6 where for each TCP version we have considered an adaptive mode: the mode selection criterion is to maximize the TCP goodput. The resulting switching points in terms of Eb/N0 are: *

about 5.8 dB for Hybla, corresponding to SER(mode#2) BER(mode#2)  4.3
105;

Copyright # 2006 John Wiley & Sons, Ltd.



8
102 and to

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

337

3 Hybla Westwood+ SACK NewReno STCP

TCP goodput [Mbit/s]

2.5 2

Opt. mode transition threshold for Westwood+

1.5

Opt. mode transition threshold for STCP

1

0.5 0

Opt. mode transition threshold for NewReno and SACK

Opt. mode transition threshold for Hybla

0

2 4 Use of mode #1

6 Eb /N0[dB]

8

10 12 Use of mode #2

Figure 7. Simulation results for the different TCP versions as a function of the Eb/N0 (reference value for BPSK) in the case with adaptive modes 1 and 2; the maximum 95% confidence interval amplitude for all the curves is  7%. *

*

*

about 6.2 dB for Westwood+, corresponding to SER(mode#2)  4.2
102 and to BER(mode#2)  2
105; about 6.6 dB for STCP, corresponding to SER(mode#2)  1.1
102 and to BER(mode#2)  4.7
106; and 7.5 dB for both TCP NewReno and SACK, corresponding to SER(mode#2)  5.9
104 and to BER(mode#2)  1.7
107.

We note that Hybla, Westwood+ and STCP achieve the best performance in recovering from packet loss events; whereas, lower goodput values are obtained with SACK and TCP NewReno. Figures 6 and 7 allow us to determine the most convenient transmission mode for different Eb/N0 values and diverse TCP versions. The actual mean goodput for a given TCP version depends on the distribution of the Eb/N0 value due to the meteorological phenomena, the link budget, the antenna pattern, and the position of the Earth stations. In summary, our simulation results demonstrate that the TCP-driven optimization of lower layers can permit to improve the higher-layer goodput in satellite networks.

5. CONCLUSIONS The design of IP-based satellite communication systems integrated with the 4G scenario calls for system optimization considering the interactions among the different protocol layers. The conventional protocol stack is based on independent layers, thus precluding the adaptation of each layer according to changing system conditions. It is therefore extremely critical to study the interaction and the impact on different layers according to a new cross-layer design approach. This is a new paradigm that involves physical, network, transport and application layers for the optimized management of resources. Tasks such as physical layer adaptivity, joint layer 2 and Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

338

G. GIAMBENE AND S. KOTA

higher layer coding optimizations for end-to-end reliable transmissions, QoS mapping from layer 3 and 2, dynamic resource allocation depending on higher layer behaviours are important functions that according to the novel cross-layer approach need to be introduced for advanced satellite communication systems. These challenges require further detailed research. For example, combination of dynamic resource management and cross-layer approach can permit to meet QoS levels for multimedia traffic, guaranteeing at the same time good resource utilization. The cross-layer protocol engineering is an interdisciplinary field which warrants research of new methods and techniques.

NOMENCLATURE ACM AF ACK AQM ARQ AWGN BDP BER COPS CRC DAMA DS DVB-RCS DVB-S DVB-S2 EF ETSI ETSI TC-SES/BSM

FTP GEO IP LDPC LOS MAC MIMD NCC OSI PEPs PHY PSK PRMA-HS QPSK

Adaptive Coding & Modulation Assured Forwarding Acknowledgment Active Queue Management Automatic Repeat Request Additive White Gaussian Noise Bandwidth Delay Product Bit Error Rate Common Open Policy Service Protocol Cyclic Redundancy Check Demand-Assignment Multiple Access Differentiated Service Digital Video Broadcasting-Return Channel via Satellite Digital Video Broadcasting-Satellite Digital Video Broadcasting-Satellite 2 Expedited Forwarding European Telecommunication Standards Institute European Telecommunication Standards Institute Technical Committee } Satellites Standard Earth Stations and Systems/Broadband Satellite Multimedia File Transfer Protocol Geostationary Internet Protocol Low-Density Parity Check Line-Of-Sight Medium Access Control Multiplicative Increase Multiplicative Decrease Network Control Center Open System Interconnection Performance Enhancement Proxies Physical Layer Phase Shift Keying Packet Reservation Multiple Access with Hindering States Quadrature Phase Shift Keying

Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

QoS RED RSVP RTT RTO SACK SER SI-SAP SLA S-TCP TCP ToS UDP

339

Quality of Service Random Early Detection Resource Reservation protocol Round Trip Time Retransmission Time Out Selective Acknowledgment Segment Error Rate Satellite-Independent-Service Access Point Service Level Agreement Scalable-TCP Transmission Control Protocol Type of Service User Datagram Protocol

ACKNOWLEDGEMENTS

The authors wish to thank the SatNEx project and Dr Nedo Celandroni (CNR-ISTI, Italy) for providing the material concerning the performance of NASA convolutional codes. The authors would like to thank also Nico Candio Liberato for his contribution to simulator set-up. REFERENCES 1. ETSI. Digital video broadcasting (DVB); interaction channel for satellite distribution systems. Final draft EN 301 790, V1.3.1 (2002-11). ETSI, 2002. 2. Kota SL, Hossain E, Fantacci R, Karmouch A. Cross-layer protocol engineering for wireless mobile networks: part 1. IEEE Communications Magazine, December 2005; 43(12):110–111. 3. ETSI. Satellite earth stations and systems (SES); broadband satellite multimedia (BSM) services and architectures; functional architecture for IP interworking with BSM networks. TS 102 292, V1.1.1 (2004-02). ETSI, 2004. 4. Wang Q, Abu-Rgheff M-A. Cross-layer signalling for next-generation wireless systems. IEEE Wireless Communications and Networking Conference (WCNC), New Orleans, U.S.A., 16–20 March 2003. 5. Conti M, Crowcroft J, Maselli G, Turi G. A modular cross-layer architecture for ad hoc networks. In Handbook on Theoretical And Algorithmic Aspects of Sensor, Ad Hoc Wireless, and Peer-to-Peer Networks, Chapter 1, Wu J (ed.). CRC Press: New York, 2005. 6. Carneiro G, Ruela J, Ricardo M. Cross-layer design in 4G wireless terminals. IEEE Wireless Communications Magazine April 2004; 11(2):7–13. 7. Vardhan V, Sachs DG, Yuan W, Harris AF, Adve SV, Jones DL, Kravets RH, Nahrstedt K. GRACE: a hierarchical adaptation framework for saving energy. Technical Report UIUCDCS-R-2004-2409, Computer Science, University of Illinois, February 2004. 8. Durham D, Boyle J, Cohen R, Herzog S, Rajan R, Sastry A. The COPS (common open policy service) protocol. IETF RFC 2748, January 2000. 9. ETSI. Digital video broadcasting (DVB); second generation framing structure, channel coding and modulation systems for broadcasting, interactive services, news gathering and other broadband satellite applications (DVB-S2). Draft EN 302 307. ETSI, 2005. 10. Celandroni N, Davoli F, Ferro E, Gotta A. TCP connections via satellite: cross-layer bandwidth allocation, pricing and adaptive control. IS-MANET Project. Available at the URL: http://zeus.elet.polimi.it/is-manet/Documenti/ pap-isti-8.pdf 11. Celandroni N, Davoli F, Ferro E. Static and dynamic resource allocation in a multiservice satellite network with fading. International Journal Satellite Communication and Networking 2003; 21(4–5):469–487. 12. Celandroni N, Davoli F, Ferro E, Gotta A. Networking with multi-service GEO satellites: cross-layer approaches for bandwidth allocation. International Journal of Satellite Communications and Networking 2006; 24(5):385–400. 13. Vieira F, Va´zquez Castro MA, Seco Granados G. A tunable-fairness cross-layer scheduler for DVB-S2. International Journal of Satellite Communications and Networking 2006; 24(5):433–446. 14. Kota SL, Pahlavan K, Leppanen P. Broadband Satellite Communications for Internet Access. Kluwer Publications: Dordrecht, 2004. Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

