When considering the main requirements of an advanced QoS-oriented MAC, it is important to begin by describing the basic principles and goals of the novel advanced MAC-based architecture MPLS/ATM-MFMAC. In this section the advanced MAC-based architecture and the requirements to the advanced QoS-oriented MAC will be considered .
The known MPLS over ATM architecture is illustrated in Figure 1a. MPLS networks use a superset of the protocols used in ordinary IP routers. The ordinary IP may be thought of as a forwarding protocol, since it moves packets, rather than as a type of forwarding mechanism called label switching . Label switching relies on the setup of switched paths (LSPs) through the network. A process of label distribution does the setup of LSPs.
Switch-based MPLS/ATM's label distribution protocol (LDP) supports hop-by-hop label routing (termed ATM "Pipeline" to ATM "Pipeline" by Markhasin, 1996). LDP is used in combination with the resource reservation protocol plus some extensions. Switch-based LSRs used to date have been based on ATM switches. The LSRs have ATM virtual path (VPI) and circuit (VCI) identifier as label encapsulation. The ATM cell forwarding mechanism also supports hop-by-hop ATM label switching .
The advanced-MAC-based MPLS/ATM-MFMAC alternative is illustrated in Figure 1b. This alternative is based on the integration of the L2 ATM MFMAC technology, of the L2 completely distributed multifunctional ATM hyperbus architecture, and of the L3 MPLS over ATM routing. The suggested MPLS/ ATM structure includes least of L3 LSR/MFMAC routing nodes (RN) for ATM hyperbuses interworking or for edge internetworking and also more of L2 ATM/MFMAC transit nodes (TN) for data units (ATM cell) or label (VPI-VCI) selecting .
Markhasin (1996, 2001) and Brandt et al. (2000) noted that the advanced broadband ATM MFMAC ensures a completely distributed and adaptive access to the long-delay ATM medium and also provides dynamic bandwidth control and network configuration, traffic parameters, and QoS for each user , services class, date, and label unit, and finally ‚for each time slot. The ATM hyperbus builds a virtual bus between all the users or L2 nodes. Brandt et al. (2000) argued that the topology of this virtual bus can be represented by a PON tree, a classical PON, wireless or satellite spacemedium hyperbus bus, cross-hexagonal (see Figure 1b), ring-radial ("metro" topology). This virtual bus can support wide and global areas with long-delay distances.
The advanced broadband ATM MFMAC technology is based on adaptive long-delay MAC protocol of RS-token broadcast reservation (RS-TBR; see Markhasin, 1996). This protocol is based on the concept of recurrent M-sequences (RS) and offers an effective medium access control and a low-cost selective addressing of the unique protocol subjects (up to the so-called RS-address bit per each time slot/reservation mini-slot/L2 station identifier/ATM cell header/request/label, etc.).
The dimension of RS-address bit space may support up to M = 2 n 1 unique addresses, within n standing for the degree of the generating polynomial (n = 48 64). That is, the label dimension M ranges between M ~ 2,8E+14 and M ~ 1,8E+19 per hyperbus. By using the Shannon-Fano method, Markhasin (1988) noted that the RS-addresses of each protocol subject can be dynamically encoded using optimum non-uniform codes. In other words, the RS-addressing mechanism can support simultaneously label routing ("ATM hyperbus to ATM hyperbus") and the dynamic resource reservation protocol. In fact, one can make certain that the method of RS-addressing is very promising for MPLS label encapsulation, which is of a special interest in generalized (see Benerjee et al., 2001) MPLS.
The advanced-MAC-based MPLS/ATM does not require hop-by-hop switching mechanisms. The advanced-MAC-based MPLS/ATM label distribution protocol uses a label selecting mechanism by TN and label routing mechanism by RN for setup of selected paths through the network. The ATM cell forwarding protocol uses also a non-hop-by-hop label selecting mechanism by transit and routing mechanism "ATM hyperbus to ATM hyperbus".
The diagram of an all-MPLS/ATM-MFMAC-based architecture applied to satellite networks is illustrated in Figure 2. The all-MPLS/ATM-MFMAC satellite network architecture is based on an implementation of distributed virtual spacemedium ATM hyperbus internetworking and interbusworking and uses the common universal dynamically adapted MFMAC protocol through the entire network's hierarchy ‚ core , backbone, and access networks.
At this point, it is perhaps appropriate to explain the abbreviations used in Figure 2:
L1_UAMI ‚universal adaptive medium interface with medium type cartridge,
L2_MFMAC ‚universal adaptive multifunctional distributed ATM medium access controller,
L3_LSR ‚label switch router,
RN ‚routing node,
TN ‚transit (selecting) node,
AN ‚hyperbus QoS and AAL adaptation manager node,
L2_QoS ‚the hyperbus manager of the distributed dynamic QoS control, traffic parameters (TP), and bandwidth resource assignment (BRA),
Th_QoS ‚the interhyperbus path manager of the distributed dynamic QoS control, TP, and BRA.
