Part III: Overview of Technologies, ATM, and IP. Networks

Transcription

1 Part III: Overview of Techlogies,, and IP Overview of Networking Techlogies Developments High-speed Characteristics Switching Techniques (Asynchrous Transfer Mode) Principles of Cell Switching Connection Management Layer and Adaptation Layer IP (Internet Protocol) Key Elements Protocol Stack 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 1 ETH Zürich Networks Networks provide an infrastructure for: Interconnecting machines or services (connectivity), Making available scarce resources (resource sharing), Equalizing traffic volumes (load balancing), and Providing alternative fallbacks (reliability). Structuring networks according to dimensions: Expansion (LAN, MAN, WAN), Topology (star, ring, bus, meshed), Performance (low-speed, high-speed, real-time), Administration (public, private), and Task (internet or physical net, intranet or virtual net) B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 2 ETH Zürich

2 Inventions in Telecommunications Momode optical fiber 16 Gbit/s Multimode optical fiber 140 Mbit/s Momode optical fiber 565 Mbit/s Multimode optical fiber 45 Mbit/s Bandwidth bit/s voice channels over cable voice channels over cable 3600 voice channels over cable and microwave 600 voice channels over cable (T3) and microwave 60 voice channels over coaxial cable Carrier telephony carries 12 voice channels on wire First carrier telephony First telephone channels constructed Baudot multiplex telegraph (6 machines on one line) Printing telegraph systems Early telegraphy (Morse code dots and dashes) Oscillating needle telegraph experiments Year voice channels over cable and microwave 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 3 ETH Zürich Transmission Media 10 2 Copper (Coaxial) Attenuation [db/km] 10 1 Copper (UTP) Fiber (Multimode) Fiber (Gradient) Fiber optical media: Low error rates, long distances, low attenuation. Fiber (Momode) Frequency [MHz] 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 4 ETH Zürich

3 Evolution of Bandwidth Last Mile Problem: Connection between the home and the backbone is serious. Requires a huge capital investment. Fiber igres the last mile problem. Bandwidth [kbit/s] LAN LAN POTS WAN Local Area Network Plain Old Telephone Service Wide Area Network WAN POTS Time [year] B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 5 ETH Zürich High Speed Networks Overview (1) High Speed Local Area Networks (LAN): Fast Ethernet: 100 Mbit/s Gigabit Ethernet: 1.0 Gbit/s HIPPI: 800 Mbit/s Fiber Channel: 100, 200, 400, or 800 Mbit/s High Speed Token Ring: 100 Mbit/s (1 Gbit/s) High Speed Metropolitan Area Networks (MAN): FDDI: 100 Mbit/s FDDI-II: 100 Mbit/s (incl. isochrous channels) DQDB: 34 Mbit/s, 155 Mbit/s, or 622 Mbit/s High Speed Wide Area Networks (WAN): -based B-ISDN: e.g., 2, 34, 155, 622, or 2,400 Mbit/s 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 6 ETH Zürich

4 High Speed Networks Overview (2) Carrier support (physical layer): Plesiochrous Digital Hierarchy (PDH): e.g., DS-0: Mbit/s (= 1 voice channel) T-1 (USA): Mbit/s ( 24 voice channels) (also called DS-1) DS: Digital Signal T: Telephone Line E-1 (Europe): Mbit/s E-3 (Europe): Mbit/s T-3 (USA): Mbit/s (also called DS-3) Synchrous Digital Hierarchy (SDH) or Synchrous Optical Network (SONET): e.g., STS/OC-1: MBit/s STS/OC-3: MBit/s ( STM-1) OC: Optical Carrier STS: Synchrous Transfer Signal Level 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 7 ETH Zürich High Speed Networks Overview (3) Carrier support (physical layer) continued: Wavelength Division Multiplexing (WDM) Access techlogies: Cable TV Frame Relay (FR) xdsl (Digital Subscriber Line) Satellite networks and wireless local loop Service techlogies: Switched Multimegabit Data Service (SMDS) 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 8 ETH Zürich

