9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Group (Multicast) Communication in Wide Area Networks Mostafa Ammar Don Towsley College of Computing Dept. of Computer Science Georgia Tech University of Massachusetts Atlanta, GA Amherst, MA

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Definition u Multicast u Multicast: is the act of sending a message to multiple receivers using a single local “transmit” operation

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley A Spectrum of Paradigms Unicast Broadcast Multicast Send to one All Send to some

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley The Layering of Multicast MulticastbyUnicastMulticastbyBroadcastMulticastbyMulticast

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley The Many Uses of Multicasting u Teleconferencing u Distributed Games u Software/File Distribution u Video Distribution u Replicated Database Updates

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Application Models u Point-to-Multipoint: Single Source, Multiple Receivers u Multipoint-to-Multipoint: Multiple Sources, Multiple Receivers u Sources are receivers u Sources are not receivers

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Scope of Tutorial u Support for multicast communication in –Transport –Network –Link Layer u Also important: enforcing reception semantics across receivers (ordering, atomicity) -- in distributed computing literature

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Support Challenges u Overhead for network layer support u Scalability u Dealing with heterogeneity

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Outline u IP Multicast and the Mbone u Multicast Routing u QoS and Real-Time support u Reliable Multicast Transport u Multicast Flow Control u ATM and IP/ATM Multicast u Summary

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IP Multicast and the Mbone

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Server-Oriented Multicast u source sets up one-to-many multicast group u each source responsible for its own group u examples: –ATM (explicit connection to each receiver) –ST-II (receivers listed in setup message) u for connection-oriented services (packet header size!) u discourages dynamic groups

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Receiver-Oriented Multicast Deering, 1991 u senders need not be members u groups may be of any size u no topological restrictions on membership u membership dynamic and autonomous u host groups may be transient or permanent

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IP Multicast u host-group model u network-level; same packet format, different address u routers do all of the work u special IP addresses:  28 bits  268 million groups (plus scope) u ttl value limits distribution

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Forwarding 1. check incoming interface: discard if not on shortest path to source 2. forward to all outgoing interfaces 3. don’t forward if interface has been pruned 4. prunes time out every minute 5. routers may send grafts upstream Routing done by DVMRP

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IP Multicast (cont.) Basic idea is to flood and prune R R router packet S no receiver

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IP Multicast (cont.) Prune branches where no members and branches not on shortest paths R R 2nd packet S prune

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IP Multicast (cont.) Add new user via grafting; departure via pruning R R S graft R

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Mbone MBONE tunnel endpoint IP router WS u Mbone = multicast backbone u virtual network overlaying Internet u needed until mcast capable routers deployed u IP in IP encapsulation u limited capacity, resilience

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Mbone Protocols u IP u UDP: best effort u RTP: real-time transport u RSVP: resource reservation protocol u SDP/SAP: session description, announcement protocols physical layer IP UDP TCP SAP HTTP SMTP SDP session directory RSVP RTP, RTCP conf. control audio video shared tools

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Session Protocols u session description protocol (SDP) –used to describe (not necessarily) mcast session »name, purpose »start time, duration »media (type, transport protocol, format) »how to receive media u session announcement protocol (SAP) –mcast protocol for SDP –periodic transmission to known mcast address –frequency depends on other announcements and scope

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Session Directory u used to allocate multicast addresses to sessions –birthday problem, N addresses support sessions using random address allocation with negligible collision probability –random allocation currently used in popular session directory tools sd (LBL), sdr UCL) u advertises multicast sessions –uses SDP

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Example: sdr

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Mbone Applications u freeware vic, nv (video), vat, nevot (audio), wb (whiteboard) IVS (teleconferencing) u commercial IP/TV - teleconferencing (Precept) Most group applications use IP unicast E.g., CuSeeMe builds own mcast tree

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley MBone Behavior Motivation: quantitative understanding of session dynamics, and loss behavior u session dynamics: –sizes, durations of sessions –join/leave behavior u loss behavior: –loss rates –spatial and temporal correlation

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Mbone Session Dynamics Goal: to characterize the dynamics of Mbone sessions -- Join/Leave Behavior Mlisten u capture data using Mlisten tool u data pre-processed to account for –Mbone un-reliability, –abnormal usage

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Typical Sessions u long-lived (Shuttle) u short-lived (Seminars)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Session Stats Statistics for u interarrival of rcvrs u rcvr membership duration