Referring to Figure 2, the spacemedium ATM hyperbus builds a virtual bus between all the users, which are in the footprint of a given satellite. The satellite in this case is merely a re-translator, implying that users in different geographical locations within the footprint can share the same virtual bus. The ATM hyperbus provides a wide range of services with different traffic streams and different QoS characteristics. QoS guarantees are essential for the support of real-time services such as telephone and video.
It is possible to configure various wireless and satellite IP/ATM and ATM networks which differ according to their topology scale (W-LAN, W-MAN, W-WAN, S-GAN), function hierarchy (core, backbone, and access networks), technical characteristics, and kinds of medium used.
The main advantages of the MFMAC-based MPLS/ATM integration technology over ATM-switching-based technology are:
through dynamic path control and guaranteed QoS provisioning,
through path bandwidth on dynamic demand (up to real time),
multifunctional, universal, and adaptation capabilities,
all-MPLS/ATM completely distributed architecture, based on common multifunctional universal logical homogenous MAC sub- layers through all networks' function hierarchy ‚access, core, and backbone,
minimum value of expensive data units switching ‚maximum value of inexpensive data units selecting,
minimum of cells delay ‚maximum of reliability,
highly reconfigurable and scalable ‚from MAN up to GAN,
low cost, simplicity, support of multimedia service mass-market.
As we already briefly mentioned, it is necessary to overcome three principal barriers to the implementation of advanced ATM MAC technology, dynamic QoS control techniques, and all-IP/ATM architecture for future global all-IP/ ATM wireless environments. The time barrier manifests itself in a degradation of the long-delay MAC capacity when the round-trip time is increased. The dynamic barrier is the principal obstacle in implementing dynamic on-the-fly QoS provisioning and control. Finally, Markhasin (1996) argued that the economic barrier is due to unacceptable costs incurred by any centralized architecture, making a low-cost wireless broadband ATM mass-market implementation difficult.
The time barrier appears due to the effect of long-delay ATM MAC efficiency characteristics degradation when the round-trip time t p increases considerably. In Markhasin (1984, 1996) it was shown that the MAC protocol efficiency depends on normalized medium's bandwidth resources expenses on multi-access control v o = Y MAC / Y INF , on normalized round-trip (signal propagation) time v p = t p / T INF , and also from method of the multi-access control instructions processing:
where Y INF is average value of the medium's bandwidth resources expenses used for information traffic (packet or cells) transmission, Y MAC is average value of the medium bandwidth resources cost needed to organize the multi-access control (reserving, request, carrier sense, etc.), T INF is the average duration of information time slot (traffic packet or cell burst), a is a coefficient that depends on the type and system parameters of MAC protocol, a ° 1; b is a coefficient depending on the method of MAC instructions processing.
It is shown in Markhasin (1984) that if the processing is serial, b=1, and that b = 0 if the processing is parallel. The MAC protocols' efficiency was defined in Markhasin (1984) as:
where G is input traffic intensity, S is throughput, or output traffic intensity.
The value of the media resources expenses can be represented generally as
where T x , F x , P x are expenses of time, frequency bands, and power of media's resources. Assignment and variation of the frequency bands and power expenses are essential accordingly for frequency division (FDMA) and code division (CDMA) multiple access methods . Further we consider time division multiple access (TDMA) principles and assume that T x = var( x ), ‚ F x = const ( x ), ‚ P x = const ( x ). Therefore, we can represent the medium bandwidth resources normalized cost of multi-access control as:
where T MAC is the average duration of a MAC mini-slot (reservation, carrier sense, etc.).
Further we will range the normalized medium bandwidth resources expenses on multi-access control as small MAC expense if v o ° 0.02, moderate if 0.02 < v o ° 0.10, and large if v o > 0.10.
Thus, equation (1) confirms that the efficiency of MAC protocols in the case of serial processing (i.e., b = 1) goes to 0 with the increase of the round-trip time:
while, for the case of parallel processing (i.e., b = 0), we have:
It follows from (6) that the round-trip time of the MAC with parallel processing shows itself as transport delay.
Markhasin (1996) showed that the long-delay characteristics of the wireless broadband ATM MAC are conveniently expressed in "space-time-bit rate" coordinates, using the medium bit delay (MBD) characteristic:
where B is the bit rate (bit/s), L is the distance (km), and c is the signal propagation rate (km/s). Further we say that the MBD is small if d p ° 10, moderate if 10 < d p ° 100, large if 100 < d p ° 1,000, and hyper large if d p > 1000. The time barrier problem is illustrated in Figure 3.
