IP QOS and Traffic Management 2001 Copyright SCUT DT&P Labs

3. Differentiated Services Contents 1. What is QoS & Why QoS? 2. Integrated Services 3. Differentiated Services 2001 Copyright SCUT DT&P Labs

1. What is QoS & Why QoS？ 2001 Copyright SCUT DT&P Labs

1. What is QoS & Why QoS？ 1.1 Circuit Switching Allocate each stream a channel with the fixed bit rate Traffic characteristic: Constant Bit Rate 2001 Copyright SCUT DT&P Labs

1.2 Packet Switching (store-and-forward) 1. What is QoS & Why QoS？ 1.2 Packet Switching (store-and-forward) All the user share a common channel Traffic characteristic: Variable & Burst Bit Rate 2001 Copyright SCUT DT&P Labs

1.3 Traditional Internet Service 1. What is QoS & Why QoS？ 1.3 Traditional Internet Service Traditional Internet Applications and their features Traditional Applications such as: FTP (File Transfer Protocol) Email Telnet WWW ……… These kinds of services cannot tolerate packet loss, but are less sensitive to delay and delay variation. 2001 Copyright SCUT DT&P Labs

1. What is QoS & Why QoS？ 1.4 Basic Concept of QoS A QoS is a set of parameters that describer the Quality of Service , for example, bandwidth, loss rate, delay and delay variation. …… Different QoSs may mean different network resources allocation. The resources may includes link bandwidth, buffer usage, CPU usage, priority, etc. 2001 Copyright SCUT DT&P Labs

OSI 参考模型中的QoS参数 吞吐量 传输延时 出错率 连接失败的概率 传输失败的概率 重置率（在给定的时间内服务者释放连接或重置连接的概率） 释放延迟（从释放请求开始到释放确认为止的时间延迟） 释放失败概率 （其它：访问权限、优先级、成本） 2001 Copyright SCUT DT&P Labs

ITU (International Telecommunications Union )（CCITT） QOS参数 传输级参数： 分组的峰值到达率、峰值持续时间、分组平均到达率、分组丢失率、分组插入率和比特出错率 呼叫级控制级参数： 呼叫次数、失败率 连接级参数: 连接延迟、连接失败率；释放延迟、释放失败率 2001 Copyright SCUT DT&P Labs

ATM 网络QoS定义 服务类别 恒定位速率服务（CBR） 实时可变位速率服务（rt－VBR） 非实时可变位速率服务（nrt－VBR） 可用位速率服务（ABR） 未指定位速率服务（UBR） 2001 Copyright SCUT DT&P Labs

ATM 网络QoS定义 QoS参数 峰值信元速率（PCR） 持续信元速率（SCR） 最小信元速率（MCR） 信元丢失率 （CLR） 信元传输延时（CTD） 信元传输时延方差（CDV） 突发容许长度（BT） 2001 Copyright SCUT DT&P Labs

IETF QoS模型及定义 综合服务模型（Integrate Service：IntServ） 可控负载型服务 保证型服务 可控负载型服务 区分服务模型（Differentiated Service: DiffServ） 奖赏服务 确保服务 …… 2001 Copyright SCUT DT&P Labs

用户角度 QoS 描述 信息流特征（用户－网络间的承诺） 信息流性能要求（网络侧的承诺） 网络的吞吐量、传输的延迟、抖动和丢失率等。 信息流的峰值速率和平均速率，突发长度等。 信息流性能要求（网络侧的承诺） 网络的吞吐量、传输的延迟、抖动和丢失率等。 信息流的同步要求 多个相关信息流之间的同步关系，如会议系统同时记录的视频信号的同步播放；图像信息与声音信息的同步等。 2001 Copyright SCUT DT&P Labs

用户角度 QoS 描述（续） 服务层次（端到端QoS的保证程度） 如：可控负载型服务、保证型服务、 区分服务和尽力而为型服务等。 用户－网络间接入协商机制； 带宽、延时、抖动和丢失率等状况的通知等。 服务成本（避免网络资源的滥用） 不同类别与层次的服务与服务成本的关系。 2001 Copyright SCUT DT&P Labs

1.5 Basic IP Protocol Service Function 1. What is QoS & Why QoS？ 1.5 Basic IP Protocol Service Function The Basic IP Protocol stack provides only one kind of QoS: best-effort. With the best effort traffic model, Internet (router) handles the transmission service with first-come, first-served strategy. All requests have the same priority and are dealt with one after the other. No possibility of making bandwidth reservations for specific connections or raising the priority for special requests. 2001 Copyright SCUT DT&P Labs

1.6 Real-time Traffic in Internet 1. What is QoS & Why QoS？ 1.6 Real-time Traffic in Internet New Internet Applications (multimedia) and their features Multimedia Applications such as: Real-Audio (Audio Broadcast) Real-Video (Video Broadcast) Internet phone Videoconferencing These kinds of real-time applications show different behavior with traditional service. Can compensate for a reasonable amount of packet loss Be usually very critical toward high-variable delays. 2001 Copyright SCUT DT&P Labs

1.7 The Requirement for the New Service 1. What is QoS & Why QoS？ 1.7 The Requirement for the New Service Without bandwidth control, the quality of real-time steams depends on the bandwidth that is just available. Unstable bandwidth leads to poor quality real-time transmissions, such as dropout and hangs. As a consequence of these limitations, new strategies are required to provided predictable services for the Internet. 2001 Copyright SCUT DT&P Labs

1.8 Two Classes of QoS Services 1. What is QoS & Why QoS？ 1.8 Two Classes of QoS Services Now, there are two main rudiments for bringing QoS to the Internet: * Integrated Services; * Differentiated Services. Integrated Services bring enhancements to the IP network to support real-time transmissions and guarantee bandwidth for specific flows. Differentiated Service mechanisms do not use per-flow signaling. Different service levels can be allocated to different groups of Internet users. 2001 Copyright SCUT DT&P Labs

1.9 QoS 设计的原则 集成原则 置、预测和维护的（理想情况）。 分离原则 区分控制信令和媒体数据的传输－－通常两者的传输 服务要求不同。 透明原则 基于友好的QoS的API函数为用户提供服务，屏蔽网络 底层的复杂服务细节。 2001 Copyright SCUT DT&P Labs

