PT105305B - A METHOD FOR ADVANCED CONTROL OF RESERVOIRS BASED ON SERVICE CLASSES AND SYSTEM FOR ITS IMPLEMENTATION - Google Patents

Abstract

This invention provides a method and apparatus for dynamically configuring bandwidth parameters on service-grade (COS) -based networks by means of a network-over-network RESERVATION STRATEGY without QOS violations, while minimizing the efficiency of the QoS SIGNALING, BAND WASTE AND PROBABILITY OF LOCKOUT SESSIONS, WHEN COMPARED WITH PREVIOUS SOLUTIONS BASED ON OVER-RESERVE. IT IS CHARACTERIZED BY: - ATTRIBUTING THE TOTAL BANDWIDTH OF CONNECTION DESIGNED TO COSS OF SERVICES IMPLEMENTED IN CONNECTIONS IN THE INITIALIZATION OF THE NETWORK; - ALLOW THE NETWORK DECISION CONTROL POINT TO AVOID THE VIOLATION OF QOS ,; - WHEN RECEIVING A REQUEST FOR A COS IN A CERTAIN PATH, IF THE AVAILABLE BAND WIDTH IN THE COURSE IN THE GARGALO DO CAMINHO INTERFACE, INDICATED BY THE LOCAL DATABASE, IS INSUFFICIENT, DEFINE NEW RESERVE PARAMETERS FOR EACH COST IN EACH INTERFACE ON THE PATH.

Description

A METHOD FOR ADVANCED OVER RESERVE CONTROL BASED ON SERVICE CLASSES AND SYSTEM FOR THEIR EXECUTION

Background

Future Networks is expected to support new features and capabilities to provide value-added sessions (eg, multimedia, custom services, immersive video, etc.) about heterogeneous technologies (transport, access, etc.) with Quality of Service support ( - QoS). In this scenario, Class of Service-based (CoS) QoS models are less complex than flow-through methods (such as

reservation of appeal - [1]). In networks based on CoSs (as services Differentiated DitfServ [2], Commutation in hang tags multiprotocol (MPLS) - [3]), the streams are classified into a limited set of CoSs very good defined (ITU-T Y.1541 [23] and TS 123 107 [24]) in the borders network or central stations, through in politics predefined. At the However, technologies based on in bundles originally cannot allow the shipping in

quality data above the best attempt, forcing session streams to experience unwanted transport delay, delay variation (a.k.a. Jitter), and packet loss. Therefore, there is a need to allocate bandwidth and introduce signaling to install, maintain, or remove resource state for CoSs when needed. For QoS guaranteed services (sessions mapped to a

CoSs compatible with the DiffServ dispatching service (VoIP), domain behavior can be ensured through bandwidth reservation, which is implemented by the appropriate configuration of packet schedulers [4] [5] in nodes, so that each CoS is allocated with the desired amount of bandwidth, which is called Dynamic Bandwidth Reserve by Class. However, flow-signaled controlled class reservation mechanisms (as in [6]) face scalability issues as QoS control operations are triggered as the number of session requests increases. In other words, excessive control signaling messages generated with each stream not only entail a heavy CPU processing cost of the core routers, but also consume more bandwidth, memory and power while affecting the time to process the routers. session requests.

Alternatively to this (per flow) control, bandwidth over provisioning provides for reserving for each CoS more bandwidth than is really needed at any given time, thereby denoting overbooking per class. As a result, multiple streams are accepted without signaling on the network while reserved resources are still available, which may allow overall optimization of system performance. However, the unpredictability of session bandwidth requests makes the overbooking mechanism very challenging for bandwidth wastage, QoS violations (when a stream receives a different QoS than required), and increased likelihood of session blocking. In other words, some CoSs may be deprived of bandwidth while others would have unnecessary reservations, and this leads to a conflict between signaling rate optimization and the increased likelihood of session blocking. Therefore, bandwidth overbooking should be dynamically and efficiently managed to minimize resource waste and bandwidth, and thus avoid unnecessarily increasing the likelihood of session blocking and QoS violations, while significantly minimizing the cost of signaling. to improve system scalability. Driven by the importance of minimizing QoS control signaling costs to improve overall system performance, bandwidth overbooking techniques have been researched for many years.

Pan et al. proposed the Portal Edge Reservation Protocol (BGRP) [7], a mechanism for controlling bandwidth overbooking by a fixed amount. In BGRP, overbookings are adjusted when depleted, doubled, triple and so on. Although the present proposal reduces signaling load compared to per-flow solutions, its static mechanism for overbooking is not an efficient way to eliminate bandwidth wastage as it does not consider network state and need for resources. Similarly, the Semples Interdomain QoS Signaling Protocol (SIDSP) [9] has been proposed, which controls over reserves (called virtual trunks) of aggregate flows based on a prediction algorithm which considers system histories. SIDSP does not support any mechanism for dynamically controlling residual bandwidth (over-reserved but not used at any given time) between multiple trunks. Consequently, SIDSP leads to high bandwidth wastage, thus increasing the likelihood of session blocking. Dynamic aggregation of reserves for Internet services [8] over-reserves bandwidth for aggregate streams over multiple domains and reduces stored states and signaling messages over per-flow protocols. This mechanism does not detail the techniques and overbooking used.

In Sofia and others [18], static and dynamic bandwidth overbooking mechanisms have been extensively studied, and it has been stated that overbooking poses a major challenge if a reasonable operating point can be reached to deal effectively with conflict. that exists between signaling cost savings and increased likelihood of session blocking. Thus, overbooking is only perceived in the situation where an incoming request can be merged into an existing aggregate, but the referred aggregate does not have sufficient bandwidth reserve to automatically accommodate the request (see more details at [18]). ). The paper proposes the Protocol for Cross-Domain Shared Segment Aggregation (SICAP) to reinforce the provision of important functions for dynamic over-reserve control. However, SICAP leads to an unwanted increase in the likelihood of session blocking. As we have investigated, the fact that communication paths can share connections within the network is an important factor that makes overbooking bandwidth very challenging in finding a reasonable operating point to deal with the minimization conflict. signaling load and increased likelihood of blockage. In particular, the dynamic behavior of sessions on a given path not only affects that particular path, but also all correlated paths, cross traffic [26]. Any technique that correlates between paths its control mechanism.

SICAP's well-known approach issues do not provide benefit from information to improve the performance of patented work, advanced resource allocation for QoS (ASAC) [10], proposed specific functions to dynamically control overbooking for CoSs (aggregate flows) to improve system scalability. As with the SICAP solution, ASAC defines a function to calculate how much to overbook for a CoS, with the intention of overbooking sufficient bandwidth without achieving a high blocking probability. In addition, ASAC sets a maximum bandwidth reserve limit for each CoS, such that whenever a CoS is congested (with depleted reserved bandwidth unavailable), how much overbooking for a CoS is calculated for increase overbooking for CoS within the limit. Thus, when the boundary of a given CoS is congested, ASAC handles the residual bandwidth by readjusting the boundaries of existing CoSs. In particular, it removes a portion of the unused bandwidth in each remaining CoS and uses the sum to increase the boundary of the affected CoS so that the overbooking readjustment calculation can be done successfully. This mechanism provides important functions for dynamically controlling bandwidth overbooking in class-based networks. However, further studies of the ASAC mechanism have presented serious drawbacks to feature readjustment functions, which are explained below. First, the mechanism for calculating the new overbooking leads to wasted bandwidth, which is still undesirable, and thereby increasing the likelihood of session blocking. An illustrative example is in Table 1, where we assume that 3 CoSs are available in a bottleneck connection. As ASAC boot parameters, we set the bottleneck binding capacity to 60 bandwidth units, and the reserve limit compromised to 10 bandwidth units. The maximum reserve limit is x = 20 bandwidth units for each CoS, where u is the bandwidth used in a CoS, r is the new binding request for a CoS, and δ is the unused bandwidth taken. each of the remaining CoSs to decongest the affected CoS. Then, upon receipt of the request r = 0.85 bandwidth units for CoSl (congested with only 0.7 unused bandwidth units), the δ function removes 0.0169 bandwidth units, and 0.0124 units of CoS2 and CoS3 respectively, and add the sum of 0.0293 bandwidth units to the congested CoSl as shown in Table 1 with available Before and after readjustment.