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Tree Size u multicast vs unicast tree cost u sensitivity of multicast tree to source location u Mbone vs Internet tree

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Audio/Video audio receivers vs video receivers

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Mbone Loss Behavior Goal: to characterize loss behavior on Mbone u metrics –loss rates, spatial and temporal correlation u methodology –rcvr processes on ~25 MBone hosts –rcvrs listen to WRN (& other sources) which sends audio at 80ms intervals; record mcast packet receptions –off-line analysis of trace data

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Measurements

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Where does MBone loss occur? u 4/16/96 dataset u across data sets: backbone loss small compared to overall loss u no loss in end- systems

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Simultaneous Packet Loss u 4/19/96 dataset: 47% pkts lost somewhere u similar results across datasets u models of packet loss: –star: end-end loss independent –full topology: measured per link loss independent –modified star: source-to- backbone plus star   Q: distribution of number of rcvr’s losing pkt?

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Temporal Loss Correlation u Q: do losses occur singly or in “bursts” u 12/11/95 dataset, rcvr loss vs time –occasional long periods of 100% loss –spatial correlation

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Burst Loss Length Distribution u sample distribution from “alps” –12/1/95 dataset u generally isolated losses with occasional longer bursts

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Temporal Loss Correlation u 12/11/95 dataset(above) –similar results over other datasets u “most” loss bursts are 1 packets long u but often significant loss in long bursts

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley MBone Measurements Summary u identifiable backbone loss small wrt overall loss u spatial loss correlation: limited u temporal loss correlation: –most loss bursts have length one –significant loss occurs in long bursts u reference: ftp://gaia.cs.umass.edu/pub/Yajn96:loss.ps u lessons for multicast protocols/applications? u loss/delay correlation? u sensitivity to period, probe size, 1-many, many-many

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Routing

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Theoretical Basis The Steiner Tree Problem is NP-Complete –Graph G = (V, E) –Positive Edge Weights W(e) –R a subset of V –B positive integer bound Is there a subtree of G that includes all R with cost no more than B? –Heuristic less than twice optimum

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Principles of Multicast Routing u Addressing –List Addressing »Not Scalable –Group Addressing »Less Control

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Principles of Multicast Routing u Reuse of unicast routing infrastructure –Desirable –Too Constraining u Multicast routing overhead –Needs to be minimized

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Principles of Multicast Routing u Evaluation –Bandwidth Usage –Delay -- Average, Maximum, Variance –Concentration –Router/Switch overhead

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Basic Multicast Routing Protocols Source AB C D E u Problem: Given a source and a set of destinations, Route same packet to at least (or exactly) this set of destinations

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Basic Multicast Routing Protocols u Multicast by Broadcast –Filter above network layer –Natural in Broadcast Networks (Satellite, Bridged LANs) –Use Flooding in PSN –Bandwidth inefficient, Security concerns

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Basic Multicast Routing Protocols u Separately addressed packets Source AB C D E to A to B to C to D to E

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Basic Multicast Routing Protocols u Multidestination Addressing ABC D E Source to (A,B,C,D,E) to (A,B) to (C,D,E) to (A,B) to (D,C) to (B) to (C)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley u Spanning Tree Forwarding –Shared or Source-Based Basic Multicast Routing Protocols ABC D E Source

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Basic Multicast Routing Protocols u Reverse Path Forwarding –Dalal and Metcalfe –Basis for DVMRP (the original Mbone routing protocol) –Main advantage: Use of existing unicast routing infratsrutcure

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Reverse Path Forwarding u A Broadcast Protocol u Group addressing used u Routers/Switches Forward based on source of multicast packet

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Reverse Path Forwarding u Flood if packet arrives from Source on link that router would use to send packets to source u Otherwise Discard u Rule avoids flooding loops u Uses Shortest Path Tree from destinations to source (reverse tree)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Reverse Path ForwardingS xy z w Destinations touse S y Routing Table to Group

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Reverse Path Forwarding ABC D E Source Shortest Path Tree to Source

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Shared Tree VS Source-Based Tree source-based tree u RPF routes over source-based tree –Good delay properties –Per source and group overhead shared tree u Spanning Tree Forwarding uses shared tree –Per group overhead –Higher delays –More Traffic Concentration