1.9 QoS 设计的原则（续） 异步资源管理原则 资源管理：调度、流控、路由和QoS管理－－在各种 资源之间定期（周期性）地交换信息。 异步：管理和控制的发生在时间上可能是不同的。 性能原则 既要保证QoS的实现，又要使网络的资源有高的利用 率。 资源分配的效率； 控制信息传输的代价； …… 。 2001 Copyright SCUT DT&P Labs

1.10 QoS 的控制与管理 资源的管理的实现过程 静态阶段：处理信息流的建立和端到端的QoS的协商, 动态阶段：在接入控制之后，通过特定算法实现资源 调度和流控,实现用户所需QoS的过程。 2001 Copyright SCUT DT&P Labs

1.10 QoS 的控制与管理（续） 预留和分配机制 悲观方式：根据应用可能出现的最大需求来预留资源 例如，按应用的峰值速率来预留带宽。 优点：可靠性最高； 缺点：网络的资源利用率可能收到影响。 乐观方式：根据平均负载预留资源。例如根据平均速 率预留带宽。 优点：资源利用率较高； 缺点：难有高的服务质量。 实际方式：介于上述两种方式之间。 2001 Copyright SCUT DT&P Labs

1.10 QoS 的控制与管理（续） QoS 的控制机制 信息流整型：在网络边缘进行信息流整型，改变进入 网络的业务流量的特性。 采用合理的、隔离度好的调度策略。 信息流监控：监控用户是否按照流量的特性要求发送 信息；监控网络是否能够提供承诺的服 务。 对违约信息分组的处理方式。 2001 Copyright SCUT DT&P Labs

1.10 QoS 的控制与管理（续） QoS 的控制机制 信息流同步：控制事件顺序和多媒体交互行为的精确 同步。如所谓的“嘴唇-声音”同步。 2001 Copyright SCUT DT&P Labs

2. Integrated Services （综合服务） 2001 Copyright SCUT DT&P Labs

2.1 Integrated Service and RSVP (Reservation Protocol) 2. Integrated Services 2.1 Integrated Service and RSVP (Reservation Protocol) The Integrated Service model was defined by an Internet Engineering Task Force (IETF) working group as being the keystone of the planned Integrated Service Internet. The Integrated Service are described by the ReSerVation Protocol (RSVP). To support the Integrated Service, an Internet Router must be able to provide an appropriate QoS for each flow in accordance with RSVP. 2001 Copyright SCUT DT&P Labs

IntServ QoS 概述 两种服务方式 质量保证型（guaranteed service） 负载控制型（controlled－load service） 同时保留“尽力而为”型的服务方式 实现方法 在路由器的控制信息路径上处理每个流的信息消息， 维护每个流的路径和资源预留状态。 在路由器的数据路径上执行基于流的分类、调度和缓 冲区管理。 2001 Copyright SCUT DT&P Labs

IntServ QoS 概述 技术方案 依据IETF的资源预留协议（RSVP）的QoS协商机制， 逐点地建立或拆除每个数据流的路径状态和资源预留 状态。 依靠接纳控制来决定各节点是否有足够的资源满足用 户的特定请求。 依靠传输路径上各节点的分类、调度算法和监控机制 来保证用户的QoS。 2001 Copyright SCUT DT&P Labs

2.2 The Integrated Services (RSVP) Model Host Router RSVP Process RSVP Application RSVP Process Routing Process Policy Control Policy Control Admission Control Admission Control Classifier Packet Scheduler Packet Scheduler Data Classifier 2001 Copyright SCUT DT&P Labs

2.3 The Components for the RSVP Traffic Control 2. Integrated Services 2.3 The Components for the RSVP Traffic Control Packet scheduler: It manages the forwarding of the different packet streams in hosts and routers, based on their service classes, using queue management and various scheduling algorithms. It must ensure that the packet delivery corresponds to the QoS parameters for each flow. It can also police or shape the traffic to conform to a certain level of service. 2001 Copyright SCUT DT&P Labs

2. Integrated Services 2.3 The Components for the RSVP Traffic Control (continued) Packet classifier: It identifies the packets of an IP flow in hosts and routers that will receive a certain level of service. To realize effective traffic control, each incoming packet is mapped by the classifier into a specific class. All packets that are classified in the same class get the same treatment from the packet scheduler. The choice of a class is based upon the source and destination IP address and source and destination port number in the existing packet header or an additional classification number（such as IP protocol number）that must be added to each packet. 2001 Copyright SCUT DT&P Labs

TCP/IP Protocol Stack Telnet, FTP, TFTP,HTTP,SNMP,SMTP, and so on Application Layer Telnet, FTP, TFTP,HTTP,SNMP,SMTP, and so on Port Number BGP RIP Transport Layer OSPF EGP TCP UDP ICMP IGMP Protocol Number Internet Layer ARP IP (address) RARP Type code Data Link Layer Ethernet, Token Ring, PPP, and so on 2001 Copyright SCUT DT&P Labs

2.4 The Components for the RSVP Traffic Control (continued) 2. Integrated Services 2.4 The Components for the RSVP Traffic Control (continued) Admission Control: The admission control contains the decision algorithm that a router uses to determine if there are enough routing resources to accept the requested QoS for a new flow. Admission control is invoked at each router along a reservation path so as to make a local accept/reject decision at the time a host requests a real-time service. 2001 Copyright SCUT DT&P Labs

2.4 The Components for the RSVP Traffic Control (continued) 2. Integrated Services 2.4 The Components for the RSVP Traffic Control (continued) Policy Control: The policy control ensures that a host does not violate its promised traffic characteristics. Some control policies are used to check the user authentication for a requested reservation. Unauthorized reservation requests can be rejected. The police control makes sure that the QoS guarantees are honored. Admission control is concerned with enforcing administrative policies on resource reservations. 2001 Copyright SCUT DT&P Labs

2.5 The Classifiers and the Scheduler 2. Integrated Services 2.5 The Classifiers and the Scheduler Queue for Session 1 Queue for Session 2 Queue for Session 3 Classifier Queue for Session 4 Packets Queue for Session 5 Packets Queue for Session 6 Queue For Best Effort T Scheduler 2001 Copyright SCUT DT&P Labs

2.6 Weighted Fair Queuing Scheduling （WFQ）Algorithm 2. Integrated Services 2.6 Weighted Fair Queuing Scheduling （WFQ）Algorithm WFQ is also named as General Processor Sharing (GPS). WFQ is a service discipline designed to allocate capacity (bandwidth) equally and fairly among multiple connections sharing a link. Suppose the capacity on a given link is C b/s; N connections share the link; Connection j, 1 <= j <= N, is assigned a parameter фj such that it is guaranteed service at a rate gj given by : 2001 Copyright SCUT DT&P Labs