Thus, the δ function increases the available CoSl bandwidth from 0.7 to 0.7293 bandwidth units, which remains insufficient to support the request for r = 0.85 bandwidth units. As a result, the request is blocked while the total unused bandwidth on all three CoSs equals 2.1 units of bandwidth (connection almost congested, but not higher than the blocked request.) bandwidth is wasted and the likelihood of session blocking increases unnecessarily.

Table 1. Example of resource inefficiency

Before Adjustment After adjustment CoSl CoS2 CoS3 CoSl CoS2 CoS3 % 20 20 20 20,0293 19.9831 19.9876 u 19.3 19.2 19.4 19.3 19.2 19.4 r llllslili NULL NULL llllsllil NULL NULL δ NULL 0.0150 0.0052 NULL NULL NULL

Second, for network topology and related resource status information, a Control Decision Point (CDP) configured with ASAC (inbound router) stores the list of interfaces in its local state table. paths, along with only the bottleneck interface capabilities (least resource reserved interface of all path interfaces) of each path. Therefore, whenever the local state table indicates available reserve with sufficient resources on a given path, it admits, resets or terminates session requests on the path without flagging it, and simply updates with the new bottleneck interface capabilities of that path. Thus, the use of ASAC raises an important issue with regard to correlations between paths and traffic load, since dynamic session behavior on a given path affects not only this particular path but also others sharing the router somewhere in the path. route (called correlation between paths). As a result, changes to a path can easily cause inaccurate information in the status table as to the capabilities of correlated path bottleneck interfaces, leading to QoS (different than demanded QoS provision) violation, which compromises system performance. Third, to readjust reservations on a given path, ASAC sets the amount of bandwidth to add or remove to CoSs based on the resource conditions of the path bottleneck interface only, as well as readjusting reserve parameters all interfaces belonging to the path by signaling. To help you understand this better, suppose that three consecutive interfaces Ii, Π, and I3 belong to one path P _lr while I2 belongs to another path P ₂ , which means that paths P ₂ and P ₂ are correlated by sharing the path. P ₂ interface. So suppose I3 is the bottleneck interface of P ₂ , and P ₂ is P ₂ , a situation that can easily occur in normal network scenarios. Thus, the resource adjustment ASAC mechanism P _2, ie for all connections P _2, P ₂ and I3 using defined parameters based on resource conditions affect the R ₃ neck interface I ₂ Q _2. This is essentially due to the fact that ASAC does not consider the resources of all path interfaces for its calculations, so any change in a particular path will certainly affect the other correlated paths, which is aggravated with the scale. from the Web.

According to these studies, we claim that an efficient bandwidth overrun mechanism needs to adopt appropriate functions to: (a) take into account the patterns of correlation between paths and traffic; (b) calculate the appropriate bandwidth to over-reserve each CoS: (c) deal with residual bandwidth to avoid wasting resources, class monopoly, and thus avoid increasing the likelihood of session blocking unnecessarily.

Summary of the Invention

This invention provides a method for dynamically controlling bandwidth in class-based networks without QoS violation, while efficiently minimizing signaling load, bandwidth wastage and the likelihood of session blocking.

It allows the system's Decision Control Point (CDP) to build concrete patterns of correlations between communication paths by comparing the output interfaces that make up each of the existing paths. Briefly, a CDP detects all paths that

share interface (s) get out create a list of ways correlated with each one with cost benefit • 0 CDP keeps on your bank of data place identifications of all interfaces outputs that make up

the existing paths, along with their current conditions regarding their capacity, reserved bandwidth, bandwidth used, and configured CoS available bandwidth. To this end, whenever a CDP admits a session, it adjusts its QoS in a given path, it automatically updates (in its local database) the resource status of all outgoing interfaces that make up that path according to the granted QoS. of the session during your trading. Thus, it is guaranteed to accurately maintain the resource information of each CoS associated with each outgoing interface in the local state tables, considering that traffic behavior is unpredictable, not only by conditioning the network boundary for load control. , but also by forcing all packets belonging to a stream to be aggregated into one CoS / path to follow the same path through appropriate technology (such as MPLS, IP multicast, VPN, etc.).

In addition, upon admission of a session, it updates the status of the outbound interface features of the paths, resets the QoS of a session, or terminates it in a given path, a CDP correctly maintains the available bandwidth for each configured CoS. at the bottleneck interface of all existing paths within a network. This is done by taking into account path and traffic correlations to avoid QoS violation or wrong admission control decisions.

CDP provides proper calculation for bandwidth overbooking (functions from (1) to (5)) to avoid wasting bandwidth and increasing the likelihood of blogging while significantly minimizing signaling load.

New bandwidth override parameters are dynamically set (as needed) for each CoS of each interface and a given path according to the resource conditions of the interface itself, not those of the path bottleneck interface. . Thus, the defined parameters are encapsulated in well-supplied QoS Specifications QSPECs and transmitted to nodes along the way by a proper signaling message. Thus each node intercepts and interprets the message to correctly configure its outbound interface according to the QSPEC derived information intended for it. In this way, system performance is improved by avoiding resource configuration on a given interface (which can be shared across multiple paths) with heavy reliance on resource conditions on the path bottleneck interface.

This invention is applicable to Next Generation Networks (NGNs) to improve their performance and QoS control with lower signaling, status and processing costs, as well as connection request setup time. The solution is simple while supporting current and future service paradigms (eg data, multimedia, custom services, etc.), and operators can easily provide more value-added services as well as increase their profitability. It can be implemented by software, hardware, or by a combination of both. The implementation of the modules and features described in the present invention may be integrated into a node (router for example) of a class-based network with facilities for setting bandwidth parameters. In centralized network architectures (eg TISPAN [20], 3GPP [25], etc.), your topological information is maintained by a CDP (like a centralized CDP in our illustration), which is in charge of locally controlling all the ways.

The beneficial and economic effects of this invention are various:

• Bandwidth is over-reserved by CoS while resource waste is optimized and the likelihood of session blocking not increased unnecessarily. In addition, processing reserve states are thus improving the

QoS, signaling costs and significantly minimized system scalability. Operators will be able to provide more services, and more users, for the same bandwidth with less complexity in configuring service connections.

• Users will experience short call request times and sessions of an acceptable quality level throughout their duration. This way both QoS and user perception will be improved.

DESCRIPTION OF THE FIGURES

Figure 1 - Illustrates ACOR by means of a flowchart.

Figure 2 - Illustrates the reference architecture of QoS control and ETSI admission (TISPAN).

Figure 3 - Illustrates a sequence of session establishment flows within the TISPAN reference architecture.