5 Network Dimensions Transmission rate [Mbit/s] HIPPI 1990 System Bus 1980/90 LAN HSTR Fast Ethernet xdsl Token Ring Ethernet Local Area Network FDDI, DQDB, CRMA Metropolitan Area Network 1995 Wide Area Network X.25 N-ISDN -based B-ISDN B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 9 ETH Zürich STM Distance [km] Protocol Layer Architecture Examples Voice... telnet Voice IP... IP... PPP Video Voice Transmission Convergence HDLC WDM SDH/SONET ADSL SDH/SONET ADSL: Asymmetric Digital Subscriber Loop : Asynchrous Transfer Mode HDLC: High Data Link Control IP: Internet Protocol PPP: Point-to-Point Protocol WDM: Wavelength Division Multiplexing 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 10 ETH Zürich

6 Characteristics of High Speed Networks Characteristics of high speed networks: Low bit error rate (fiber optical media), Higher packet error rate (buffer overflow), Existing Jitter (different buffer lengths), Small transmission units (cells), Many connections (context data), High bandwidth (fiber optical media), and Extreme bandwidth-delay product. Protocols have to deal with these issues to alleviate influences of delayed, corrupted, or lost data B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 11 ETH Zürich File Transfer Example File size: 1 Mbyte Link: San Diego Boston Signal delay: 25 ms 64 kbit/s channel: 64 kbit/s * 25 ms = 1,600 bit 0.02 % of file on link 2 Mbit/s channel: 2 Mbit/s * 25 ms = 50,000 bit 0.6 % of file on link 1 Gbit/s channel: 1 Gbit/s * 25 ms = 25,000,000 bit 8 ms for transmission, 17 ms idle 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 12 ETH Zürich SAN SAN 5000 km 1,600 bit 50,000 bit 25,000,000 bit BOS BOS Bits are smaller t faster!

7 Effects on Data in Transit Signal delay is dominating the transmission delay. This grows worse for high transmission rates. Buffer overflows dominate occurring errors. The effect grows worse for fast and real-time traffic. Error recovery has to be a trade-off between a waste of bandwidth or extending delays. Bandwidth is much cheaper than tolerating high delays. Multimedia applications rmally don t like delays. A huge amount of data is in transit within high speed and long distance networks Path Capacity P C : P C = B * D signal B: Bandwidth, D signal: Signal delay B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 13 ETH Zürich Switching Techniques Overview Switching Techniques: Based on circuits, cells, frames, or packets. Simple Intricacy Complex Fixed Behavior of Bandwidth Variable Circuit Switching (STM) Multirate Circuit Switching Fast Circuit Switching Fast Packet Switching Frame Relay Frame Packet Switching Switching Circuit switching suffers from fixed bandwidth constraints for bursty traffic. Packet switching allows for variable bandwidth B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 14 ETH Zürich

8 Circuit information are stored during establishment times in a translation table. Delay is determined by the propagation delay and the processing in switches. Bounded to 450µs by ITU-T. Bit error rates are caused by single bit errors (switching malfunction) or bursts (loss of synchronization). Inflexible, e.g., G.703 PCM. Fixed bandwidth. Circuit Switching Incoming Link 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 15 ETH Zürich l 1 l 2 l m Time Slot 1 2 m 1 2 m 1 2 m Outgoing Link O 1 O 2 O 3 O 1 O 2 O 3 O 1 O 2 O 3 Time Slot m 1 m 1 Packet Switching Based on user data that are encapsulated in packets. Concept is based on techlogy available in the 60s: Erroneous links: Link-based error control in complex protocols. Low bandwidth links: High delays (due to retransmissions) and low speed (due to protocol processing) Lacking support of real-time and multimedia traffic Software-based protocol implementations: Variable sized packets require a complex buffering. X.25 is the oldest example of packet switching nets B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 16 ETH Zürich