2.6 The Weighted Fair Queuing Scheduling Algorithm 2. Integrated Services 2.6 The Weighted Fair Queuing Scheduling Algorithm gj is the minimum guaranteed service rate for connection j; WFQ is defined to be work conserving: if one connection has traffic ready to be served, while another has none, capacity from the latter will be used to serve the former. The actual connection j traffic, served in an interval (t1, t2), denoted by Sj(t2,t1), will be obey the condition: 2001 Copyright SCUT DT&P Labs

2.6 The Weighted Fair Queuing Scheduling Algorithm 2. Integrated Services 2.6 The Weighted Fair Queuing Scheduling Algorithm For the system being steady, it is required that: The maximum delay of traffic on a given connection can be bounded based on its own characteristics, independent of other sessions. Consider the special case in which each connection on a link regulated in the leaky bucket flow-constrained: where Ai (t2,t1) is the quantity of the data transmitted in the connection i during t1~t2. And if 2001 Copyright SCUT DT&P Labs

2.6 The Weighted Fair Queuing Scheduling Algorithm 2. Integrated Services 2.6 The Weighted Fair Queuing Scheduling Algorithm then the worst-case backlog (data buffered) and delay bounds are established: and , respectively. Let Lmax be the maximum packet length, the end to end delay, over an m-hop path, neglecting the propagation delay: Ck is the total link capacity on link k. 2001 Copyright SCUT DT&P Labs

2. Integrated Services 2.7 Message & Data Stream The Data Stream and Message Stream (in the same data link) of The Integrated Service (IS): RSVP control message stream: The IS instances communicate via RSVP to create and maintain flow-specific states in the endpoint hosts and routers along the path of a flow. Data Stream: the data packets in a reserved flow were sent as usually, the flow is recognized with the the source and destination IP address and port number in the existing packet header. 2001 Copyright SCUT DT&P Labs

2.8 Service Classes in the IS Model 2. Integrated Services 2.8 Service Classes in the IS Model Two Service Classes in the IS Model: Controlled Load Service (defined in RFC 2211). Guaranteed Service (defined in RFC 2212). 2001 Copyright SCUT DT&P Labs

2.8 Service Classes in the IS Model (continued) 2. Integrated Services 2.8 Service Classes in the IS Model (continued) Controlled Load Service: Applications that make QoS reservation using Controlled Load Service are provided with service closely equivalent to the service uncontrolled (best-effort) traffic under lightly loaded conditions. The Controlled Load Service does not provide functions for reserving a fixed bandwidth or guaranteeing minimum packet delay. The term lightly load conditions means that a very high percentage of transmitted packets are successfully delivered to the destination. 2001 Copyright SCUT DT&P Labs

可控负载型服务的特点 使用户感到网络是在一种很轻的负载或具有很大容量 的条件下运行。 本质上是一种半定量半定性的服务： 以很高的概率（百分比）将数据分组转发到接收端； 绝大部分转发成功的分组的延时在一个可接受的范围 内； 2001 Copyright SCUT DT&P Labs

2.8 Service Classes in the IS Model (continued) 2. Integrated Services 2.8 Service Classes in the IS Model (continued) Guaranteed Service: Functions provided by the Guaranteed Service model assure that datagrams arrive within a guaranteed delivered time. The Guaranteed Service model represents the extreme end of delay control for networks. This control mechanism is only useful if it is provided by every router along the reservation path. 2001 Copyright SCUT DT&P Labs

质量保证型服务的特点 模拟某种特定速率所提供的服务，提供定量的带宽和 延时保证。 要求传输路径上的每个节点都支持质量保证型的服务。 2001 Copyright SCUT DT&P Labs

2.9 The Flow Descriptor Flow Descriptor: 2. Integrated Services 2.9 The Flow Descriptor Flow Descriptor: The Flow Descriptor defines the traffic and QoS characteristics for a specific flow of data packet. The information contained in the Flow Descriptor consists of Filter Specification (Filterspec), which is used in the packet classifier to identify the packets that belong to a specific flow with the send IP address and source port. Flow Specification (Flowspec), which can be assorted into two groups Traffic Specification (Tspec), Service Request Specification (Rspec). 2001 Copyright SCUT DT&P Labs

2.9 The Flow Descriptor (continued) 2. Integrated Services 2.9 The Flow Descriptor (continued) Filter Specification Flow Descriptor Traffic Specification Flow Specification Request Specification 2001 Copyright SCUT DT&P Labs

2.9 The Flow Descriptor (continued) 2. Integrated Services 2.9 The Flow Descriptor (continued) Traffic Specification (Tspec) The Tspec describes the traffic characteristic of the requested service. That is represented with a token bucket filter to define a data flow control mechanism. 2001 Copyright SCUT DT&P Labs

2.9 The Flow Descriptor (continued) 2. Integrated Services 2.9 The Flow Descriptor (continued) The token bucket system is specified by two parameters r b Token rate r, which represents the rate at which tokens are placed into the bucket. Bucket capacity b, which is the size of the bucket. 2001 Copyright SCUT DT&P Labs

2.9 The Flow Descriptor (continued) 2. Integrated Services 2.9 The Flow Descriptor (continued) The parameters of the token bucket system The parameter r specifies the long-term data rate and is measured in bytes of IP datagrams per second (from 1 byte per second to 40 terabytes per second). The parameter b specifies the burst data rate allowed by the system and is also measured in bytes (from 1 byte to 250 gigabytes). Traffics that passes the token bucket filter must obey the rule that over all time periods T, the amount of data sent does not exceed rT + b. Other token bucket parameter: The minimum policed unit m: size of the min. IP datagram. The maximum packet size M: size of the max. IP datagram 2001 Copyright SCUT DT&P Labs

2.9 The Flow Descriptor (continued) 2. Integrated Services 2.9 The Flow Descriptor (continued) Service Request Specification (Rspec) The Rspec specifies the QoS that the application wants to request for a specific flow. It may consist of a specific bandwidth, a maximum packet delay and/or maximum packet loss rate，etc. 2001 Copyright SCUT DT&P Labs