Figure 4 - Illustrates the mapping of TISPAN functional modules in illustrative network architecture.

Figure 5 - Illustrates the mapping of ACOR agents in illustrative network architecture.

Figure 6 - Illustrates the state of the initial path resource reserves.

Figure 7 - Illustrates the simplified drawings of the

Figure 6 for ease of understanding illustrative descriptions.

Figure 8 - Illustrates the dynamism of sessions allowed by the edge router (BR1).

Figure 9 - Illustrates resource readjustments in requesting paths of function (4).

Figure 10 - Illustrates resource readjustments on requesting paths of functions (4) and (5).

DETAILED DESCRIPTION

This invention provides a mechanism for processing Class-based Advanced Bandwidth Over-provisioning (ACOR).

In order to demonstrate how this invention can be used for cost-effective QoS provisioning and admission control in Next Generation Networks (NGN), the following are the proposed features for bandwidth overbooking control, using the QoS reference architecture and admission control of the European Telecommunications Standards Institute (ETSI) ES 282 003 [21], ETSI ES 282 001 [20], ETSI TS 183 060 [19], known as telecommunications and converged services Internet and advanced network protocols (TISPAN); NGN Functional Architecture; Appeals and Admission Control Subsystem (RACS).

1.1 TISPAN Architecture Overview

As shown in Figure 2 (taken from ES 282 003 [21]), the QoS TISPAN control architecture is made up of four main subsystems: (1) Function to application (AF) subsystem is the functional entity that should provide the explicit session description, with its QoS requirements, to express the expected service of a network; (2) Resource Admission Control Subsystem (RACS) is TISPAN's NGN subsystem responsible for implementing procedures and mechanisms for dealing with policy-based resource reservation and admission control of unicast and multicast traffic on access, core and client installations, allowing applications to request resource reservation on transport networks in the RACS control domain; (3) Function subsystem for the transport process, includes basic elementary functions to support resource reservation execution, packet forwarding, and more specifically group of functions defined as functional entities; (4) 0 Network Connection Subsystem (NASS), which is used for dynamic provisioning of IP addresses, other endpoint configuration parameters, as well as network access authorization based on user profiles. The main functional entities and interfaces that make up each of the subsystems described above (as shown in Figure 2) are briefly described below.

Application Role (AF): This subsystem must provide explicit session description with QoS requirements (for example: IP domain identifier, Requester name and class of service, media description, service priority) a Decision Role Service Policy (SPDF) through the Gq 'interface [22], to express the intended service of the Admission Control and Subsystem Control (RACS) subsystem. As such, requests may include resource allocation for new sessions, Resource Modification such as enhancing or demoting the quality of existing sessions, and release of finished session resources;

Service Policy Decision Function (SPDF): This functional element is responsible for making policy decisions (for example, service authorization) on requests received from an interconnected AF or SPDF, based on service policy rules defined by the network administrator. In addition, SPDF performs a coordination function between AF, RACF, edge portal function (BGF), interconnected SPDFs, or any combination of them. In particular, when a policy decision given order flow is rated as positive, the SPDF must instruct the BGF (through the interface I _a) to implement policies (such as traffic conditioning filter, QoS marking, bandwidth limitation , bandwidth usage measurement, etc.) that limit / control flow at an allowable level indicated by the SPDF. In addition, SPDF forwards to RACF session and QoS-related information (received by interface Gq ') via interface Rq in order to request resources on the network;

Edge Portal Protocol (BGF): 0 BGF provides interface between two IP transport domains. It may reside at the network boundary, and implement packet marking, modeling, policing, or network address translation (NAT);

Feature Control and Admission Function (RACF): The RACF entity receives resource requests from SPDF (such as bandwidth) via the Rq reference point in Push mode, or from the RCEF via the Re reference point in Pull mode. Requests provide indication of desired QoS characteristics (bandwidth, IP domain identifier, requester name, intended Class of Service, etc.). Further details on Push or Pull modes are available in ES 282 003 [21].

RACF then performs admission control based on information such as session-wanted QoS, user profiles received from the e4 interface subsystem (NASS), and resource availability gained from knowing its topology network and associated resource status. In this regard, RACF can have a complete or partial view of the network topology and its resources, including congestion points, current reserves, etc. Thus, in order to dynamically control QoS reservations (ie, reserve / readjust / release resources on the transport channel when necessary), RACF derives resource reservation policies (eg, reservation processing or overrun functions). -reserves) and sends appropriate instructions (for example, command or signaling) to the resource control enforcement (RCEF) function, which is responsible for executing the maintenance, readjustment, or removal policies of what it is (installation, reservation) on the transport. Thus, appropriate QoS control decisions are applied to communication paths in a network to ensure QoS to unicast or multicast sessions during their duration.

Network Connection Subsystem (NASS): NASS is responsible for providing control information, which includes: dynamic provisioning of IP addresses; terminal configuration parameters; IP layer authentication before or during the address assignment procedure; network access authorization depending on the user's profile; and IP-level location management;

Resource Control Enforcement Function (RCEF): RCEF is a transport processing functional entity charged with, under the command of RACF, applying policy functions for unicast and / or multicast. Therefore, RCEF applies the policies autonomously or in conjunction with the basic transport function (BTF) to enable transport control service actions. In this sense, RCEF can interact with BTF to enforce policies that impact data sending behavior, such as multicast stream data replication. In addition, RCEF includes other features, such as resource allocation for upstream and downstream traffic. And, depending on local control policies, RCEF may notify RACF of predetermined events (such as response messages, connection / node failures, control malfunction, etc.) that have occurred on network elements. Then RACF may decide to change / remove existing policies, install new policies, or escalate the case to the top tier.

Basic Transport Function (BTF): All network transport segments include one or more basic transport functions and implement the Elementary Dispatch Function (FEP) to route data traffic and the Elemental Control Function (ECF), to process the unicast and multicast data control protocol (e.g. signaling control, routing protocol, etc.). Therefore, the ECF on one node may decide to send control protocol data to another ECF or interact with one or more EFFs in order to establish new or modify existing data sending behavior by manipulating the routing bases in the nodes. related. In general, almost all physical elements of the network (bridges, routers, etc.) usually contain a BTF, and may even contain more functional entities such as RCEF.

Figure 3 illustrates a scenario with interactions that may occur between different entities to handle a push request from a QoS-sensitive session. As summarized below (more details are available in ES 282 003 [21]), you can see that:

1) An AF login message is received from a User Equipment (- UE) generated on its own or another AF. AF identifies that this session requires resources on the transport network to support the associated media streams;

2) AF sends SPDF unicast or multicast service request information;

3) The SPDF processes and authorizes the request;

4) Once the service is authorized, the SPDF contacts the access RACF (A-RACF) to control the admission of the resource;

5) RACF maps SPDF request to network resources and local admission control policies. Thus, ARACF decides whether there are traffic policies to be installed in RCEF (for example, whether resource readjustment or reservation is required) based on user profiles, session QoS requirements, and knowledge of resource availability;

6) In this example, A-RACF derives appropriate control policies (such as new resource configuration parameters in the path) and requests RCEF to install traffic policies to be applied to associated streams;

7) RCEF confirms the installation of traffic policies. For multicast services, traffic policies (e.g., packet replication) may be required for proper distribution of multicast stream (s);

8) A-RACF sends a message to inform SPDF if resources are reserved;

9) SPDF has determined that serving this request requires a request to be sent to the appropriate BGF, and therefore SPDF sends a message to BGF;

10) BGF performs the requested service (traffic conditioning, address translation, etc.), and confirms the operation to the SPDF;

11) The SPDF forwards the result to AF and traffic transmission can start if admission control allows.