9 Frame Relay (1) Frame Relay supports connection-oriented services. Subscriber Network Interface (SNI) defined between customer (router) and PTO equipment. Support of pure data, t particularly voice etc. Multiplexes flows of data being divided in data blocks. Flows are carried in virtual channels which may exceed their bandwidth as other channels are idle. Frame Relay may carry X.25 packets/frames. Performance: Different implementation approaches exist, however, 2 Mbit/s access speeds are common. Insufficient guarantees on bandwidth and delay variation. PTO: Public Telecommunication Operator 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 17 ETH Zürich Frame Relay (2) Physical and data link layer specifications available. The data link layer is based on LAPD (ISDN): Data link services: addressing (DLCI, local significance). Error control is left out as an end-to-end function. Core LAPD frame: Byte 1 2 variable 2 1 Flag Address Payload FCS Flag DLCI C/R EA DLCI FECNBECN DE EA bit DLCI: Data Link Connection Identifier C/R: t used EA: Extended Address FECN: Forward Error Congestion Notification BECN: Backward Error Congestion Notification DE: Discard Eligibility 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 18 ETH Zürich

10 Frame Relay (3) Frame Relay DLCI assignments: 0 Reserved for Call Control Signaling 1-15 Reserved Assigned to Permanent Virtual Circuits (PVC) Reserved 1023 Local Management Interface A simple sample Frame Relay network: 1007 User A Frame Relay Interface Frame Relay Interface User B 16, 145, 1007: DLCI Permanent Virtual Circuits 145 Frame Relay Interface User C 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 19 ETH Zürich Comparison Summary of important functional differences: Connection-oriented Connectionless Frame Boundaries Bit Stuffing CRC Error-Control ARQ Flow-Control Multiplexing of log. Channels X.25 Frame Switching or Frame Relay Heavy weight Light weight Techlogy CRC: Cyclic Redundancy Check 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 20 ETH Zürich

11 Part II: Overview of Techlogies,, and IP Overview of Networking Techlogies Developments High-speed Characteristics Switching Techniques (Asynchrous Transfer Mode) Principles of Cell Switching Connection Management Layer and Adaptation Layer IP (Internet Protocol) Key Elements Protocol Stack 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 21 ETH Zürich Cell-based Switching Integration of a variety of services: Bursts are smoothened. Isochrous data are delivered according to their jitter. Data may be multiplexed statistically, if the overall bandwidth is sufficient. Efficiency: statistical multiplexing gain. Delay problems occur in case of sending packets of different length. Extremely long blocking can be avoided, if cells of fixed-size are used. If the overall cell length is too big, a similar problem appears B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 22 ETH Zürich

12 Handling Cells Segmentation/Reassembly To sent packets over cell-based networks, packets have to be segmented and reassembled again. Packet Cells Packet The segmentation and reassembly (SAR) functionality is placed right above the cell level. User Data HLP User Data Cell Header SAR HLP UD SAR User Data... SAR User Data SAR Header Cell... SAR HLP UD Cell SAR User Data Cell SAR User Data HLP Higher Layer Protocol Header Cell Payload Length Total Cell Length 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 23 ETH Zürich Structure of an -based B-ISDN Network -connected MM-Workstation Customer Premises Equipment UNI Core -Switch Edge -Switch UNI - Switch NNI - Switch - Switch NNI - Switch NNI - Switch NNI NNI - Switch : MM: NNI: UNI: Asynchrous Transfer Mode Multimedia Network Node Interface User Network Interface UNI -connected MM-Workstation 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 24 ETH Zürich