2.10 The Basic Concept of RSVP 2. Integrated Services 2.10 The Basic Concept of RSVP RSVP is defined in RFC 2205. RSVP is only a control protocol, it required an existing routing protocol to operation. RSVP runs on top of IP and UDP and must be implemented in all routers on the reservation path. RSVP provides the same service for unicast and multicast flows. RSVP does not understand the contents of a flow describer. It is carried as an opaque object by RSVP and it is delivered to a router’s traffic control functions (packet classifier and scheduler, etc. ) for processing. 2001 Copyright SCUT DT&P Labs

2.10 The Basic Concept of RSVP (continued) 2. Integrated Services 2.10 The Basic Concept of RSVP (continued) The reservations of RSVP are only done in one direction. For duplex connections, it is necessary to set up two RSVP sessions for each station. RSVP is receiver-initiated. Using RSVP signaling messages, the sender provides a specific QoS parameter (related to a specific application) to the receiver. The receiver sends an RSVP reservation message, depending on its system capabilities, back with the QoS that should be reserved for the flow from the sender (via routers) to the receiver. 2001 Copyright SCUT DT&P Labs

A basic part of a resource reservation is the path. 2. Integrated Services 2.11 RSVP Operation A basic part of a resource reservation is the path. The RSVP operation includes three phases Establish the path; Maintain the path; Release the path. All packets that belong to a specific flow use the same path. 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) 1. Establish the resource reservation path For establish the path, the sender hosts periodically send Path message for each data flow they originate. Router2 Path mes. Sender Router1 Receiver Path mes. Path mes. Path mes. Router3 Router5 Router4 The Path message contains traffic information that describes the QoS for a specific flow. 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) The router operation when it receives the Path message (a). It stores the IP address from the last hop field in the message, which is the address of the sender. (b). Then the router inserts its own IP address into the last hop field, sends the path message to the next router. (c). The process repeats itself until the message has reached the receiver. (d). At the end of this process, each router will know the address from the previous and the path can be accessed backward. 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) The router operation when it receives the Path message (e). Routers that have received a path message are prepared to process resource reservation. 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) When a receiver receives the Path message and wants to reserve QoS for this flow, it sends a request, the reservation message (Resv. message), which contains the information of flow describer. The RSVP process in a router passes the request to the admission control and policy control for the authorization checking. Router2 Resv. mes. Sender Receiver Resv. mes. Resv. mes. Resv. mes. Router3 Router1 Router5 Router4 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) After a successful admission and policy check, a reservation request is propagated upstream toward the sender. In a multicast environment, reservations from different downstream branches but for the same sender are merged together as they across the upstream path. The merging is necessary to conserve resources in the routers. Router2 Receiver1 Router1 Sender Router3 Receiver2 Path Message Packet flow Resv. Message 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) 2. Maintaining the resource reservation path RSVP resource reservations maintain a soft state in routers and hosts. Soft state means that a reservation is canceled if RSVP does not send refresh (Path & Resv.) messages along the path for an existing reservation. 3. Releasing the resource reservation path PathTear messages ResvTear messages 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) PathTear messages: They travel downstream from the point of initiation to all receivers, deleting the path state as well as all reservation states in each RSVP-capable device. ResvTear messages: They travel upstream from the point of initiation to all senders, deleting reservation states in all routers and hosts. 2001 Copyright SCUT DT&P Labs

2.11 RSVP Operation (continued) 2. Integrated Services 2.11 RSVP Operation (continued) Other RSVP control messages Confirmation messages (ResvConf)：They are used to tell the receiver that the reservation has been established successfully (optional). Reservation Error Messages (ResvErr): If one of the admission and policy checks fails, the reservation is rejected and the RSVP process returns a ResvErr error message to the appropriate receiver. Path Error Messages (PathErr) 2001 Copyright SCUT DT&P Labs

2.12 RSVP Reservation Styles 2. Integrated Services 2.12 RSVP Reservation Styles Background: Users of multicast multimedia application often receive flows from different senders, such as videoconferencing, telephone-conferencing. It is necessary to do the resource reservation for these multi-flows. RSVP provides a more flexible way (comparing with one flow one reservation control signaling) to reserve QoS for flows from different senders. These reservation request options are called reservation style. Reservation style Explicit; Wide-card. 2001 Copyright SCUT DT&P Labs

2.12 RSVP Reservation Styles (continued) 2. Integrated Services 2.12 RSVP Reservation Styles (continued) Explicit styles: Fixed-Filter (FF：固定) Style: A distinct reservations is created for data packets from a particular sender. Shared-Explicit (SE：显式共享) Style: A single reservation covers flow from a specified subset of senders (A sender list must be included in the receiver’s reservation request). In an explicit sender-selection reservation, a filterspec must identify exactly one sender. 2001 Copyright SCUT DT&P Labs

2.12 RSVP Reservation Styles (continued) 2. Integrated Services 2.12 RSVP Reservation Styles (continued) Wide-card-Filter (WF：通配符) Style: Reservations from different senders are merged together along the path so that only the biggest reservation request reaches the senders. This reservation style establishes a single reservation for all senders in a session. In a Wind-card sender selection the filterspec is not needed. 2001 Copyright SCUT DT&P Labs