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Routing Performance Evaluation u Simulation, Analysis and Experimentation u Need –Network Models (Waxman, GT-ITM, Tiers) –Application Models (Mlisten) –Traffic Models –Data Loss Models

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Internet Multicast Routing u Group Addressing –Class D IP addresses u Link Layer Multicast u Two Protocol Functions –Group Management »IGMP –Route Establishment »DVMRP, MOSPF, CBT, PIM

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Link Layer Multicast u Example Ethernet: –Ethernet multicast addresses –Algorithmic mapping between IP mcast address and Ethernet mcast address u To join group: –Recognize IP mcast address –Interface recognizes link layer mcast address –Inform local router using IGMP

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Internet Group Management Protocol end-system router(s) u Used by end-system to declare membership in particular multicast group to nearest router(s) –Version 1: Timed-out Leave –Vesrion 2: Fast (Explicit Leave) –Version 3: Per-Source Join

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IGMPv1 u Joining Host send IGMP Report u Leaving Host does nothing u Router periodically polls hosts on subnet using IGMP Query u Hosts respond to Query in a randomized fashion

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IGMPv2 u ADDS: –Group Specific Queries –Leave Group Message u Host sends Leave Group message if it was the one to respond to most recent query u Router receiving Leave Group msg queries group.

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IGMPv3 u Unclear Status?? u ADDS: –Group-Source Specific Queries, Reports and Leaves u Inclusion/Exclusion of sources

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Routing Protocols u Source -based Tree Protocols: –DVMRP –MOSPF –PIM-DM u Shared-Tree Protocols –CBT –PIM-SM

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley DVMRP u Distance Vector Multicast Routing Protocol –An enhancement of Reverse Path Forwarding that : »Uses Distance Vector Routing Packets for building tree »Prunes broadcast tree links that are not used (non-membership reports) »Allows for Broadcast links (LANs)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Forwarding in DVMRP 1. check incoming interface: discard if not on shortest path to source 2. forward to all outgoing interfaces 3. don’t forward if interface has been pruned 4. prunes time out every minute

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley DVMRP Forwarding (cont.) Basic idea is to flood and prune R R router S no receiver packet

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley DVMRP Forwarding (cont.) Prune branches where no members and branches not on shortest paths R R S prune 2nd packet

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley DVMRP Forwarding (cont.) Add new user via grafting; departure via pruning R R S R Report graft

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Overheard on Mbone Mailing List! u “Help, we are unable to send prunes” u Response: “Well, have you tried to send plums? Raisins or grapes? …… Perhaps your multicast implementation does not support fruit at all?”

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast OSPF u Link-State Multicast Routing u Routers maintain topology DBs u Group-Membership-LSA broadcast by routers to advertise links with members u Routers compute and cache pruned SPTs

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Protocol Independent Multicast u Motivation: –DVMRP good for dense group membership –Need shared/source-based tree flexibility –Independence from Unicast Routing u Two PIM modes: –Dense Mode (approx. DVMRP) –Sparse Mode

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley PIM- Dense Mode u Independent from underlying unicast routing u Slight efficiency cost u Contains protocol mechanisms to: –detect leaf routers –avoid packet duplicates

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley PIM - Sparse Mode u Rendezvous Point (Core): Receivers Meet Sources u Reception through RP connection = Shared Tree u Establish Path to Source = Source-Based Tree

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley PIM - Sparse Mode ReceiverSourceRendez-Vous Register Join SourceJoin Prune

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley PIM - Sparse Mode ReceiverSourceRendez-Vous

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Core-Based Trees u A shared-tree protocol Core u One node on shared tree is Core u Sender sends to Core u Core forwards over multicast tree

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Core-Based Trees ABC D E Source Core-based tree Core

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley PIM and CBT Issues u Unidirectional VS Bidirectional Shared Trees u Core/RP Placement and Selection u Multiple Cores/RPS u Dynamic Cores/RPs

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Real-Time Multicasting and Quality-of-Service (QoS)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley The Problem sender receiver timet 0 t 1 t 2 pkts generated (received) How to get smooth, continuous playout u adaptation by appl. –playout delay adj. –loss concealment u perf. guarantees by network

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Requirements for Real-Time Applications u transport protocol must provide timing information u call admission/reservation protocols needed to –determine availability of resources for particular performance requirement –reserve (allocate) resources u support for multicast –heterogeneity, scalability