Three agents are specified to facilitate the use of the ACOR solution on Next Generation Networks compatible with the standard TISPAN architecture, as described in the following sections.

1.2 ACOR Agents

Taking into account the positions and features of network nodes, Figure 4 presents a mapping of TISPAN functional modules over the illustrative network architecture. As can be seen from Figure 4, SPDF and RACF are mapped to the Decision Control Point (CDP), allowing the latter to take full control of the network domain. In addition, each edge / border router (- BR) includes BGF, RCEF, and BTF functions, while RCEF and BTF functions are mapped to each interior router.

Using the standard TISPAN architecture as an illustration, ACOR defines three agents that can be easily deployed across multiple nodes to ensure proper interaction and low cost QoS and admission control. As shown in Figure 5, these agents include: (1) ACOR-FULL (ACOR-F), a full state agent specified primarily for network Decision Control Points (CDPs); (2) ACOR-Light (ACOR-L), a specified lightweight agent with basic functionality required by all routers; (3) ACOR-Edge (ACOR-B), a half-full state (reduced state) agent specified primarily for network edge routers (BRs). More details about ACOR agents are provided in later subsections.

1.2.1 ACOR-Filled Agent

ACOR-F is a full-state agent designed to be hosted primarily at the Decision Control Point -CDP- (eg QoS Broker) to ensure bandwidth overbooking and admission control across the network.

For this purpose, ACOR-F implements the Service Policy Decision (SPDF) function to enable CDP to receive session requests via the Gg 'interface (eg, call requests, readjustment, or call termination). as well as handle them properly for authorization depending on policies set by your network administrator.

In addition, ACOR-F includes the Feature Control and Admission Function (RACF). Therefore, it applies the ACOR functions from (1) to (5) and is appropriately used (when necessary) to set bandwidth override parameters to be configured (installed, readjusted or removed) on the various CoSs deployed on the nodes. along data paths within the network to efficiently minimize bandwidth and therefore the likelihood of blogging while significantly minimizing the overall cost of QoS control.

ACOR-F coordinates the creation of communication paths within a network with the flexibility to use any technique specified by the network administrator (such as shorter paths [14] [13], flood [15], tree population [16], etc.). ACOR-F also maintains a good and accurate knowledge of the underlying network topology (over its control) with the associated resource status, with a view to improving admission control and the efficiency of network resource utilization without QoS violation, while significantly minimizing the cost of signaling. In particular, it stores in local databases the ID of all output interfaces within the path, along with capacities, reserved bandwidth, bandwidth used and bandwidth available for each CoS of each interface. Additionally, it stores the list of output interfaces that make up each path and associates all paths with appropriate variables that store the bottleneck interface's QoS capabilities (called path capabilities). These path and interface capabilities are maintained in real time in local databases. To that end, whenever ACOR-F automatically admits, resets, or terminates session requests to a CoS on a given path, the new resource states (for example, reserved bandwidth, bandwidth used, and bandwidth each interface upgraded to available) automatically from the path, are the CoS in the local database according to the QoS granted to the session. Then the capabilities of all paths that share one or many of these interfaces (ie, the correlated paths) are also automatically updated by simply comparing (in the local database) the current capabilities of the output interfaces that make up the paths. . In addition, ACOR-F records each active session along with its stream information, in terms of session description and QoS requirements (for example, session ID, stream IDs, session source and destination IDs). , CoS and the mapped path to a stream, the bandwidth granted to each stream, etc.). This is called per-flow session information accounting, where information is used to properly instruct BGF functions, as well as efficiently control active flows, to ensure the differentiated QoS granted to each flow during QoS negotiation. Flow states are required in the network CDC, especially when it is designed that the control complexity is intended for the network boundary and the interior nodes remain simple (for storing only reserve state by CoS and not by flow). Thus, when a change occurs, ACOR-F easily updates and maintains its local database with the accurate (broadband) resource state of all outgoing interfaces on the paths, as well as the capabilities of all existing paths on real time, while addressing communication path correlations and passing traffic issues, to avoid degradation of QoS control performance. This assumes that the minimum QoS requirements of active sessions (delay, delay variation, or loss are guaranteed until the required bandwidth is properly provisioned and the traffic load is controlled at the network boundary.

properly

In addition, ACOR can be easily adapted to the local model used by the operator for QoS control and policies. An example is: ACOR can be integrated with the use of periodic control messages to monitor network performance as needed. Thus, ACOR-F ensures proper overbooking and admission control in a network by jointly exploiting ACOR functions, session QoS requirements obtained from the application function (FA) via interface

Gq ', NASS user profiles via the e4 interface, and network resources and the functions of the via the Ia interface accordingly.

obtained from systems their good knowledge of state paths while instructing) GF for traffic control

In addition to storing information for intra-domain control as described, one may also store path segments or paths between boundary nodes with related information as needed, and use the same approach for end-to-end QoS control. over various domains with heterogeneous technology. This good view of the state of cost-effective network resources is important for many system optimization purposes: it is used for traffic engineering to detect and avoid congestion points to improve session-to-QoS mapping / paths, and also for network management and capacity planning.

1.2.2 ACOR-Light Agent

ACOR-L is a lightweight agent that implements the necessary functionality of a normal network node (for example, normal router) to execute resource processing instructions from

control such as interception and messaging, standby facility resource readjustment configuration, resource removal configuration, data transmission, data replication to the multicast tree, execution response, and failure notification of network.

To this end, ACOR-L implements the Resource Control Application Function (RCEF) (which interacts with ACOR-F via the Re interface) and the Basic Transport Function (BTF) to include appropriate intercept functions. , interpret and process messages, and ensure QoS data transport over a network. Thus, to ensure that the overbooking control instructions transmitted via signaling commands or messages are intercepted at nodes visited along the way, the ACOR-L agent allows UDP port recognition (routers are constantly listening for the UDP port) or the IP Router Alarm Option (RAO) [11] on the nodes. In addition, it is capable of interacting with resource management functions - RMF (eg [12]) to configure programmers on nodes [4] [5] to ensure that each CoS receives the amount of bandwidth allocated to it. In addition, it interacts with legacy protocols (eg routing protocols) or manipulates appropriate databases on nodes (eg information management base - MIB, information routing base -RIB or Multicast information routing base). -MRIB, basis of dispatch of information -FIB, etc.) according to the control instructions received. As an example, ACOR-L allows you to collect interface IDs in paths (for example, path registry object -RRO- [13] [17]) along with interface QoS capabilities (for example, interface capability) , to help the network Decision Control Point build its local database for link topology and resource details. In addition, it allows nodes to store the ID of the previous interface visited by the signaling message, which information is used to avoid reverse path problems [1] [13], while it implements the ACOR (1) function in support of override. -reserve the initial bandwidth at system startup phase.

In summary, ACOR-L implements fewer functions, but sufficient to ensure QoS data packet routing, and multicast service packet replication while detecting and reporting link / node failure events to the decision control entity , which in turn makes the appropriate decision to process events (for example, change of itinerary).

1.2.3 ACOR-Edge Agent

ACOR-B is a lightweight agent that implements the necessary features at the network boundary (for example, Edge Router -BR).