2. Integrated Services 2.13 RSVP Messages Format An RSVP message consists of a common header followed by a body consisting of a variable number of objects. The message objects contain the information that is necessary to realize resource reservation (e.g., the flow descriptor or reservation style). RSVP Common Header Version Flags Message Type RSVP Checksum Send_TTL (Reserved) RSVP Length 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) RSVP Common Header Version: 4-bit RSVP protocol number, current version: 1. Flags: the field is reserved for flags. No flags are defined yet. Message Type: (a). Path (b). Resv (c). PathErr (d). ResvErr (e). PathTear (f). ResvTear (g). ResvConf 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) RSVP Common Header RSVP Checksum: it can be used by receivers of an RSVP message to detect errors in the transmission of this message. Send_TTL: it contains the IP TTL value the message was sent with. RSVP Length: it contains the total length (counted in bytes) of the RSVP message, including the common header and all objects that follow. 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) RSVP Objects RSVP Object header. Length (Bytes) Class-Number C-Type (Object Contents) Length: the object length. Class-Number: Identifies the object class. The following classes are defined. NULL: the content is ignored by the receiver. Session (required): it contains the IP destination address, the IP protocol ID, and the destination port to define a specific session for the other objects that follow. 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) RSVP Object Header Class-Number (continued) RSVP_HOP: this object contains the IP address of the node that sent this message. For downstream messages, the RSVP_HOP object represents previous hop object; For upstream messages, it represents an next hop object. Time_Values: it contains the refresh period for path and reservation messages. Style: it defines the reservation style and some style specific information that is not in flowspec or filterspec. Flowspec: this object specifies the required QoS in reservation messages Filterspec: it defines which data packets receive the QoS specified in the flowspec. 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) RSVP Object header Class-Number (continued) Sender_Template: the object contains the sender IP address and additional demultiplexing information used to identify a sender. Sender_Tspec: it defines the traffic characteristics of a data flow from a sender. Adspec: this object is used to provided advertising information to the traffic control modules in the RSVP nodes along the path. Policy_Data: it contains the information that allows a policy modules to decide whether an associated reservation is administratively permited or not. Resv_Confirm: this object contains the IP address of a receiver that requests confirmation for its reservation. 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) RSVP Object Header Class-Number (continued) Integrity: it contains cryptographic data to authenticate the originating node and to verify the contents of an RSVP message. Scope: the object contains an explicit list of sender hosts to which the information in the message is sent. C_Type: it specifies the object type within the class number. Different object type are used for IPv4 and IPv6. Object contents depend on the object type and have a maximum length of 65,528 bytes 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) Example 1: an RSVP path message format Common Header (Integrity) Session RSVP_HOP Time_Value (Policy_Data) Sender_Template Sender_Tspec (Adspec) 2001 Copyright SCUT DT&P Labs

2.13 RSVP Messages Format (continued) 2. Integrated Services 2.13 RSVP Messages Format (continued) Example 2: an RSVP Resv message format Common Header (Integrity) Session RSVP_HOP Time_Values (Resv_Confirm) (Scope) (Policy_Data) Sender_Template Sender_Tspec (Adspec) 2001 Copyright SCUT DT&P Labs

2.14 An Experiment of the RSVP Application 2. Integrated Services 2.14 An Experiment of the RSVP Application Link Bandwidth: 4 Mbps Host Host Host Host Host Host Network 1 Network 2 Link Router Router RSVP Router RSVP Router Host Host Host Video Clients Video Server 2001 Copyright SCUT DT&P Labs

2.15 IntServ 在整个Internet上实现的局限性 RSVP 信令协议的实现复杂，系统扩展性不好； 2. Integrated Services 2.15 IntServ 在整个Internet上实现的局限性 RSVP 信令协议的实现复杂，系统扩展性不好； 一般基于单个流的分类、监管和调度控制系统开销过大； 有关IntServ的QoS的研究仍然是一个开放的问题： 网络系统状态和可用资源变化的预测； 高效高性能的调度策略； QoS 路由选择 （单播、组播） 2001 Copyright SCUT DT&P Labs

3. Differentiated Services （区分服务） 2001 Copyright SCUT DT&P Labs

区分服务 DiffServ 主要针对综合服务 IntServ 实现上面临的问题而提出； DiffServ 的目标在于使QoS的功能实现简单有效； DiffServ 的主要特点： 简化网络内部的服务机制，内部节点只作调度转发，流状态的信息保存与监控等只在边界节点上进行； 采用聚集传输控制，内部节点的服务对象是流聚集，而非单个的流，单个流的信息在边界保存和处理。 2001 Copyright SCUT DT&P Labs

网络内部对聚集流提供特定质量的调度转发服务。 区分服务 DiffServ 服务体系 （1）两个层次结构： 网络边界作单流分类聚合； 网络内部对聚集流提供特定质量的调度转发服务。 区分服务 DiffServ 服务体系 （1）两个层次结构： DS区域（DS domain）：在每个DS区域内具有相同的 服务提供策略； DS区（DS region）：在DS区内的不同DS区域可以支 持不同的服务提供策略； DS区内不同的服务提供策略之间通过服务层协议 （SLA）和传输调节协议（TCA）进行协调。 2001 Copyright SCUT DT&P Labs

服务层协议（SLA：Service Layer Agreement） 3.0 区分服务 DiffServ 概述（续） 服务层协议（SLA：Service Layer Agreement） 传输调节协议（TCA：Traffic Conditioning Agreement） （2）总体集中控制策略 确定在边界如何分类聚合流，在内部如何调度转 发聚集流。 （3）路由 区分服务在节点处只是提供队列调度和缓冲管理， 不涉及路由选择机制（与综合服务同）。 2001 Copyright SCUT DT&P Labs

3.1 Basic Concept of Differentiated Services The Differentiated Services (DS) concept is currently under development at the IETF DS working group. The goal of the DS development is to make it possible to provide differentiated classes of service for Internet traffic and to support various types of application and specific business requirement. DS offers predictable performance (delay, throughput, packet loss, etc.) for a given load at a given time. 2001 Copyright SCUT DT&P Labs

3.1 Basic Concept of Differentiated Services (continued) DS does not need for per-flow state and signaling at every hop. In DS, the QoS guarantees are static and long-term in routers. With DS, the Internet traffic is split into different classes with different QoS requirement. Service Layer Agreement (SLA): it is a service contract between a customer and a service provided. 2001 Copyright SCUT DT&P Labs

3.1 Basic Concept of Differentiated Services (continued) DS byte: It uses the space of the TOS (Type Of Service) octet in the IPv4 IP header and the traffic class octet in the IPv6 header. It is used to mark the packet that receive a particular forwarding treatment at each network node. 0 1 2 3 4 5 6 7 DSCP CU Differentiated Services Code-Point (DSCP): the field is used to select the traffic class that a packet experiences at each node. Currently Unused (CU): this field is reserved and can be assigned later. 2001 Copyright SCUT DT&P Labs

3.1 Basic Concept of Differentiated Services (continued) The DS byte is set at network edges and administrative boundaries. The DS byte is interpreted by the routers in the network. The currently defined DS architecture only provides service differentiation in one direction and is therefore asymmetric. 2001 Copyright SCUT DT&P Labs

3.2 Differentiated Services Architecture Per-Hop Behavior (PHB) A PHB is a description of the forwarding treatment a packet receives at a given network node. The DSCP value in the DS byte is used to select the PHB that a packet experiences at each node. The PHB can be described as a set of parameters inside of a router that can be used to control how packets are scheduled into an output interface (scheduler). The scheduler can be a number of separate queues with settable priorities, parameters for queue lengths or drop algorithms, and drop preference weights for packets. 2001 Copyright SCUT DT&P Labs