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RTP: Real-Time Transport Protocol u timing info. for playout u reordering information u loss detection for quality estimation, recovery u synchronization –network jitter –clock drift –intermedia (lip sync) u QoS feedback u source identification

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RTP: Real-Time Transport Protocol Schulzrinne, et al. RFC 1889 u product of IETF Audio Video Transport Working Group (AVT WG) u goals –lightweight, interoperability –easy integration with application –mechanism - not policy –scalability - unicast, multipoint s participants –separation of control/data

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RTCP: RTP Control Protocol u provides monitoring capablities –quality of routes –state of participants (talker indication) u feedback to application –QoS feedback - adjust sender rate u scalability- randomized control traffic; rate decreases as number increases

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Performance Guarantees to Application u deterministic guarantees –absolute guarantees on loss, delay or jitter u statistical guarantees –probabilistic guarantees –cell loss probability <  –prob. pkt delay exceeds D is less than  u different services for different performance requirements

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Example: Internet Services Guaranteed: no loss, upper bound on delay –invoked by specifying traffic (TSpec) Controlled load: negligible losses, like unloaded network => delay-adaptive applications –invoked by specifying traffic (Tspec) Best Effort: traditional service

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Guaranteed Service u user specifies traffic (TSpec) –token bucket spec. ( r - token rate, b - bucket depth in bytes) – p - peak rate – m - minimum policed unit – M - max. packet size u and desired service (Rspec) –desired bandwidth R > r –slack term S b r

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Guaranteed Service (cont.) u delay is decreasing function of R for weighted fair queueing u user does not provide delay bound; delay bound controlled by choice of R,S u call admission, scheduling policy unspecified –service oriented towards WFQ

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Controlled Load Service u user sees unloaded network –buffer loss on order of loss due to noise, faults, etc. –delays should be mostly = propagation delay + processing costs u TSpec is same as for guaranteed service; no RSpec u call admission, scheduling, buffer management unspecified

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Call Setup u contract between network and application –network guarantees performance –application guarantees traffic behavior »peak rate »average rate »burst size u approaches –one pass –two pass

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley One Pass u sender or receiver oriented u source (rcvr) sends resource reservation to rcvrs (source) u cannot specify/guarantee QoS u soft state possible –periodic refreshes u e.g., RSVP u one pass w. advertisement rsvtn source rcvr1 rcvr2 rsvtn source rcvr1 rcvr2

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Two Pass u sender/rcvr initiated u forward phase: check for resources at each link u reverse phase –inform routers if call admitted –reserve resources u reserve max resources u resource reclamation phase can be added u passive two-phase - notify originator source rcvr1 rcvr2 fwd phase rev phase

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RSVP: ReSerVation Protocol Zhang, etal u receivers control reservations –consistent with IP multicast –good scalability –supports heterogeneity u separate resource reservation from usage –packet filters control usage u soft state –end system periodically refresh state

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RSVP Operation u rcvr joins group via IGMP u source sends PATH messages to rcvrs u rcvrs send RESV messages back to source(s) u reservations can be lowered, merged between senders (audio) u one pass => rcvr does not know final QoS => one pass with advertising R R R D D S data (mcast) PATH RESV

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Other Issues u policing/traffic shaping –leaky bucket –Generic Cell Rate Algorithm (GCRA) »peak cell rate, mean cell rate, cell delay variation tolerance u interaction with routing u what does a guarantee really mean? u pricing u performance evaluation

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Grand-Unified Multicast (GUM) u Putting it all together u A protocol for inter-domain multicast routing u Bidirectional Shared Tree for inter-domain routing u Receiver domains can utilize choice of protocol u Very New -- Still needs digesting

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Summary u Multicast Routing is a well researched problem. u Challenge now is –deployment –inter-operability –management

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Reliable Multicast

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Problem How to transfer data reliably from source to R receivers u scalability: 10s s s s s of receivers u heterogeneity u feedback implosion problem

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Feedback Implosion Problem... ACK sender rcvrs

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Issues u level of reliability –full reliability (data) –semi-reliability (video) u ordering –no ordering –ordering per sender –full ordering (distr. computing)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Approaches u shift responsibilities to receivers u feedback suppression u multiple multicast groups u local recovery u server-based recovery u forward error correction (FEC)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Applications u application requirements –low latency »file transfer (finite duration) »DIS, teleconferencing –high throughputs - streaming applications (billing, etc.) u application characteristics –one-many: one sender, all other participants rcvrs (streaming appl. teleconferencing) –many-many: all participants send and receive (DIS)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Sender Oriented Reliable Mcast Sender: mcasts all (re)transmissions selective repeat use of timeouts for loss detection ACK table Rcvr: ACKs received pkts Note: group membership important X sender receivers ACK