For this effect, O ACOR-B includes all at basic functionality implemented in ACOR-L agent (for example, processing in posts, The installation / readjustment / removal gives reserve, notification in failure, etc.) In addition, B implements Function in Portal Edge -BGF- in order to execute the conditioning in traffic (for example measure O use of bandwidth, remodeling and policing in traffic out of profile, etc.) to control the debt in traffic of each flow (at border of network) according with the control instructions

which it receives from ACOR-F via interface Ia. In addition, ACOR-B is responsible for its own inter-domain connection, which may be achieved on the basis of Service Level Agreement, SLA or specific contracts between network operators, while ensuring the need for tunneling or network address translation (NAT).

To further facilitate the understanding of this invention, the rest of the document illustrates ACOR functionalities and appropriate interactions between different agents using the single domain network infrastructure of Figure 6 for simplicity. Thus, in a large network comprised of multiple domains (multiple Autonomous Systems (ASs)), each domain is allowed to implement its own QoS model and connect to neighboring domains based on service level agreement (SLA). or specific contracts through their ACOR-B agents deployed at the network boundary. In this case, Decision Control Points -CDPs from different domains should interact appropriately to construct end-to-end paths over the various domains, as suggested in the TISPAN reference architecture (ES 282 003 [21]).

1.3 ACOR Illustration

For simplicity, Figure 6 introduces a Decision Control Point -CDP- that hosts the ACOR-F agent as responsible for overall network control, 5 Edge Routers ((BR1 through BR5) hosting the ACOR-B agent each one, configured as an ingress / egress router such as network traffic domains and 4 indoor routers (Cl through C4) hosting ACOR-L, as detailed below

So with the proper interactions between us. we implemented four CoSs: one Control Signaling (CS) CoS, with a fixed reserve b = 1 bandwidth unit, and three service CoSs such as expedited dispatch (EF), right dispatch (AF) and best attempt (BE) , respectively, with weights of 40%, 30% and 30%.

1.3.1 Network Boot

At system startup, the network administrator has a strong influence on network startup. Thus, at system startup, when a session is not yet active, CDP instructs all Edge Routers (BR1 through BR5) under its control to create edge-to-edge paths from each BR to any other BR within the system. same domain. Thus, it collects the information collected during path creation and builds its local database, while each BR builds its database as well, as illustrated in Figure 6, and in more detail below.

Therefore, CDP sends initialization message to each BRi (BR1 through BR5) within the domain (in event (a)). Thus, upon receiving the CDP boot instruction or command, each BRi implements the appropriate signaling protocol (eg [13]), using short-path techniques as an example, creates multiple paths for itself to other BRis within the domain ( see Figure 6), while collecting important path-related information, such as each path's list of interfaces, the capacity of each interface, and the multicast channel built into the path (for example, (BRi, BRj)). In this work, a multicast channel (BRi, BRj) built on a path is such that BRi is the identification of the BR-ticket that originated the path creation when BRj is the BR-egress where the edge-to-edge path ends. . A multicast channel is configured per path to force all packets (in one path) to follow the same associated path. Similarly, you can use Multi-Protocol Label Switching (MPLS). Thus, as shown in Figure 7, a directed connection is represented by an arrow indicating the ingress-to-egress direction of the paths it is in, where the ID of an interface denoted I _t is the identification (e.g., IP address or MAC) of the interface on the way. Thus, after a BR creates its paths (in event (b)), it assigns a unique ID per path and stores the path IDs in its local database (in event (c)), the associated multicast channels. and the list of interfaces. This information is required by each BR to map active streams to the appropriate CoSs / paths and correctly control the traffic load of each data stream according to the instructions it receives from CDP. After that, it sends (in event (d)) to CDP all paths with related information, such as the list of interfaces in the path, the multicast channels, and the capacity of each interface. Thus, upon receiving this information, CDP locally deducts the initial reserves for each CoS on each interface in the path, using the capacity of each interface, the reserve b defined for the control signaling CoS, and the ACOR function (1). . Thus, it is not necessary to collect the parameters of the initial reserves of nodes in the path, an important strategy to avoid unnecessarily increasing signaling message size. It then correctly populates its local database on events (e), (f) and (g), as shown respectively in

Table A, Table B, and Table C from Figure 6. In Figure 6, Table A stores and maintains the ID of each interface in the path with the associated resource capabilities. Thus, Table B is used to store the list of interfaces in the path (with path-related information) where the capabilities of the path bottleneck interface are easily obtained by comparing the capabilities (stored in Table A) of the interfaces that make up the path. In addition, Table C is used to store the list of paths that correlate with any given path within the domain. Thus, to update the capabilities of correlated paths when changes occur in a given path, ACOR simply retrieves (from Table C) the list of paths that correlate with the path and updates the information according to less processing overhead.

For bandwidth overbooking, the C parameter used for an interface is considered to be the total bandwidth for the service CoSs implemented on the interface. In other words, suppose that (k + 1) classes of Services (CoSs) are implemented on each interface in a network, among which one CoS is dedicated for Signal Control (CS) or other purposes, and must reserve a width of static band b defined by the network administrator in support of reliable control traffic transport, which means that only the remaining CoSs k are used to transport QoS differentiated services. Thus, if Bw is the total capacity of an interface, then the quantity (C = Bw-b) is the total bandwidth for the remaining CoSs k on the interface. Thus, as shown in Figure 6 at system startup, each service CoSi (1 bi <k) on each network path interface is initialized with a B ± reserve obtained based on a percentage cq also called the CoSi weight, using function (1) as follows:

H = a.C such that:

k

The. = 1 i = 1

As can be seen, the weight of a CoSi can be flexibly set by the network administrator (for example, according to specific policies, or according to usage). In other words, ACOR functions can be easily adapted to the specific operator QoS control policies. In addition, the initial reservation for each CoS on each interface in the path can be made in several ways, taking into account that duplication of the reservation should be avoided.

One possibility is as follows: All routers within a network domain can be configured with the same CoSs weight set as well as the Signal Control CoS CS bandwidth b. Thus, when a router starts, initial reservations are automatically set and installed on all its interfaces. Thus, when a BR uses a signaling protocol to create paths, the signaling message simply collects the list of interfaces on each path, the associated multicast channel, and the capacity of each interface. However, this option requires a network administrator to reconfigure all routers whenever the weight set is to be changed and is therefore not scalable.

In this sense, an alternative is: the weight set for CoSs and the amount of bandwidth b are set or configured in CDP only. Thus, when CDP instructs BRis to create paths, it provides BRis with the set weight and bandwidth b defined for CoSs (identified based on associated CoS IDs), allowing BRis to encapsulate this information in signaling messages. . Thus, each node visited by the signaling messages in the path collects the weights and b, and defines and installs the initial reserves on their appropriate interfaces, if these reservations do not yet exist. The signaling messages collect the list of interfaces of each path, the associated multicast channel, and the capacity of each interface. In this way, the weights and the b-value can be set dynamically or adapted as needed by the network CDP, provided that new parameters can be easily applied to all nodes via signaling.

As you can see in Figure 6 after the network is initialized, the paths as well as the initial reservations are created, and the network can be put into operation for users.