逐点行为 （Per-Hop Behavior (PHB)） 逐点行为PHB是一个DS节点调度转发特定流聚集的外部特性描述，说明单个节点为特定流聚集分配资源的方式； PHB可以用调度转发流聚集时的流特性参数（如延时、丢失率）来描述； PHB针对具体的流聚集，流聚集用IP包头的DSCP标识； 多个PHB共存于一个节点时，不同的PHB有不同的相对优先等级。 2001 Copyright SCUT DT&P Labs

缺省型PHB－BE（Best Effort）：相当与传统的“尽力而为”调度转发行为，属于BE的IP包仅在带宽空闲未被其它流聚集使用时发送。 准尽力而为型PHB－LBE（Lower than BE）:优先度比BE更低，使在发生拥塞时能更有选择性地丢弃包，LBE可用于类似备份的相对不太重要的后台数据传输。 类选择型PHB－CS（Class Selector）：CS是为使DiffServ向后兼容原IPv4 TOS字节的前3位作为优先级队列调度的选择标志。 2001 Copyright SCUT DT&P Labs

逐点行为（PHB）的类型（续） 加速型PHB－EF（Expedited forwarding ）：保证在任何时候接受此服务的流离开速率大于等于设定速率，优先级最高，不受其它传输流的影响； 允许丢失的加速型PHB－EFD（Expedited forwarding with dropping）：相对EF型，EFD的外特性是低延时，有一定的丢失； 协同PHB组（Interoperability PHB Group）：将所有应用归结为若干级别的重要性（Importance）－I级和紧迫性（Urgency）－U级。以描述丢失率和延时。 2001 Copyright SCUT DT&P Labs

Queue 7 (Highest Priority) Queue 1 (Lowest Priority) 3. Differentiated Services 3.2 Differentiated Services Architecture (continued) The scheduler in a router Example 1: A router with 8 different queues that have all different priorities Queue 7 (Highest Priority) Queue 6 Queue 5 Classifier Queue 4 Queue 3 Queue 2 Queue 1 (Lowest Priority) 2001 Copyright SCUT DT&P Labs

3.2 Differentiated Services Architecture (continued) Example 2: A router that has a single queue with multiple drop priorities for data packets. Example 3: Another possible configuration is four queues with two levels of drop preference in each. DS requires router that support queue scheduling and management to prioritize outbound packets and control the queue depth so as to minimize congestion on the network. The packet treatment inside of a router depends on the router’s capabilities and its particular configuration. 2001 Copyright SCUT DT&P Labs

3.2 Differentiated Services Architecture (continued) PHBs will be defined in groups. A PHB group is a set of one or more PHBs that can only be specified and implemented simultaneously. A default PHB (the standard best-effort forwarding behavior) must be available in existing routers. The IETF working group recommends the use of the DSCP(Differentiated Services Code-Point ) value 000000 in the DS byte to define the default PHB. Another PHB that is proposed for standardization is the Expedited Forwarding (EF) PHB (DSCP: 101100 ), which is a high-priority behavior that is typically used for network control traffic. 2001 Copyright SCUT DT&P Labs

奖赏服务－PS（Premium Service） 为用户提供“三低一保证”的服务，低延迟、低抖动、低丢失率和保证端到端的传输带宽； DiffServ 的典型服务类型 奖赏服务－PS（Premium Service） 为用户提供“三低一保证”的服务，低延迟、低抖动、低丢失率和保证端到端的传输带宽； 服务承诺针对用户流的最高速率，费用最为昂贵； 在每个节点处采用EF－PHB提供服务资源的保证； 在网络的输入节点边界处必要时要通过流的整形保证在网络的后续节点处均有入速率小于出速率。 2001 Copyright SCUT DT&P Labs

确保服务－AS（Assured Service） DiffServ 的典型服务类型（续） 确保服务－AS（Assured Service） 无论是否拥塞，保证为用户提供预约时的最低限量的带宽，其特点是确保一定的带宽和丢失率，不涉及延迟和抖动； 采用AS服务的分组进入网络时在边界节点作标记，预约带宽以内的流量标为IN（In profile），超出预约带宽以外的流量标为OUT（Out profile）； 在拥塞时包头标记决定分组的丢弃概率，OUT的丢弃概率大于IN。 2001 Copyright SCUT DT&P Labs

IETF的标准化工作只是针对PHB，服务类型由各ISP自行确定； DiffServ 的PHB与服务类型 IETF的标准化工作只是针对PHB，服务类型由各ISP自行确定； 同一PHB与不同的边界分类调节机制结合，可以实现不同的服务； 当新的服务类型无法用已有的PHB实现时，就需要定义新的PHB： 针对无线网络中误码率高和可能出现移动交接中断的问题，已经提出两种新的服务种类： 移动奖赏服务和低延时尽力而为服务，针对后者提出了前面介绍过的PHB－EFD。 2001 Copyright SCUT DT&P Labs

Different Services Domains 3. Differentiated Services 3.2 Differentiated Services Architecture (continued) Different Services Domains DS domains is a contiguous portion of the Internet over which a consistent set of DS policies are administered in a coordinated fashion. Different Services Domains can represent different administrative domains or autonomous system; different trust regions, and different network technologies. A DS domain normally consists of one or more networks under the same administration, for example, a corporate intranet or an Internet service provider (ISP). The administration of the DS domain is responsible for ensuring that adequate resources are provisioned and reserved to support the SLA(Service Level Agreement )s offered by the domain. 2001 Copyright SCUT DT&P Labs

3. Differentiated Services 3.2 Differentiated Services Architecture (continued) DS Boundary Nodes A DS boundary Node is such a node that connects two or more DS domains, which can be a router, a host or a firewall. A DS boundary Node that handles traffic leaving a DS domain is called an egress node, which performs conditioning function (by a traffic conditioner in it) on traffic that is forwarded to a directly connected peering domain. A DS boundary node that handles traffic entering a DS domain is called an ingress node, which must make sure that the packets enter a domain receive the same QoS as in the domain the packets traveled through before. 2001 Copyright SCUT DT&P Labs

I I B I I B B I I I B 3. Differentiated Services 3.2 Differentiated Services Architecture (continued) The relationship between the DS Boundary Nodes and Domains I I B I I B B I I I Interior Node B Boundary Node 2001 Copyright SCUT DT&P Labs