13 B-ISDN Reference Model I.321 Modeling communication systems is done in a logically hierarchical structure, e.g., ISO/OSI BRM. Relation between OSI and B-ISDN/ undefined. The plane approach has been used within B-ISDN. Management Plane Control P. User P. Higher Layer Protocols Adaptation Layer Layer Physical Layer Plane Management Layer Management P.: Plane 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 25 ETH Zürich Connections (1) Connections, links, equipment, and identifiers: Links Equipment Identifiers Connections End-system Cross-connect Switch VCC 1 VPC 1 VPC 2 VPC 3 VCI 1 VCI 2 VCI 3 VPI 1 VPI 2 VPI 3 VPI 4 VPI 5 Switch Cross-connect VPL 1 VPL 2 VPL 3 VPL 4 VPL 5 End-system VC: Virtual Channel, VP: Virtual Path, L: Link, I: Identifier, C: Connection 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 26 ETH Zürich

14 Connections (2) Hierarchical connection concept includes: Virtual Connections are identified by two identifiers, which are significant only locally per link in the virtual connection. Error-control is done end-to-end only, if required. High quality links and a good call acceptance control. Flow-control is t provided. High bandwidth delay product. Virtual Channel (VC) is a uni-directional channel, identified by the Virtual Channel Identifier (VCI). Dynamically allocatable connections. Virtual Path (VP) contains a group of VCs, identified by the Virtual Path Identifier (VPI). Statically allocatable connections B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 27 ETH Zürich Connections (3) Simultaneous support of many thousands of VCs requires the cell to carry the VCI field. Supporting many semi-permanent connections between endpoints, carrying many grouped VPs requires the cell to carry the VPI field. VPI 1 VPI k VCI 1 VCI 2 VCI n VCI m Link Pre-assigned VPI/VCI values: 0/0 Unassigned, idle 0/1 Meta-signaling 0/3 Segment flow (between VP end-points, F4) 0/4 End-to-end F4 flow 0/5 Signaling 0/15 SMDS 0/16 ILMI VPI Virtual Path Identifier ILMI: Integrated Layer Management Interface VCI Virtual Channel Identifier SMDS: Switched Multimegabit Data Service 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 28 ETH Zürich

15 Switching (1) Two types of switching may be performed. VP switching ( Cross-connect): Switching between VPs, No evaluation and change of VCIs, Change of VPIs, and Variable number of VCs per VP possible. VC/VP switching ( Switch): Switching in close cooperation between VCs and VPs, Evaluation of VCI and VPI in an intermediate system, Change of VCI and VPI if necessary, and Incoming VCs of one VP may be distributed between many outgoing VPs B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 29 ETH Zürich Switching (2) Use of VPI in a B-ISDN network (cross-connect). VPI in VPI out VPI = 5 VCI = 1, 2, 3 VPI in VPI out PHY VPI = 7 VCI = 1, 2, 3 B A VPI = 7 VCI = 1, 2, 3 VPI = 9 VCI = 3, 4 PHY 2 B-ISDN Network VPI = 7 VCI = 3, 4 PHY VPI in 7 3 VPI = 3 VCI = 3, B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 30 ETH Zürich 3 VPI out C

16 Switching (3) Use of VPI-VCI in a B-ISDN network ( switch). VPI-VCI in VPI-VCI out A VPI = 7 VCI = 1, 2, 3 VPI = 9 VCI = 3, 4 VPI = 5 VCI = 1, 2, 3 PHY 2 VPI-VCI in VPI-VCI out PHY B-ISDN Network VPI = 7 VCI = 3, 4 VPI = 7 VCI = 1, 2, 3 VPI-VCI in VPI-VCI out PHY VPI = 3 VCI = 3, B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 31 ETH Zürich 1 3 B C Layer UNI Cell Format GFC: Generic Flow-Control used at the service interface. PT: Payload Type defines contents of a cell: User data congested, User data n-congested, Operation And Maintenance (OAM) cells, and Resource management cells. CLP: Cell Loss Priority to identify low/high priority cells. HEC: Header Error Control. GFC 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 32 ETH Zürich Byte bit 5 Byte 48 Byte Header VPI VCI VCI HEC Payload PT VPI VCI CLP UNI: User-Network Interface