1.3.2 Overbooking Control

With the network initialized in Figure 6 in service, users' unpredictable requests would be received by the CDP, which will process and map requests for access to the appropriate BR (BR1 through BR5) using the overbook QoS and the admission control mechanism. THE COLOR. Based on this network in Figure 6, the ACOR control mechanism is illustrated below, where only mapped session requests to BR1 are studied for simplicity. This scenario is chosen because it clearly shows how ACOR takes account of path correlations and cross-traffic to correctly maintain the local database in real time. In this sense, Figure 7 shows mainly the CDP database information that will be influenced by session dynamics in paths originating from BR1. Thus, Table B of Figure 7 stores not only paths originating from BR1, but also all paths from other

BRs that relate to them, and correlation information that is obtained from Table C in Figure 6. Thus, after system startup, CDP checks the available bandwidth of 39.6 bandwidth units, 29.7 bandwidth and 29.7 bandwidth units, respectively in CoSs EF, AF and BE on edge-to-edge paths created by BR1 (BR1: pathO to path3) while the bandwidth parameters used in Table A are all null as no session is currently running. Consequently, the RACF entity can accommodate many session requests on appropriate BR1 paths without signaling the paths as long as the available bandwidth is sufficient in the paths. In this way, overbooking significantly minimizes the QoS reservation control signaling number. Thus, upon session admission, session QoS readjustment or session termination on a given path, CDP automatically updates the width bandwidth used on all interfaces belonging to that path, as shown in Table A of Figure 8, where the bandwidth used on multiple CoSs on multiple interfaces indicates the amount of bandwidth granted to active sessions on the related paths. In addition, CDP automatically updates the capabilities, not only of the paths in which sessions run, but also the capabilities of all correlated paths, as in Table B of Figure 8. From this Table B, it is important to note that, Although no session is performed on paths originating from other BRs (BR2, BR3, BR4 and BR5), the capabilities of some of its paths have changed according to the patterns of correlations within the network. Therefore, in dynamic network scenarios, especially in multiple input router environments, path correlations and cross-traffic must be taken into account to effectively support bandwidth overbooking techniques. Thus, CDP correctly maintains the capabilities of all paths and interfaces within a real-time network. In addition, in request events (eg, session admission, session readjustment, or session termination), CDP instructs the appropriate BR (BR1 in this example) to include the BGF function via interface Ia to traffic conditioning and traffic control (eg remodeling or policing of out-of-profile traffic) can be properly performed at the network boundary.

However, if upon receiving a χχ request for a given CoSj on a given path, the admission decision functions detect that the available bandwidth on the path is insufficient to ensure acceptable quality, the appropriate ACOR functions are triggered for reconfiguration. of resources along the way. In Figure 9 the CDP receives (in event (a)) a request for 3 bandwidth units to connect to CoS EF on path3 (linking BR1 to BR2), while the available bandwidth in class EF on path3 is only 1 unit of bandwidth, ie insufficient. Obviously, the request cannot be admitted to CoS EF without violating QoS, but CoS EF's reserve can be sufficiently increased with unused resources in bottlenecks. In this sense, CDP, especially the RACF entity, uses ACOR functions (2), (3), (4) and (5) in conjunction with the resource state of the interfaces concerned (12, 15 and 18) in their local database (Table A) of Figure 9), and locally defines (in event (b) in Figure 9) new Bi reservation parameters for each CoSi on each interface on the affected path (path3, such as I2 connections). , I5 and Ig (Table A) in Figure 9). For each CoSi on each of the I2, I5, and Ig interfaces, the new bandwidth buffer parameters are defined taking into account the resource conditions of the interface itself, as follows. First, RACF gets the total unused bandwidth A _T at the interface with the following function:

Δ _τ = Σ (Βϊ- ^υ ι) i = l

Where ui is the bandwidth used in each CoSi.

Then, if the request rj for CoSj is less than or equal to the weighted proportion of unused interface bandwidth (rjSaj.A _T , RACF obtains the new buffer parameters Bg for each CoSi on the interface with function (4) as follows. Follow:

B = Uj + .Δ _τ

However, if the request rj for CoSj is greater than the weighted proportion of unused interface bandwidth (rj> otj. Δ _τ ), the RACF obtains the new reserve Bj parameters for the affected CoSj, and the new parameters of the

Bg reserve for each one of CoSi remaining (i B j) in interface with function (5) as follows: B j = Uj + η + α _Γ (Δ _τ -Γ Bi -Uj + ajXAy-q), l <i <k, (i © j) Thus, 0 RACF defines locally the new reservation parameters that should be set up for each

CoS on each of the Ig, I5, and Ig interfaces (see event (b) in Table A) in Figure 9).

Thus, if the request rj is greater than the total unused bandwidth (rj> A _T ) on one of the interfaces (I2,

Ι5 or Tg), which belong to path3, path3 is considered congested as there is insufficient unused resources on the path to accommodate the request with acceptable quality. In this case, RACF denies the request, or renegotiates the session QoS requirements, depending on the service level agreement (SLA) or the operator's local control model and policies.

However, as long as the zg request is less than or equal to the total unused bandwidth (rjkA _T ) on all interfaces in path3 (I2, I5, and fg), the ACOR reservation reset mechanism leads to session admission. successfully. Thus, ACOR efficiently reuses the total reserved bandwidth that is not used in the different CoSs (Δ _τ ) at each interface in the path, in a way that minimizes wasted bandwidth and blocking probability while significantly minimizing the Number of signaling messages.

With a numerical example with Table A) of Figure 9, the new reserve parameters for each CoS on I2, I5, and Tg connections after receiving (in event (a)) the request for 3 CoS EF bandwidth units on Path3, while the available bandwidth on the bottleneck link is 1 unit of bandwidth, are given below:

12: Δ _τ : (399.6-64) + (299.7-30) + (299.7-38) = 867;

η _ΕΕ ·. 867 * 40% = 346.8.

rjSq _EE , then the reserves are obtained for each CoSi using function (4), such as:

B _ee = 64 + 867 * 40% = 410.8 B _af = 30 + 867 * 30% = 290.1 B _af = 38 + 867 * 30% = 298.1

15: Δ _τ : (39.6-39) + (29.7-10) + (29.7-15) = 35;

η _ΕΕ : 35 * 40% = 14 rjdqEp, so B ₂ reserves are obtained for each CoSi using function function (4), such as:

B _ef = 39 + 35 * 40% = 53 B _af = 10 + 35 * 30% = 20.5 B _af = 15 + 35 * 30% = 25.5

18: Δ _τ : (39.6-10) + (29.7-3) + (29.7-10) = 76; η _ΕΕ : 76 * 40% = 30.4 rjbzfr, so Bi reserves are obtained for each CoSi using function (4), such as:

B _ef = 10 + 7 6 * 4 0% = 4 0.4 B _af = 3 + 76 * 30% = 25.8 B _af = 10 + 76 * 30% = 32.8