We now address the dynamic barrier problem . To overcome the MAC degradation in packet satellite and radio networks, the superframe formats reservation (SFR) protocol of Rubun (1979), Markhasin (1988) described using the parallel processing of the multi-access instructions. Its frame format can be defined as fixed (FPODA, FRAC, and the like) and also as traffic-adaptable (DFRAC, ARDA, and the like). The efficiency (6) of these methods is:
where J is the (fixed or variable) number of the information slots, N is the number of the reservation mini-slots in the superframe, and coefficient a = N / J . Equation (8) tells us that for minimization of the SFR's cell delay, when the input traffic intensity G changes, it is necessary to adapt the value of N / J .
Markhasin (1996) showed that as the intensity G increases, the dependence of the delay time on the SFR protocol efficiency becomes more and more critical with respect to the value of the control parameter a = N / J . And the graph of a delay time function has a U-like shape (see also Figure 4).
The operational field where the delay time is minimal decreases rapidly as the traffic intensity G increases. The avalanche-like increase in the delay time on the left branches of the U-like plots is due to the saturation of the information cells queue ( G C parallel MAC ) and on the right branches due to the saturation of the reserving mini-cells queue ( G reserving 1 C parallel MAC ). In the operational field the following relations hold G < C parallel MAC and G reserving < 1 C parallel MAC.
As G increases, the characteristics must be kept within the optimum field by jumping to new optimum superframe format. Simultaneously catastrophic decreases the stability of superframe formats adaptation. Therefore SFR protocol supports real throughput (or realistic dynamic bandwidth utilization) only up to 0.30, 0.50. This is the principal shortcoming of the dynamic capabilities of the SFR long-delay MAC protocols with parallel processing.
The best dynamic capabilities can be represented as the asymptotic ideal dependence TBR (see Figure 4b) of advanced long-delay ATM MAC protocol, which provides the stable automatic on-the-fly adaptation of the M-periodical hyperframe to traffic intensity, based on parallel-conveyer processing of MAC instructions. This will be looked at further later in the chapter.
The economic barrier problem is very important from the perspective of developing low-cost broadband mobile and satellite mass-market services. This barrier manifests itself in unacceptably high costs characteristic of any centralized architecture.
Akyildiz and Jeong (1997) and Ivancic et al. (1999) described that the wellknown broadband ATM long-delay satellite multimedia systems are based mainly on centralized architecture (ATM "Pipeline", classical ATM satellite switch, centralized ATM/IP, ATM switch/router/multiplexer, ATM backbone, subscriber's ATM concentrator, ATM satellite "Rays", etc.). However, the cost of such a centralized ATM architecture is unacceptably large for deployment with the huge user populations in the mass market of broadband multimedia services. Such centralized architecture is used mainly in transport and corporate networks. According to Gillespie, Orth, Profumo, and Webster (1997), the main reason of the high cost of such networks is large investments necessary for the broadband access networks.
An effective solution is to develop a completely distributed all-IP/ATM architecture, based on QoS-oriented characteristics of advanced MAC ATM technology and on carrying over the principles of the simple and inexpensive local areas LAN and WATM architecture up to global areas W-MAN, W-WAN, and S-GAN wireless and satellite ATM networks. It is widely known that the development of distributed wireless (WLAN, WATM) and mobile broadband (MBS) systems of the 3G initiative within the framework of the European Program IST was successfully realized. However, da Silva, Barani, Arroyo-Fernandez, Pereira, and Ikonomou (1998) noted that these elaborated systems have a limited ‚either ‚ distance L up to 1,000 meters and that due to the time barrier only small to moderate medium bit delay (MBD), restricting these systems local urban areas (see Figure 3b). However, the task of developing a fully all-IP/ATM architecture is highly nontrivial. To be cost-effective , we need to increase the operation distance of distributed multifunctional broadband ATM systems up to 10,000 times and to decrease its cost by up to 10 times.
Based on the fundamental principles of the advanced all-IP/ATM architecture of the future global wireless and satellite broadband networks and its three barriers considered in Subsections 3.1 and 3.2, during the design of the MAC protocol the following milestones must be achieved:
high efficiency and throughput of access control to long-delay wireless and space mediums (for providing of the global, wide, and metropolitan area coverage and overcoming the time barrier , i.e., all-IP/ATM aspects);
high controllability, differentiation, and guarantee of QoS, traffic parameters, and bandwidth resources allocation (for QoS provisioning );
high efficiency and stability of the dynamic QoS control , traffic parameters, and bandwidth resources (for providing real time, optimization, and adaptation, i.e., overcoming the dynamic barrier );
low-cost, homogeneous, completely distributed hyperbus architecture of the MAC sub-layers through all networks' hierarchy ‚core, backbone, and access networks (for providing the IP/ATM integration through the entire network hierarchy on the basis of the common, universal, dynamically adapted MAC protocol, i.e., overcoming the economic barrier and ensuring all-IP/ATM aspects);
high multifuctionality and universality on basis of the common dynamically controlled and adaptive ATM MAC protocol through the entire network hierarchy ‚core, backbone, and access networks (for high effective realization of the all-MPLS/ATM aspects, i.e., using multi-protocol on top sub-layers and common multifunctional and universal protocol on MAC sub-layers ).