17 Layer NNI Cell Format VPI: Virtual Path Identifier comprises of 12 bit length. PT: Payload Type defines contents of a cell: User data congested, User data n-congested, Operation And Maintenance (OAM) cells, and Resource management cells. CLP: Cell Loss Priority to identify low/high priority cells. HEC: Header Error Control. Byte bit 5 Byte 48 Byte Header Payload VPI VPI VCI VCI VCI PT CLP HEC NNI: Network-Network Interface 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 33 ETH Zürich Structure of the AAL AAL includes sublayers: Segmentation and Reassembly (SAR) between packets/cells. Convergence sublayer (CS) for servicedependent adaptation: Common Part Convergence Sublayer (CPCS) and Service Specific Convergence Sublayer (SSCS). Layers may be empty. Class A Class B Class C/D Class C/D CS-1 SAR-1 AAL 1 SSCS-2 CPCS-2 SAR-2 AAL 2 SSCS-3/4 CPCS-3/4 SAR-3/4 AAL 3/4 Adaptation Layer Layer SSCS-5 CPCS-5 SAR B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 34 ETH Zürich AAL 5

18 AAL Comparison Criteria AAL 1 AAL 2 AAL 3/4 AAL 5 AAL Service Class Message Delimiter Advanced Buffer Allocation Multiplexing CS Padding CS Protocol Overhead CS Checksum SAR Payload SAR Protocol Overhead SAR Checksum A /47 Byte 1/2 Byte B yes 0/46 Byte 2 Byte 1/47 Byte 3 Byte C/D BTAG yes yes 4 Byte 8 Byte 44 Byte 4 Byte 10 bit C/D Bit in PTI 0/47 Byte 8 Byte 32 bit 48 Byte 0 PTI: Payload Type Information 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 35 ETH Zürich Functions per Layer Summary Layer AAL PHY Subl. CS SAR TC PM Function Handles transmission errors Handles lost and misinserted cell conditions Handles timing between source and destination Handles cell delay variation Segments higher-layer information into 48 Byte fields Reassembles cell payload in higher layer information Multiplexes cells from different channels Generates cell header (first four bytes) Performs payload type discrimination Performs traffic shaping and flow control Routes and switches cells as needed Indicates cell loss priority and selects cells for discarding HEC header sequence generation and verification Cell delineation Transmission frame generation and recovery Bit timing 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 36 ETH Zürich

19 Part II: Overview of Techlogies,, and IP Overview of Networking Techlogies Developments High-speed Characteristics Switching Techniques (Asynchrous Transfer Mode) Principles of Cell Switching Connection Management Layer and Adaptation Layer IP (Internet Protocol) Key Elements Protocol Stack 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 37 ETH Zürich IP Techlogy Key elements of the techlogy used in the Internet: Packet switching, using datagrams No connection-dependent state information in the network Distributed management Many physical subnetwork techlogies One network protocol Two transport protocols Infrastructure for hundreds of different distributed applications Scalability: to accommodate exponential growth 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 38 ETH Zürich

20 IP Protocol Stack Application layer HTTP FTP DNS Transport layer TCP UDP Internet layer IP Routing Phys. Network layer Ethernet DECnet 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 39 ETH Zürich Internet Protocol (IP) IPv4 shows addressing problems: Nearly exhausted Class B addresses. Classless Interdomain Routing (CIDR) provides short-term solution only. Routing tables grow extremely fast. IP unicast address space will run out due to radipdly, e.g., increasing Internet hosts and low-end Internet devices. Next Generation Internet, IPv6, delivers solutions: Extended addressing: 128 bit addresses. Address hierarchy levels (hierarchy: subscriber, subnet,...) Anycast addresses to reach the nearest de of a group (in terms of the routing metric). Simplified IP header, including a flow label B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 40 ETH Zürich