Thus, as the new standby parameters are successfully set for all interfaces that belong to path3 (in wind (b) in Table A) in Figure 9), CDP gives instructions (in event (c)) to BR1 to new buffer parameters are set on the affected Path3. To this end, CDP provides information to BR1, such as path identification and new buffer parameters for each interface in path3 (12, 15 and 18), whose parameters are correctly encapsulated in well-marked QoS specification objects ( QSPECs). Thus, upon receiving the instructions, BR1 first installs (in event (d)) the new configurations for its J ₂ interface. After that, it composes a suitable signaling message and sends in Path3 (in event (e)) the parameters destined for the remaining interfaces that are in the path (I5 and Ig). Thus, when the message is intercepted by the inner router, node C2, the node inspects the message and retrieves the spare parameters for its interface I5, which is recognized as long as the spare parameters are encapsulated in appropriate QSPEC objects, marked with the IDs of the associated interfaces. So C2 correctly installs the new reserves in interface I5 (in event (f)). It then deletes its QSPEC object from the message and forwards the message to the path in event (g). Excluding QSPEC not only reduces the size of the message, it also indicates that the reservation was successfully made on the interface. Similarly, C3 intercepts the message, reads the new reservation parameters, and correctly installs the new reservations into its Ig interface in event (h). It then deletes its QSPEC object and forwards the message in event (i). Therefore, a node where the reservation fails due to malfunction or connection failure must send a failure notification to CDP without deleting its QSPEC object. As BR2 receives the message, it composes the corresponding response message with a flag indicating that the reservation was made successfully and sends it to BR1 in event (j). Thus, ACOR resets the reservation on each interface based on the interface resource conditions and not based on the state of resources on the interface bottleneck. Thus, upon receiving the signaling response message, BR1 notifies (in event (k)) CDP of the success of the reservation. Thus, CDP updates its local database, as in event (b) (Table A) for link states and event (1) (Table B), where CDP updates the available bandwidth in connection bottlenecks. each path sharing one or some of the affected connections (L2, L5, and L5). It is important to note that Table B of Figure 9 clearly shows that session dynamics on BR1 path3 do not only affect the capabilities of that path3 only, but also affect other correlated paths, such as paths created by BR2, BR3 BR4 and BR5, and even other paths created by BR1 (BR1:

path2 and path3). In the figures, the notation (BRi: Pj) represents a path Pj originating from BRi. Thus, path correlations are taken into account to ensure correct information in the local CDP database as long as the session dynamics on an interface directly affect all paths it is on. As a result, the available bandwidth in CoS EF at pathway bottleneck 3 has been upgraded from 0.6 bandwidth units to 14 bandwidths, and CDP can accommodate the new request for 3 bandwidth units.

Although the example in Figure 9 mainly shows the use of ACOR function (4), it does not require function (5). In this regard, the following example uses Figure 10 to illustrate how the ACOR function (5) helps to efficiently use bandwidth during periods of high network utilization to effectively prevent bandwidth wastage and thus avoid increasing the likelihood of blocking unnecessarily while significantly minimizing the signaling load. In particular, CDP receives a 1 unit bandwidth request for CoS EF on path3 as in Figure 10, when the available bandwidth on the path3 bottleneck link is 0 bandwidth units, while the total Residual bandwidth at the path3 neck connection is exactly 1 unit of bandwidth.

Thus, ACOR resets reservations on all onPath3 links 12, 15, and 18 as follows:

12: Δ _τ : (410.8-88) + (290.1-48) + (298.1-50) = 813, η _ΕΕ : 813 * 40% = 325.2 rjlr / EF, so the B ± reserves are obtained for each CoS, using function (4) such as:

B _ef = 88 + 813 * 402 = 413.2 B _af = 48 + 813 * 302 = 291.9 B _af = 50 + 813 * 302 = 293.9

15: A _T : (53-53) + (20.5-20) + (25.5-25) = 1; η _ΕΕ : 1 * 402 = 0.4

7> η _ΕΕ while rjbA _T , then the Bj reserves for CoSj (FE) and Bi of each remaining CoSi (AF and BE) are obtained using function (5) as follows:

B _ef = 53 + 1+ (1-1) * 402 = 54 B _af = 20+ (1-1) * 302 = 20 B _af = 25+ (1-1) * 302 = 25

As can be seen from ₁₅ , the readjustment of ACOR allows for admission as long as the request is less than or equal to the unused total bandwidth (rjbA _T ) at the path bottleneck links. Therefore, ACOR effectively avoids wasting bandwidth and thus avoids increasing the likelihood of session locking unnecessarily.

18: A _T : (40.4-21) + (25.8-6) + (32.8-18) = 54; η _ΕΕ : 54 * 402 = 21.6.

rjbr] _EF , then the reserves Β _± are obtained for each CoSi using function (4), such as:

B _ef = 21 + 54 * 402 = 42.6 B _af = 6 + 54 * 302 = 22.2 B _af = 18 + 54 * 302 = 34.2

Thus, as the new standby parameters are successfully set for all interfaces that belong to path3 (in event (b) in Table A) in Figure

10), CDP instructs (in event (c)) BR1 for the new spare parameters to be set in the affected path3. To this end, CDP provides BR1 with the necessary information, such as path identification and new buffer parameters for each interface (I ₂ , Is and fg) in path3, whose parameters are properly encapsulated in well-marked specification objects. QoS (QSPECs). Thus, upon receiving the instructions, BR1 first installs (in event (d)) the new settings on its I ₂ interface. After that, he composes an appropriate signaling message and sends it to the path3 (in the event (e)) with the parameters for the other interfaces that are on the way (I5 and J _g). Thus, when the message is intercepted by router C2, it inspects the message and retrieves the spare parameters for its I5 interface. C2 then correctly installs the new reservations on I5 (in event (f)). It then deletes its QSPEC object from the message and forwards the message to the path in event (g). Likewise, C3 intercepts the message, reads its new reservation parameters, and properly installs the new reservations into its fg interface in event (h). It then deletes its QSPEC object and forwards the message in event (i). When BR2 receives the message, it composes the corresponding response message with a flag indicating the successful reservation and sends it to BR1 in event (j). Thus, upon receiving the signaling response message, BR1 notifies (in event (k)) CDP of the success of the reservation. Thus, CDP updates its local database according to event (b) (Table A) for link states and event (1) (Table B), where CDP updates the available bandwidth on bottlenecks. for each path sharing one or some of the affected connections (I ₂ , I5 and I5).

As a result, the available bandwidth, in CoS EF at path bottleneck link 3, has been upgraded from 0 bandwidth unit to 1 unit and CDP can accommodate the new request for 1 bandwidth unit.

This architecture and mechanism may be implemented by software, hardware, or a combination of software and hardware. Application of the modules and features described in the present invention may comprise a node or element (routers, for example) in a class-based QoS network that is properly programmed to perform bandwidth parameter configurations as described in the present invention. . In centralized network architectures (eg TISPAN [20], 3GPP [25], etc.), topological information is maintained by a central Decision Control Point (eg a central CDP as in the illustration), which controls locally. all the ways. For decentralized architectures, such as a network deployed with multiple distributed CDPs, proper decentralization control protocol is required for distributed CDPs to properly cooperate and synchronize with relevant network control information such as dynamic QoS capabilities in all correlated paths within the network

However, an efficient decentralization control synchronization mechanism is outside the scope of the present invention. In addition, the ACOR solution can be tailored to the QoS control model and local operator policies. However, this specific implementation issue, which depends on the operators' business models, is outside the scope of the present invention.

REFERENCES [1] R. Braden, L. Zhang, S. Berson, S. Herzog,

Resource ReSerVation Protocol (RSVP) - Version 1 Functional Specification, RFC 2205, Sept. 1997

[2] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang, and W. Weiss, An Architecture for Differentiated Services, IETF RFC 2475.

[3] E. Rosen, A. Viswanathan, and R. Callon, Multiprotocol Label Switching architecture, IETF RFC 3031.

[4] A. Demers, S. Keshav, S. Shenker; Analysis and simulation of a fair queue algorithm, ACM SIGCOMM'89, V.19, Pages 1-12.

[5] S.J. Golestani; A self-clocked fair queue scheme for broadband applications, INFOCOM'94, Networking for Global Communications, Vo. 2, Pages: 636-646, Canada.