3.3 DS Traffic Conditioner 3. Differentiated Services 3.3 DS Traffic Conditioner Meter Packets Packets Shaper/ Dropper Classifier Marker 2001 Copyright SCUT DT&P Labs

3.3 DS Traffic Conditioner 3. Differentiated Services 3.3 DS Traffic Conditioner Traffic Conditioner: It consists of the following components: Classifiers: which select packets based on their packet header and forward the packets that match the classifier rules for further processing. Two classes of classifier: Multi-Field classifiers: they can classify on the DS byte as well as on the other IP header fields (such as IP address and port number); Behavior Aggregate classifiers: they classify only on the bits in the DS byte. 2001 Copyright SCUT DT&P Labs

3.3 DS Traffic Conditioner 3. Differentiated Services 3.3 DS Traffic Conditioner Meter: it measure whether the forwarding of the packets that are selected by the classifier correspond to the traffic profile that describers the specific QoS. A meter passes state information to other conditioning functions so as to trigger a particular action for each packet. Marker: it sets the DS byte of the incoming IP packet to a particular bit pattern according to the SLA between the service provider and customer. Shapers/Droppers: They use different methods to bring the stream into compliance with a traffic profile. For instance, Shapers can delay some or all of the packets. Droppers can discard some or all of the packet. 2001 Copyright SCUT DT&P Labs

3.3 DS Traffic Conditioner 3. Differentiated Services 3.3 DS Traffic Conditioner The Functions of the DS Traffic Conditioner The traffic conditioner is mainly used in DS boundary components. Make sure that packets that transit the domain are correctly marked to select a PHB from one of the PHB(Per-Hop Behavior ) groups supported within the domain. Make the PHB value remarked if necessary when packets travel through two different DS domains. If a data travels through multiple domains, the DS byte can be remarked at every boundary component to guarantee the QoS that was contracted in the SLA (Service Level Agreement). The SLA contains the details of the Traffic Conditioning Agreement that defines how the metering, marking, discarding, and shaping of packet must be done to fulfill the SLA. 2001 Copyright SCUT DT&P Labs

3.3 DS Traffic Conditioner 3. Differentiated Services 3.3 DS Traffic Conditioner Example: Remarking of Data Packets Interior Router Boundary Router Interior Router 7 7 7 3 3 3 Remarking 3 7 2001 Copyright SCUT DT&P Labs

3.4 DS Interior Components 3. Differentiated Services 3.4 DS Interior Components The interior components of a DS domain is usually a router that contains a traffic prioritization algorithm. All interior routers must use the same traffic forwarding policies to comply with the QoS agreement. The traffic conditioning inside of an interior node is done only by a packet classifier. Interior Router Packets Packets DS Byte Classifier Queue Management/ Scheduler 2001 Copyright SCUT DT&P Labs

3. Differentiated Services 3.5 Source Domains The IETF DS working group defines a source domain as the domain that contains one or more nodes that originate the traffic that receives a particular service. The traffic that is sent from a source domain may be marked by the traffic source directly or by intermediate nodes before leaving the source domain. It is important to understand that the first PHB marking of the data packets is not done by the sending application. The first PHB marking of packets that are sent from as application can be done in the source host or in the first router the packets passes. The packets are identified with their IP address and source port. 2001 Copyright SCUT DT&P Labs

3.5 Source Domains Host Router 3. Differentiated Services Example 1: Initial Marking in the Host. Application Host Data TCP / UDP TCP Data Classifier IP TCP Data Router DS Byte Marker 2001 Copyright SCUT DT&P Labs

3.5 Source Domains Host Router 3. Differentiated Services Example 2: Initial Marking in the Router. Application Data Host TCP / UDP TCP Data Router Classifier IP TCP Data DS Byte Marking 2001 Copyright SCUT DT&P Labs

3.6 Configuration and Administration of DS Components with LDAP 3. Differentiated Services 3.6 Configuration and Administration of DS Components with LDAP In a DS network, the service level information must be provided to all network elements. Network administrator can define different service level for different customer and provide this information to all boundary components. It is necessary to use standardized format to store the administrative policies in the directory server function and standardized mechanism to provide the information to the network elements. 2001 Copyright SCUT DT&P Labs

3.6 Configuration and Administration of DS Components with LDAP 3. Differentiated Services 3.6 Configuration and Administration of DS Components with LDAP The Lightweight Directory Access Protocol (LDAP:轻量级目录访问协议) is deployed for accessing directories. Policy rules for different service levels are stored in directories as LDAP schema and can be downloaded to devices (hosts, routers) that implement the policies. 2001 Copyright SCUT DT&P Labs

（1）资源充足时各单流能否享受其预约的资源； （2）有额外资源并允许竞争时各单流能否按比例分配这些额外的资源； 3.7 DiffServ 中带宽分配的公平性问题 公平性 DiffServ在区域边界将单流聚合为流聚集，内部节点PHB处理的对象是流聚集，所谓公平性是指在属于同一流聚集中的各单流能否享受同等的待遇。 主要问题： （1）资源充足时各单流能否享受其预约的资源； （2）有额外资源并允许竞争时各单流能否按比例分配这些额外的资源； （3）资源总量不足时，各单流能否按预约资源比例获得相应的降级服务。 2001 Copyright SCUT DT&P Labs

3.7 DiffServ 中带宽分配的公平性问题（续） 影响公平性的主要因素： （1）各单流的业务量变化的特性不同。一般情况下，突发程度大，末端有拥塞控制机制，流量大，回路响应时间长，连接时间短的流在带宽竞争中处于弱势； （2）服务实现机制的影响。包括传输过程中的各个环节：边界分类调节、内部PHB以及是否有反馈控制等。 2001 Copyright SCUT DT&P Labs

3.7 DiffServ 中带宽分配的公平性问题（续） 适应与非适应流共享AF（确保服务）时公平性 （1）适应流：末端系统实现流拥塞控制，能根据网络拥塞情况自动调节发送速率，如TCP； （2）非适应流：没有任何末端拥塞控制机制的流，如UDP； （3）将适应流与非适应流分配于AF组内不同的优先级，令适应流的优先级相对高于非适应流，以抵消非适应流在竞争带宽时的优势。 2001 Copyright SCUT DT&P Labs

