Transcript Document
Presentation_ID © 2001, Cisco Systems, Inc. All rights reserved. 1 Is Internet Ready for Multimedia? Petre DINI Senior Technical Leader Cisco Systems, Inc., USA Concordia University, Montreal, Canada [email protected] MMNS 2002 Panel MMNS 2002 © 2002, Cisco Systems, Inc. All rights reserved. 2 What must be ready? Today QoS, ToS Multimedia Text, Graphics Voice Video Multimedia traffic End-points Congest ion QoS Internet Sharing Core End-points Tomorrow QoS, Pricing MMNS 2002 Santa Barbara, CA User User Multicast Unicast Packetscheduling User User IntServ DifServ PolicyMPLS © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 3 Multi-target Readiness Technology Deployment -Many IP phones, -Many VoIP ports shipped - Great strides in control and signaling - H.323, SIP, MEGACO/MGCP Operations Business -Traditional telephony business models - Music exchange - Video-conferences - Video-phones MMNS 2002 Santa Barbara, CA User-aware - SLA, QoS, Costs - Service interactions - User perception © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI - IP PBX - Class 4/5, PSTN replacement - Managed services 4 Why “almost ready”? Text, Graphics - E-business - Portals - www - E-mail --? Voice - SVMSs - Legality -? Video -SLA -Billing - Legality -? Multimedia -Service guarantee -Mixed SLAs -Security/Privacy -? ?????? - ………………….. - End-to-end congestion control - Delay-tolerant multimedia networks - Fairness use (credits/penalties, costs models) - User QoS perception multimedia network aware - …………………. MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 5 End-to-end congestion control - Round-robin (high computational complexity, modification to routers) - Slow start (collaboration of all end-users) - Congestion avoidance (based on a modest performance required) - Fair queuing paradigm for end-to-end congestion control (protocol design) - TCP relies on collaboration of all the end-users (same common mechanisms) - TCP-friendly paradigm as the basis (the throughput for a flow must decrease when its loss increases) 1. - However, audio and video applications are loss-tolerant, and the degree of loss-tolerance can be controlled with forward error correction - this is supported without significant decrease in the quality perceived by the end user; but these applications don’t tolerate frequent variations of throughput 2. – TCP-friendly paradigm is not appropriate for multicast flows, since a source-based congestion control scheme for multicast flow must adapt its sending rate to the slowest receiver (in sense of throughput computed) Given the size of Internet networks, TCP-friendly assumption is questionable MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 6 Delay-tolerant multimedia networks Thinking Interplanetary Multimedia - The Evolving Interplanetary Internet/Delay Tolerant Networks - Multimedia Internet services across interplanetary distances - Long/variable delays (propagation, queuing, clocking, processing): - Interactivity is affected by delay - Variability in delay may affect applications and protocols - Privacy/Authorization for multimedia applications may be affected by cryptographic processing on simple processors - Reasonable estimation of retransmission time out value becomes complex in networks where variable/unknowing queuing delays occur - Special multimedia applications may be affected: - Image transmission with frequent link interruptions - Multimedia applications exhibiting long delays, significant data loss, potential long link outages Excessive latency affecting TCP directly by limiting its throughput performance Data rate asymmetry affects TCP by altering the smooth flow of acknowledgments MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 7 User QoS perception multimedia network_aware (i) • Objective and subjective QoS parameters • Relationship between user participation and user satisfaction • Four contingency factors (task complexity, system complexity, user influence, user-developer communication) • Measuring user satisfaction • Collect information within the scale excellent....poor, via questionnaire • User computer-assisted perception • Comparative perceptions • QoS user interface • User friendly approaches • Preferred/achieved QoS • Self-adaptable/graceful degradation QoS Internet Multimedia must be user-centric MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 8 User QoS perception multimedia network_aware –ii• Satisfy user needs within a given cost • Correct translation of QoS parameters from the user specification to the desired ´´item´´ through the network • Cover the cost for offering a given multimedia service • Stimulate the service providers to improve their services • Evaluate correctly the cost effort to create a new service • Select various models to price the QoS degradation • Establish the criteria to automatically adapt the QoS with respect to the user constraints and the network unexpected behavior • Classify properly the user within the given profiles and automatically manage the profile changes If QoS pricing should vary with the network state, how quickly? MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 9 Fairness (credits/penalties, costs models) ´´I think it´s ridiculous to have to pay extra for a fundamental service´´ such as bandwidth. Mike McQuary, President and CEO at MindSpring Enterprises Inc, Atlanta ´´How would such services be priced?´´ ´´Consumers and business are also interested in result-oriented pricing as well´´. John Patrick, IBM Internet Technology Vice-President ( several scenarios with pricing schemes ranging from charging a fee on an access-speed basis in some cases to including the fee in the cost of content in others) ´´Whatever economic models materialize, they´re not likely to include flat-fee pricing´´. Clay Rider, Director at Zona Research, California MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 10 Expectations from Multimedia Internet Protocols • Cost-based scheduling • Providing guaranteed performance or graceful degradation • Allocating appropriate resources especially when users’ demand exceeds the network capacity • Pricing mechanisms that change according to QoS parameters • Considering ´´time´´ divided into pricing intervals • Providing the following mechanisms: • Rate control techniques for synchronization and bundling of multimedia connections as well as for multimedia squeezing to control the discrete time independent media used simultaneously with continuous time dependent media • throughput scalability support based on dynamic rate control and threshold QoS monitoring • Adaptive acknowledgement and retransmission techniques considering QoS requirements of specific media data and different network conditions (ATM, satellite, etc.) Result-oriented pricing MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 11 Profiling Multimedia Users and Multimedia Applications • An application programming interface must allow a user to define its QoS profile • A user profile includes: • A MM profile which indicates the desired QoS parameter values • A MM profile which indicates the worst acceptable QoS parameter values Multi Media Profile AudioProfile VideoProfile ImageProfile TextProfile audioQuality CD, phone language F, E, R,... colour b&w, gray,.. resolution HDTV, TV frameRate HDTV rate,.. colour b&w, gray,.. resolution HDTV res TV res minimal res language F, E, R,... MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. CostProfile audioMaxCost videoMaxCost picturemaxCost testMaxCost documentMaxC Petre DINI TimeProfile delay to receive a doc. max. duration of a doc. synchronization Q good acceptable excellent 12 Multimedia QoS, where are we? Multimedia QoS is as good or as bad, as the users perceive it to be! (Jeffrey MacKie-Masson, Liam Murphy, John Murphy) QoS level ? QoS as a key strategic driver minimum level time AS MUCH QoS REGULATION AS NECESSARY, AS LITTLE QoS REGULATION AS POSSIBLE MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 13 Discussions MMNS 2002 Santa Barbara, CA © 2002, Cisco Systems, Inc. All rights reserved. Petre DINI 14 © 2002 , Cisco Systems, Inc. All rights reserved. 15