340

G. GIAMBENE AND S. KOTA

15. Allman M, Glover D, Sanchez L. Enhancing TCP over satellite channels using standard mechanism. IETF RFC 2488, January 1999. 16. Allman M, Paxons V, Stevens W. TCP congestion control. IETF RFC 2581, April 1999. 17. Chini P, Giambene G, Bartolini D, Luglio M, Roseti C. Dynamic resource allocation based on a TCP-MAC crosslayer approach for interactive satellite networks. International Journal of Satellite Communications and Networking 2006; 24(5):365–383. 18. Border J, Kojo M, Griner J, Montenegro G, Shelby Z. Performance enhancing proxies intended to mitigate linkrelated degradations. IETF RFC 3135, June 2001. 19. Floyd S, Jacobson V. Random early detection gateways for congestion avoidance. IEEE/ACM Transactions on Networking, August 1993; 1(4):397–413. 20. Peng F, Wu L, Leung VCM. Cross-layer enhancement of TCP split-connections over satellites links. International Journal of Satellite Communications and Networking 2006; 24(5):401–414. 21. Narula-Tam A, Wysocarski J, Yao H, Wang M-C, Macdonald T, Huang O, Pandya J. Cross layer design of satellite communication systems with dynamic resource allocation. International Journal of Satellite Communications and Networking 2006; 24(5):341–363. 22. Giambene G, Zoli E. Stability analysis of an adaptive packet access scheme for mobile communication systems with high propagation delays. International Journal of Satellite Communications and Networking 2003; 21:199–225. 23. Stanislaus W, Fairhurst G, Radzik J. Cross layer techniques for flexible transport protocol using UDP-Lite over a satellite network. Proceedings of IWSSC 2005, Siena, Italy, 8–9 September 2005. 24. Karn P, Larzon L-A, Degermark M, Pink S, Jonsson L-E, Fairhurst G. The lightweight user datagram protocol (UDP-Lite). IETF RFC 3828, July 2004. 25. Arnal F, Dairaine L, Lacan J, Maral G. Cross-layer reliability management for multicast over satellite. Journal on Computer Networks and ISDN Systems 2005; 48(1):29–43. 26. Chipeta M, Karaliopoulos M, Fan L, Evans BG. Integration of packet-level FEC with data carousels for reliable content delivery in mobile satellite broadcast/multicast systems. International Journal of Satellite Communications and Networking, in press. 27. Floyd S, Henderson T, Gurtov A. The NewReno modification to TCP’s fast recovery algorithm. IETF RFC 3782, April 2004. 28. Mathis M, Mahdavi J, Floyd S, Romanow A. TCP selective acknowledgment options. IETF RFC 2018, 1996. 29. Grieco LA, Mascolo S. Performance evaluation and comparison of Westwood+, New Reno, and Vegas TCP congestion control. ACM Computer Communication Review, April 2004; 34(2):25–38. 30. Caini C, Firrincieli R. TCP Hybla: a TCP Enhancement for heterogeneous networks. International Journal of Satellite Communications and Networking 2004; 22(5):547–566. 31. Kelly T. Scalable TCP: improving performance in highspeed wide area networks. ACM Computer Communication Review, April 2003; 33(2):83–91. 32. Qualcomm. Q1401: K=7 rate 1/2 single-chip Viterbi Decoder technical data sheet. Qualcomm incorporated, September 1987. 33. Celandroni N, Potortı´ F. Maximising single connection TCP goodput by trading bandwidth for BER. International Journal of Satellite Communications and Networking 2003; 16:63–79. 34. Floyd S, Fall K. Promoting the use of end-to-end congestion control in the Internet. IEEE/ACM Transactions on Networking, August 1999; 33(2):83–91.

AUTHORS’ BIOGRAPHIES

Giovanni Giambene was born in Florence, Italy, in 1966. He received the Dr Ing degree in Electronics from the University of Florence, Italy, in 1993 and the PhD degree in Telecommunications and Informatics from the University of Florence, Italy, in 1997. From 1994 to 1997, he was with the Electronic Engineering Department of the University of Florence, Italy. He was Technical External Secretary of the European Community COST 227 Action, entitled ‘Integrated Space/Terrestrial Mobile Networks’. He also contributed to the Resource Management activity of the Working Group 3000 within the RACE Project, called ‘Satellite Integration in the Future Mobile Network’ (SAINT, RACE 2117). From 1997 to 1998, he was with OTE of the Marconi Group, Florence, Italy, where he was involved in a GSM development programme. In the same period he also contributed to the COST 252 Action (‘Evolution of Satellite Personal Communications from Second to Future Generation Systems‘) research activities by studying the performance of Packet Reservation Multiple Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat

CROSS-LAYER PROTOCOL OPTIMIZATION

341

Access (PRMA) protocols suitable for supporting voice and data transmissions in low earth orbit mobile satellite systems. In 1999 he joined the Information Engineering Department of the University of Siena, Italy, first as research associate and then as assistant professor. He teaches the advanced course of Telecommunication Networks at the University of Siena. From 2000 to 2003, he contributed to the activities of the ‘Personalised Access to Local Information and services for tOurists’ (PALIO) IST Project within the fifth Research Framework of the European Commission (www.palio.dii.unisi.it). At present, he is involved in the SatNEx network of excellence (www.satnex.org) of the FP6 programme in the satellite field, as work package leader on radio resource management techniques (ja2330) and cross-layer air interface design (ja2230). He is also vice-Chair of the COST 290 Action (www.cost290.org), entitled ‘Traffic and QoS Management in Wireless Multimedia Networks’ (Wi-QoST). He has recently published a book entitled ‘Queuing Theory and Telecommunications: Networks and Applications’, Springer (May 2005). His research interests include third-generation mobile communication systems, medium access control protocols, traffic scheduling algorithms, and queuing theory.

Dr Kota received his BS Physics from Andhra University, BSEE from BITS, Pilani, MSEE from Indian Institute of Technology (IIT), India. He received the Electrical Engineer’s Degree from Northeastern University, Boston, U.S.A and PhD in Electrical and Information Engineering from University of Oulu, Finland. Since 2003 he has been a Senior Scientist in Harris Corporation involved with Corporate Technologies and Standards with special emphasis on Wireless and Mobile Ad Hoc Networks, satellite communication networks and Standardization. He is an Adjunct Professor in the Telecommunications Laboratory of University of Oulu. His research interests include wireless and mobile Information networks, satellite IP networks, QoS and traffic management, broadband satellite access, and ATM networks. Over the years, he held technical and management positions and contributed to military and commercial communication systems at Loral Skynet, Lockheed Martin, SRI International, The MITRE Corp and Xerox Corp. He has been very active in telecommunications and networking standards development. Currently, he is the US chair for ITU-R, Working Party 4B and International Rapporteur for Ka-Band Fixed Satellite Systems. He was the chair for Wireless ATM Working Group and has been an ATM Forum Ambassador. He was the recipient of the ATM Forum Spotlight award and Golden Quill award from Harris Corporation for his contributions to Broadband Satellite Communications and Assured Communications. Dr Kota is the principal author of the book Broadband Satellite Communications for Internet Access published by Kluwer Academic Publishers, and is the co-editor of book ‘Emerging Location Aware Broadband Wireless Ad Hoc Networks’ by Springer, and has contributed book chapters to Encyclopedia of Telecommunications, John Wiley & Sons, High Performance TCP/IP Networking, Prentice Hall, and Modeling and Simulation Environment for Terrestrial and Satellite Networks, Kluwer Academic Publishers,. He has published and presented over 120 technical papers in book chapters, journals, and conference proceedings. He served as a guest editor for IEEE Communications Magazine, Special Issues on Cross-Layer Protocol Engineering for Wireless Mobile Networks, Satellite ATM architectures, Broadband Satellite Network Performance, and International Journal of Satellite Communications and Networking, Special Issue on Satellite IP QoS. He currently serves on the editorial boards of International Journal of Satellite Communications and Networking (Wiley Interscience), and International Journal of Space Communications (IOS Press). He is an Industry Advisory Board member of Rochester Institute of Technology and CRUISE Project. Dr Kota has been a keynote speaker, invited speaker and panelist at various International Conferences. He also served as Tutorial chair and Asst. Technical chair of MILCOM2004, 1997, 1990; symposium chair, co-chair of satellite Communications symposium of GLOBECOM 2000, 2002 and invited session chair of PIMRC 2004, 2005 and 2006. He is the co-chair of Wireless Communications and networking symposium of GLOBECOM2006 and Technical chair of ISWPC2007. He has been a member of technical program committees of several IEEE, AIAA, SPIE and ACM conferences and workshops. He is a senior member of IEEE, Associate Fellow of AIAA, and member of ACM.

Copyright # 2006 John Wiley & Sons, Ltd.

Int. J. Satell. Commun. Network. 2006; 24:323–341 DOI: 10.1002/sat