[6] A. Neto, E. Cerqueira, A. Rissato, E. Monteiro, and P. Mendes, A Resource Reservation Protocol Supporting QoS-aware Multicast Trees for Next Generation Networks, In Proc. 12th IEEE Symp. on Computers and Communications, Aveiro, Portugal, Jul. 2007.

[7] P. Pan, E. Hahne, and H. Schulzrinne, BGRP: A Tree-Based Aggregation Protocol for Inter-domain Reservations, Trans. of Communications and Networks Journal, vol. 2, pp. 157-167, Jun. 2000.

[8] R. Bless, Dynamic Aggregation of Reservations for Internet Services, Proc. 10th Int. Telecommunication. Systems-Modeling and (ICTSM'10), vol.l, pp.26-38, Oct. 2002.

[9] P. Pinto, A. Santos, P. Amaral, and L. Bernardo,

SIDSP: Simple Inter-domain QoS Signaling Protocol, in

Conf. on Analysis

Proc. IEEE Military Communications Conference, Orlando, Florida, USA, Oct. 2007.

[Grandson A., E. Cerqueira, M. Curado, E. Monteiro and P. Mendes, Scalable Resource Provisioning for Multi-User

Communications in Telecommunications Next generation Conference, 2008. Networks, IEEE GLOBECOM Global 2008, (patent No. EP2 037 636 A1). [ii] D. Katz, IP Router Alert Option, RFC 2113, February 1997. [12] R. Hancock, G. Karagiannis, J. Loughney, S. Van den Bosch; Next Steps in Signaling (NSIS): Framework, RFC 4080, June 2005. [i3] Manner, J ., Karagiannis G., A. McDonald; NSLP

for Quality of Service Signaling, draft-ietf-nsis-qosnslp-16 (work in progess), February 2008.

[14] J. Moy, OSPF Version 2, RFC 2328, April 1998 [15] Yogen K., Dalal and Robert M. metcalfe; Reverse Path Forwarding of Broadcast Packets, Communications of the ACM, Volume 21, Issue 12 (Dec.1978) [16] www. Avvin. com / protocolSTP. html [17] J. —P. Vasseur, Z. Ali, S. Sivabalan; Definition of a Record Route Object (RRO) Node-Id Sub-Object, RFC 4561, June 2006 [18] Rute Sofia, SICAP, a Shared-segment Inter-domain Control Aggregation Protocol, PhD Thesis, Department of Informatics Faculty of Science University of Lisbon Campo Grande, Portugal [19] Draft ETSI TS 183 060 vO. 7.1 (2008-09) Technical Specification [20] ETSI ES 282 001: Telecommunications and Internet Converged Services and Protocols for Advanced Networking (TISPAN); NGN Functional Architecture Release 2.

[21] ETSI ES 282 003: Telecommunications and Internet Converged Services and Protocols for Advanced Networking (TISPAN); NGN Functional Architecture; Resource and Admission Control Sub-system (RACS).

[22] ETSI TS 183 017 Telecommunications and Internet Converged Services and Protocols for Advanced Networking (TISPAN); Resource and Admission Control: DIAMETER protocol for session based policy set-up Information Exchange between Application Function (AF) and Service Policy Decision Function (SPDF); Protocol specification.

[23] G. Ash, A. et al. , Y.1541-QOSM - Model for Networks Using Y.1541 QoS Classes draft-ietf-nsis-yl541qosm-10, February 4, 2010 [24] ETSI TS 123 107 Universal Mobile Telecommunications System (UMTS); Quality of Service (QoS) concept and architecture [25] The 3rd Generation Partnership Project -3GPP (www.3gpp.org/) [26] Lili Wang et al., Passive Inference of Path Correlation, International Network and Operating System Support Workshop for Digital Audio and Video, 2004

Claims (4)

1. A method of advanced Class of Service-based overbooking, CoS, to dynamically configure bandwidth reservation parameters, taking as base conditions the initial bandwidth division across the various classes of service through BL = 6 ^ (2, where g is the weight of each k CoSi at interface ⁼ 1. and C is the total bandwidth i = 1, whereas over the course of accepting and terminating calls, the bandwidth update will not is used through Δ _τ = ^ (Β _; -u _; ), where u ± is ai = l bandwidth used in each CoSi, and where the Network Decision Control Point automatically updates changes to relevant information. such as capacity, identification, bandwidth used, reserved, and available on each CoS for each interface within the network, as well as the bandwidth available on each CoS on the bottlenecks of all existing paths to maintain a real-time understanding of the topology and resource states of existing interfaces, and when receiving a bandwidth request for a CoS in a given path, if the available bandwidth in CoS in the interface bottleneck of the path , indicated by the local database, is insufficient to accept the request, the engine sets new reservation parameters for each CoS on each interface that belongs to the path, based on the features of each interface, considers In addition, if the request rj for CoSj is less than or equal to the weighted proportion of unused bandwidth on the rjdaj.A _T interface, CDP obtains the new B ± buffer parameters for each CoS on the interface: BL = U; + ΰρΔ _τ characterized by the following intake control: - if the request rj for CoSj is greater than the weighted proportion of unused bandwidth at the interface rj> aj.A _T , CDP obtains the new reserve parameters Bj for the affected CoSj, and the new reserve parameters for each one. of CoS, ifj remaining: ·, · ^{= Ω} / ⁺ ϋ ⁺ Μ ^Δ Γ-ύ ·) A, = ω, + -q), 1 <ϊ <Α, ij where Bj is the new bandwidth reserve for the affected CoSj, B ± is the new bandwidth reserve for each of the remaining CoSj, Uj is the current bandwidth used in CoSj receiving the request, rj is the bandwidth request for CoSj, cg is the weight of CoSj affected, and this update is made so that an authorized request can always be accepted provided that the total unused bandwidth on the path bottleneck interface is sufficient for the request. A resource control method according to claim 1, characterized by: updating the bandwidth used U (i, I _e ) and available A (i, I _e ) on each CoS, 1 didk for each interface I _and along a path P _x subject to the admission of a service request r ₂ (U (i, I _e ) = U (i, I _e ) + r ± e A (i, I _e ) = Bi ~ U ( i, I _e )), readjusting to more or less bandwidth (U (i, I _e ) = U (i, I _e ) ± r ± e A (i, I _e ) = Bi-U (ί, _{Θ θ} )) or active session removal (U (i, I _e ) = U (i, I _e ) -qe A (i, I _e ) = Β ± -υ (i, I _e )), thus , the available bandwidth on each CoS ₄ on the bottleneck interfaces of all paths traversing this interface is also updated, in real time, to maintain the consistency of the information stored in the local Decision Control Point database. traffic conditioning is installed on nodes at the network entrance to control load, or traffic policing, packets belonging to a mapped stream in a path should follow this path. A resource control method according to claims 1 and 2, characterized by the standby parameters to be installed for each CoS on an interface, which are defined based on the resource state of that interface and not based on the state. of the bottleneck interface resources of a path it is in, so the reservations defined by a Network Decision Control Point for different interfaces in the path are encapsulated in separate QSPEC objects and associated with the interface IDs to be transmitted to nodes on the path by signaling, where each node checks the parameters intended for it and performs configuration or readjustments of reservations. Class-based advanced class-based bandwidth override control system for implementing the method according to claims 1 to 3, characterized in that it includes nodes or elements, for example routers, in a class-based network which are properly programmed to perform bandwidth reservation parameter configurations as described in the method referred to in claims 1 to 3, and the method described in addition to being applied to bandwidth as a network resource may also be applied to other network resources such as memory or node processors or elements.