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Vanilla Rcvr Oriented Reliable Mcast Sender: mcasts (re)transmissions selective repeat responds to NAKs Rcvr: upon detecting pkt loss sends pt-pt NAK timers to detect lost retransmission Note: easy to allow joins/leaves X sender receivers NAK

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Sender vs. Receiver (cont.) Significant performance improvement shifting burden to receivers for 1-many; not as great for many-many

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Feedback Suppression u randomly delay NAKs u multicast to all receivers + reduce bandwidth - additional complexity at receivers (timers, etc) - increase latencies (timers) X sender NAK X

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Performance of Feedback Suppression Additional thruput improvement for 1-many; costly for many-many

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multiple Multicast Groups u mcast group per pkt –all rcvrs belong to one group for original transmissions –rcvr losing pkt j joins group for j –additional performance improvement for small no. groups (Kasera etal 1997) u receivers divided into destination groups –identify rcvrs with different capabilities –place similar rcvrs into same group –additional improvement (Ammar, Wu 1992)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Server-based Reliable Multicast server sender receivers u first transmisions mcast to all receivers and servers u each receiver assigned to server u servers perform loss recovery u servers can be subset of rcvrs or provided by network u can have more than 2 levels LBRM (Cheriton)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Benefits of Adding Servers u 2 levels, 4 rcvrs/server u equal losses on each link u rcvr oriented, NAK suppression at all levels Comments: u clear performance benefits u how to configure –static/dynamic –many-many

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Local Recovery Lost packets recovered from nearby receivers u deterministic methods –impose tree structure on rcvrs with sender as root –rcvr goes to upstream node on tree RMTP (Lucent) u self-organizing methods – rcvrs elect nearby rcvr to act as retransmitter using scoped multicast and random delays (SRM) u hybrid methods

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Issues with Server- and Local Based Recovery u how to configure tree u what constitutes a local group u how to permit joins/leaves u how to adapt to time-varying network conditions no definitive resolution

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Some Examples of Protocols

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RAMP (TASC) Reliable Adaptive Multicast Protocol u supports sender and rcvr oriented reliability u mixture of reliable/unreliable senders/rcvrs supported u late joins and leaves supported u rate-based flow control

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley RMTP (Lucent) Reliable Multicast Transport Protocol u imposes a tree structure on rcvrs corresponding to multicast routing tree u leaves send periodic status msgs to upstream nodes u nodes inside tree –provide repairs to downstream nodes –send aggregate status msgs upstream u late-joins supported thru 2-level cache u rate- and window-based flow control

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley SRM (LBL) Scalable Reliable Multicast framework u integrated with application u rcvr-oriented using NAK suppression and self- organizing local recovery u supports late-joins and leaves u as built in wb, uses rate-based flow control u has been used with 100s of participants over the Internet

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Other Examples u Single Connection Emulation (SCE, Georgia Tech) –sender oriented, offers semantics of TCP (sender ordering, etc) –late-joins are not supported. u ISIS (Cornell, Isis Dist. Systems) –general purpose toolkit for providing reliable data transfer to dist. applications –sender and rcvr oriented –wide range of ordering semantics supported –late-joins and leaves supported

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Other Examples Local Group based Multicast Protocol (LGMP, Karlsruhe) –self-organizing local- recovery based scheme Xpress Transport Protocol (XTP) –sender-oriented extension of unicast protocol

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Forward Error Correction (FEC) Add redundancy in order to reduce need to recover from losses (e.g., Reed Solomon codes) (k,n) code –for every k data pkts, construct n-k parity pkts –can recover all data pkts if no more than n-k losses + reduce loss probability - greater overheads at end-hosts Q: can FEC reduce network resource utilization?