3.7 DiffServ 中带宽分配的公平性问题（续） 公平性的通用解决办法 （1）避免在网络内部发生拥塞：在边界节点严格控制进入区域的总流量； （2）加入动态反馈机制：内部节点搜集周边情况并通报给边界节点，边界节点根据反馈获得的有关各单流的适宜速率，调控进入网络的流量以避免拥塞。 2001 Copyright SCUT DT&P Labs

4. Combine of IntServ and DiffServ to offer end-to end QoS 2001 Copyright SCUT DT&P Labs

4.1 IntServ 和 DiffServ 体系结构比较 对比项 IntServ DiffServ 服务区分的粒度 单个流 流聚集 路由器中的状态维护 基于每个流 基于每个流聚集 分组的分类依据 分组的多个头部域 IP头部的DS字节 服务区分的类型 确定性的或统计上的保证 绝对的或相对的保证 接纳协议 需要 仅对绝对的区分需要 信令协议 需要（RSVP） 相对保证不要，绝对保证需要 协调 端到端 局部的（逐跳） 可扩展性 受流的数量限制 受服务的种类限制 网络计费 基于流的特性和QoS需求 基于服务类的使用 网络管理 近似于电路交换网络 近似于现有的IP网络 域间配置 需要多边协议 需要双边协议 2001 Copyright SCUT DT&P Labs    

4.2 DiffServ 网络区支持 IntServ/RSVP 的意义 在DiffServ区实现基于资源的接纳控制 在DiffServ的网络区指定一个支持IntServ的接纳控制代理可以优化资源的使用，提高DiffServ区对于定量QoS应用的服务质量。 在DiffServ区实现基于策略的接纳控制 在DiffServ的网络区采用RSVP接纳控制代理，可以在决定资源分配时采用针对特定客户的策略，为特定的用户和应用有效地分配资源。 2001 Copyright SCUT DT&P Labs

4.2 DiffServ 网络区支持 IntServ/RSVP 的意义（续） 传输识别及分类中的辅助作用 在DiffServ网络内部，资源分配基于IP包头的DSCP值，主机往往需要知道网络如何翻译DSCP，此时主机可以使用显式信令协议（RSVP）通过询问网络来获取。 2001 Copyright SCUT DT&P Labs

4.3 DiffServ 网络区支持端到端 IntServ 的方案 逻辑框架结构 BR2 BR1 Rx Tx ER2 非DiffServ区 非DiffServ区 ER1 2001 Copyright SCUT DT&P Labs

4.3 DiffServ 网络区支持端到端 IntServ 的方案（续） 实现机制 （1）发送方Tx和接收方Rx都使用RSVP来传达特定应用所需的定量的QoS请求； （2）在非DiffServ区可能支持IntServ RSVP的资源预留；也可能不支持IntServ RSVP，此时RSVP消息不受妨碍地进行传输； （3）根据不同的实现，RSVP信令可能被、也可能不被DiffServ区内的路由器处理（至少透明穿过）； 2001 Copyright SCUT DT&P Labs

4.3 DiffServ 网络区支持端到端 IntServ 的方案（续） （4）在DiffServ区不识别RSVP的情况下，边界路由器ER1和ER2作为DiffServ区的接纳控制代理，根据DiffServ区的资源情况和与客户协商的协议实施流聚集传输控制； （5）在DiffServ区能够识别RSVP的情况下，边界路由器ER1和ER2根据DiffServ区的资源情况和客户定义的策略实施接纳控制，边界路由器BR1和BR2参加RSVP的信令过程并作为DiffServ区的接纳控制代理。 （6）在DiffServ区采用流聚集传输控制，一般不进行RSVP的分类。 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案 静态资源管理： （1）网络中DiffServ区内不含能够识别RSVP的设备； （2）在DiffServ网络区的客户和网络所有者之间通过服务层描述SLS（Service Layer Specification）协商建立一个静态契约； （3）边界路由器ER作为DiffServ网络区的接纳控制代理，配置了SLS所表示的信息； （4）特点：相对简单；但灵活性差，不容易支持SLS的灵活改变，会使DiffServ网络区的资源难以有效利用。 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案（续） 动态资源管理： （1） 在此方案中，DiffServ网络区的边界路由器BR和区内的路由器都可支持RSVP； （2）上述的DiffServ网络区路由器只是以某种形式利用RSVP信令，仍然使用IP包中的DSCP值对流聚集进行识别、分类和调度； （3）当一个新单流要加入到行为流聚集中时，使用动态提供机制和显式（RSVP）信令进行接纳控制； （4）使用RSVP信令将流的描述和期待的DSCP传给DiffServ区中的路由器。 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案（续） （5）特点： DiffServ区中路由器的控制平面是RSVP，数据平面仍然是DiffServ； 即利用了RSVP信令的优越性，又保持了DiffServ的可扩展性； 通过RSVP将DiffServ区内的可用资源的改变通知给DiffServ区外部的IntServ节点，提高DiffServ区内资源利用的效率和接纳控制的可信度。 DiffServ区可根据外部资源的请求情况改变区内资源提供的服务方式。 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案（续） 动态资源管理的实现方式： （1）聚集RSVP方案 在此基础上，每个区域都可选择其独特的资源管理的机制。 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案（续）： （2）面向单个流的RSVP方案 DiffServ网络区内的路由器对发起于区外的IntServ节点的、面向单个流的RSVP信令请求进行响应； 使用面向单个流的接纳控制，资源的使用可能获得更高的利用率，但处理RSVP信令资源的需求量也较大； 在DiffServ网络区内面向单个流的RSVP和聚集RSVP可结合使用：可在区内的边界和外围部分使用面向单个流的RSVP，在核心区域使用聚集RSVP。 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案（续）： 2001 Copyright SCUT DT&P Labs

4.4支持端到端 IntServ 的 DiffServ 网络区 资源管理方案（续）： （3）集中式动态资源管理方案 在DiffServ区内建立专用的集中管理实体BB（Bandwidth Broker）－－带宽中介服务器，该实体负责记录本区域的资源占用情况； 边界路由器可采用包括RSVP在内的有关定制协议，与集中管理实体交换信息，进行有关的接纳控制。 带宽中介服务器BB也可以采用分布式的体系结构，这时需要其它协议来维护不同BB数据库之间的一致性。 集中式动态资源管理目前也是一个开发的研究课题。 2001 Copyright SCUT DT&P Labs

The End 2001 Copyright SCUT DT&P Labs