21 Internet Network Model SOHO Large Enterprise International ISP Backbone Small Enterprise Regional ISP B ISP: Internet Service Provider SOHO: Small Office and Home 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 41 ETH Zürich Example Backbone ISP: UUNET B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 42 ETH Zürich

22 Example Regional ISP: Switch B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 43 ETH Zürich Switch UUNET Interconnection B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 44 ETH Zürich

23 Co-location Model Extensions broadens the model by other services: More than a pure routing and traffic exchange role. Content provider supported, e.g., with high volumes, selection of n-local transit providers. E.g., Multicast, Web, DNS, policy-based route services. Router Router Router Router Router Router Multicast Router Router Router Router Router Router Route Server Router Web Server DNS Server 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 45 ETH Zürich Services Integrated Internet QoS Guarantees Configuration Zone State Information Required Protocols Status Best-effort IntServ DiffServ ne entire network ne ne operational per data stream per session (dynamic) end-to-end data stream per data stream, in router signaling (RSVP) matured aggregated log-term (static) domairiented (ne, in BB, in edge router) bit fields (BB, COPS) worked on IntServ: Integrated Services, DiffServ: Differentiated Services, QoS: Quality-of-Service RSVP: Resource Reservation Protocol, BB: Bandwidth Broker, COPS: Common Open Policy Service 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 46 ETH Zürich

24 IntServ Implementation Integrated Services Architecture (IntServ) supports best-effort and guaranteed services. Traffic control functions: Admission control, Packet classifier, and Packet scheduler. Optional: Policy control (COPS). Protocol support: Admission Control Application Packet Classifier Resource Reservation Packet Scheduler Resource reservation, E.g., RSVP (Resource Reservation Protocol). Host and router require similar functionality. Control Data Policy Control 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 47 ETH Zürich The DiffServ Approach Requirements for Differentiated Services proposals: Aggregated bandwidth allocation without the need of per-session signaling and complex router state, Aggregated QoS guarantees with edge-complexity, Long-term service contracts within a single domain, Integrated and simplified accounting, Better traffic isolation for performance predictability, and Better services for users willing to pay more. A set of new proposals for DiffServ: Expedited Forwarding (EF) and Assured Forwarding (AF) B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 48 ETH Zürich

25 DiffServ Implementation 2000 B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 49 ETH Zürich References (1) F. Fluckiger: Understanding Networked Multimedia; Prentice Hall, London, England, 1995, ISBN R. Steinmetz: Multimedia Techlogie; Springer Verlag, Bonn, Germany, 1999, ISBN M. de Prycker: Asynchrous Transfer Mode Solution for Broadband ISDN; 3rd Edition, Prentice Hall, Englewood Cliffs, New Jersey, U.S.A., 1995, ISBN ITU-T Maintenance Principles: Frame Relay Operation and Maintenance Principles and Functions; I.620, October B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 50 ETH Zürich

26 References (2) R. J. Vetter: Concepts, Architectures, and Protocols; Communications of the ACM, Vol. 38, No. 2, February 1995, pp B. G. Kim, P. Wang: Network: Goals and Challenges; Com. of the ACM, Vol. 38, No. 2, February 1995, pp M. de Prycker: Asynchrous Transfer Mode Solution for Broadband ISDN; 3rd Edition, Prentice Hall, Engle-wood Cliffs, New Jersey, U.S.A., 1995, ISBN T. Braun: Die Internet-Protokollfamilie der nächsten Generation; Praxis der Informationsverarbeitung und Kommunikation, Vol. 19, No. 2, 1996, pp X. Xiao, L. M. Ni: Internet QoS: A Big Picture; IEEE Network Magazine, Vol. 13, March/April 1999, pp B. Stiller, B. Plattner ETHZ-TIK, D. Bauer IBM Research CM III 51 ETH Zürich