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Potential Benefits of FEC D1 D3 D2 D1 D2 D3 D1 D3 D2 D1 D2 D3 P P P Initial Transmission Data Retransmission Parity Retransmission One parity pkt can recover different data pkts at different rcvrs P=D1 D2 D3 X X X

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley An Integrated Approach using (k,n) Erasure Codes u transmit packets in blocks of k u send redundancy packets in response to NAKs Assumptions: independent losses, p significant performance gain achievable with small no. of parity pkts, n-k p=0.01

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley End-Host Overheads Software encoding/ decoding rates for 1KB pkt size (Rizzo) => FEC in SW is feasible for many applications

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley End-Host Thruput Comparison: No FEC vs FEC u No FEC, rcvr oriented with NAK suppression u FEC, rcvr oriented with NAK suppression –SW encoding at sender –HW encoding at sender SW FEC thruput determined by sender HW FEC thruput determined by rcvr

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Performance Evaluation - no. transmissions to get pkt j to all rcvrs - no. transmsissions of pkt j needed for rcvr i u well understood for spatially and temporally independent losses u spatial correlation: computationally expensive for general topologies see Towsley85, Tripathi94, Nonnenmacher97 u temporal correlation: has not been touched on (see Nonnenmacher etal 97 for one treatment)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Summary u reliable mcast is a hot topic u unresolved issues –proper integration of different ideas wrt different applications –integration with flow control –interaction with group memebrship –notion of semi-reliability

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Multicast Applications

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Problem u Match transmission rates to –Network capacity –Receiver “consumption” rates

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Flow Control Challenges u Accommodating heterogeneity among receivers and paths leading to them u Preserving fairness among –receivers of same flow –distinct flows u Scalability of feedback

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Flow Control Solutions u Loss-Tolerant Applications (e.g., Video) –Information content per unit time can be preserved at lower data rates u Applications demanding data integrity –lower data rates => lower information content per unit time u Goal u Goal: Co-Existence with TCP?

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u Scalable Feedback Control (Bolot, Turletti and Wakeman) –Receivers measure loss rates –Randomly generated feedback –Source estimates receivers’ state and adjusts video rate by changing compression parameters u Problem: fairness among receivers

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u Improving fairness using Destination Set Grouping (Cheung, Li, Ammar) –Send replicated video streams at different rates –Receivers can control rate of each stream within limits –Receivers can move among streams u Fairness at the expense of increased bandwidth consumption

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u DSG and Single Stream Comparison Single Group Receivers Intra- Stream Protocol DSG High Med. Low Intra- Stream Protocol Inter- Stream Protocol

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u Receiver-Driven Layered Multicast (McCanne, Jacobson and Vetterli) –Single video stream subdivided into layers –Receivers add and drop layers depending on congestion –Challenge: Distributed Consensus, Layer Synchronization

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u Receiver-driven Layered Multicast Source Receivers

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u Receiver-Driven Layered Multicast –Drop Layer: indicated by loss –Add Layer: –No such indication –Use join experiments with shared learning »Reluctance to join layers that failed »Inform others via multicast of failed experiments

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multicast Video Flow Control u Layered Video Multicast with Retransmissions (LVMR) (Li, Paul, Ammar) –Uses agents within network to maintain information about join experiments –Reduces overhead of shared learning

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Less Understood/Mature Area u Some Possibilities: –Window flow control (a la TCP) »Not Scalable, Not Fair (across receivers) –Explicit rate control (e.g., ATM ABR) »Scalable but still not fair –Multiple Multicast Groups

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u The Single Connection Emulation (SCE) Architecture (Window Flow Control) Application TCP SCE IP

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u ATM Available Bit Rate Source Receivers Explicit Rate Messages

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u ATM Available Bit Rate –Consolidation Algorithm –Consolidation Noise –Transient Time –Volume of Feedback

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Multiple Multicast Groups –Simulcasting or Destination Set Splitting (Ammar, Wu and Cheung, Ammar) –Data Partitioning (Bhattacharyya, Towsley, Kurose, Nagarajan) –Cumulative Layering (Vicisano)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Simulcasting: –Similar to DSG protocol for Video –Send multiple (uncoordinated streams) at different rates –Each stream carries all data –Receivers join appropriate stream - one at a time

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Data Partitioning: –Send multiple streams at different rates –Synchronous start time for receivers –Partition data among streams with replication among streams allowed –Schedule data for optimum completion time

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Data Partitioning Example ADCB Flow 1 rate = 1 BC Flow 2 rate = 1 C D Flow 3 rate = 2 R1 R2R3

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Data Partitioning –Can achieve ideal completion time with as many channels as receivers –Can achieve close to ideal with a few channels –Same completion time as simulcast but less bandwidth consumed –Can improve by dynamic rate adjustment –Requires coordination and scheduling among channels

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Cumulative Layering –Multiple data streams at different rates –Each stream contains entire data –Receivers join asynchronously -- Streams transmit for indefinite duration –Schedule to minimize reception time –(Scheme allows for FEC encoding)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast Channel 1 Rate = 1 A BC D Channel 2 Rate = 1 C D A B Channel 3 Rate = 2 B D A C B D A C Receiver 1, BW = 1 Receiver 2, BW = 2 Receiver 3, BW = 4

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Flow Control for Reliable Multicast u Cumulative Layering –Can achieve minimum reception time with asynchronous receivers –Schedulability requires some parameter relationships –Synchronization among channels needed

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Summary hard u Flow control for multicast communication is a hard problem: –Scalability –Heterogeneity –Added dynamic dimension (receivers and their join behavior) –Plenty of room for innovation!

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley ATM and IP over ATM Multicast

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley ATM Multicast u ATM: Connection-Oriented (VC) u Multicast connection establishment –UNI 3.0/3.1 »Source-Controlled »Full Knowledge of Receivers required –UNI 4.0 »Leaf-Initiated Join (LIJ) »Group identified by (Source, Tree ID)

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley IP Multicast over ATM u Problem (for UNI3.x): u Mapping IP’s connectionless, receiver-controlled model to ATM’s connection-oriented, source- controlled model

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley The MARS Architecture u MARS: Multicast Address Resolution Server u Stores mapping between IP group address and Unicast addresses of ATM endpoints

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley MARS -- The Mesh Approach u IP receivers joining group register their ATM addresses with the MARS u IP source consults MARS for list of ATM addresses u IP source opens multipoint connection to ATM end-points u Receivers joining/leaving need to inform MARS

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley MARS -- The VC Mesh ATM Cluster Source Receivers ATM Multipoint Connection MARS MARS Request Reply

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley MARS -- The Multicast Server u The VC Mesh approach requires one multipoint VC per source per group u Alternative: –Provide a multicast server (MCS) per group –One multipoint VC from MCS to receivers –Source sends to MCS –MARS now returns ATM address of MCS

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley MARS -- The Multicast Server Receivers ATM Cluster Source MARS MARS Request Reply Multicast Server

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley VC Mesh VS MCS u VC Mesh –Higher VC consumption –Higher signaling overhead –Better Delay –Less Vulnerability u MCS : Requires Reflection Suppression u Comparison similar to Shared Tree VS Source-based Tree in multicast routing

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley VC Mesh VS MCS

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley The MARS for UNI 4.0 u LIJ is closer to IP model than UNI3.x u MARS is still needed to map IP group address to the (source, tree ID) pair

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Multipoint-to-Multipoint Support in ATM u How to support multiple sources on same multipoint VC and shared tree u Problem: Interleaving of Cells from different AAL PDUs within switch u Solution: –No problem for AAL 3/4 –For AAL 5 : SEAM and SMART

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Simple and Efficient ATM Multicasting (SEAM) u Proposes a Core-Based Tree Approach u Solves cell interleaving by “cut through” switching –Forward cells belonging to same packet together and –Buffer other cells for same VC

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Shared Many-to-Many ATM Reservations (SMART) u Many-to-Many connection over a single VCC u Addresses –Cell-interleaving –On-demand bandwidth sharing (a la RSVP) u Solution: Take Turns (essentially) u Signalling Protocol

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Summary u ATM and Internet multicast mechanisms are incompatible u Convergence approaches have been defined u ATM multicast still has problems: –many-to-many –QoS –routing

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Summary

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Topics Covered in Tutorial u overview of multicast at –network layer (routing, congestion control) –transport layer (reliability, flow control) –session layer (Internet centric view) u examples taken from –Internet, MBone –ATM

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Other Topics u application issues –stronger semantics: ordering requirements, atomicity, etc. (Isis, Horus, Berman et al, Cornell) u session semantics u group membership u performance evaluation: how to evaluate large networks supporting large numbers of multicast applications

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley Other Topics Control Issues: u control issues related to TCP have generated 100s of papers and are still not resolved u control aspects of multicast add at least one additional layer of complexity –a very fruitful area for research and development

9/14/97SIGCOMM97 Tutorial, Copyright Ammar and Towsley