Multicast for Video Streaming .


35 views
Uploaded on:
Category: Funny / Jokes
Description
IP Multicast Overview. Semantics 1 - > Many or Many - > ManyApproach Build tree interfacing source and beneficiaries on Current Infrastructure in Net [1]Group Addressing
Transcripts
Slide 1

Multicast for Video Streaming EE290T Spring 2002 Puneet Mehra pmehra@eecs.berkeley.edu

Slide 2

IP Multicast Overview Semantics 1 - > Many or Many - > Many Approach Build tree interfacing source and recipients on Current Infrastructure in Net [1] Group Addressing – gives adaptability Receivers/senders unconscious of each different Packets conveyed all through tree. Dynamic changes to tree New Receiver - > unite way onto tree Receiver leaving - > pruning way from tree Uses UDP – so no dependability Challenges Efficient steering of information to collectors

Slide 3

Video Multicast Over Net[1] Issues in Multicast over Best Effort Fixed Frame Rate – paying little mind to deferral/jitter Losses – corruption, conceivably clumsy Heterogeneity of beneficiaries Approaches to Multicast QoS asset booking for Multicast Adaptive Rate Control Techniques for Rate Adaptation Single Stream Video Multicast Replicated Stream Video Multicast Layered Video Multicast

Slide 4

Single Stream Video Multicast Only send 1 stream to all recipients. Aces: Easy To Implement Cons: Ignores Receiver Heterogeneity Feedback Implosion INRIA Video Conferencing System Feedback Problem dealt with through probabilistic beneficiary reaction Tradeoff granularity of control versus B/W effectiveness

Slide 5

Efficiency Tradeoff in Single Stream Approach

Slide 6

Replicated-Stream Video Multicast Destination Set Group (DSG) Small # of video floods of shifting quality sent to various multicast bunches Intra-stream Rate control to change stream rate by beneficiaries Inter-stream convention utilized by recipients to switch streams Pros: manages heterogeneity – all the more reasonable Scalable since collector driven Cons: Network conveys excess information

Slide 7

Layered Video Multicast Receiver-Driven Layered Multicast (RLM) Send distinctive "layers" to multicast gatherings, and beneficiary subscribes as required - > adaptable arrangement Congestion - > layer dropping Spare B/W - > layer including Receivers lead assemble join investigations and impart data to others.

Slide 8

Layered Video Multicast Cont. Layered Video Multicast with Retrans. (LVMR) Improve gathering w/in a layer by retransmission Deal w/blockage utilizing Hierarchical Rate Control Hierarchical Rate Control (HRC) Congestion information disseminated at both sender/recipients Intelligent dividing of data - > simultaneous examinations w/less overhead Use pecking order to just illuminate the individuals who need to think around a trial – influenced areas Collaborative layer drop – better way to deal with clog

Slide 9

Error Control in Video Multicast Pure FEC ARQ – From LVMR Local Recovery - assigned collectors at every level in tree help w/rtx. of pkts - > bring down dormancy Don\'t rtx bundles past due date Receivers can exchange unwavering quality/inactivity by picking guardian with wanted properties

Slide 10

Multicast Routing [2,3] Routing – build productive tree from source to collectors Theoretical Results [3] Steiner Tree – limit add up to cost of a multicast tree. NP-Complete. So utilize heuristics to give a "decent" approx. to Steiner Tree. Compelled Steiner Tree – force b/w postpone requirements on connections to collectors. Additionally NP-Complete. So should utilize heuristics All reasonable calculations in view of most brief way tree – limit total of weights on connections along every way from source to collector

Slide 11

Intra-Domain Routing Source-based Routing Tree established at source Dense-mode steering – works best when topology thickly populated with beneficiaries Core-based Approach Select a Rendezvous Point (RP) to root the tree Sparse Mode Routing – More productive than thick mode when couple of, boundless recipients

Slide 12

Dense Mode Protocols Distance Vector Multicast Routing Protocol Uses communicate & prune method to manufacture switch briefest way trees (RSP) Steps: Src bcasts pkt on Lan. Neighborhood switch fwds pkt on all ifaces If pkt got on RPF iface, then it is sent. Leaf switches send prune toward src if no connected recipients Prune message sent to source, and send claim prune if get prune message on all ifaces. A considerable measure of state information kept in ALL switches in net. Multicast expansions to OSPF Uses IGMP locally, then surges information alongside connection state to net. PIM-DM Less intricate than DVMRP since no RPF check is finished. More wasteful thus

Slide 13

Tree Construction in DVMRP [3] S = Source. Dark Circles = Receivers Periodically surge net w/datagrams Leaf switches send prune toward source if there are no gathering individuals on leaf subnet Final Tree is appeared in (d).

Slide 14

Core-Based Routing General Approach A center, or meet point (RP) is designed for a multicast bunch Info about the RP & mapping from gathering to RP is found by switches utilizing bootstrap convention (additionally discovers interchange RP if there should arise an occurrence of disappointment) Receivers unequivocally join tree - > contact RP Src sends information to RP which sends down tree More effective since state just kept in switches on way from src/collectors to RP. Cases CBT – Core-Based Trees PIM-SM – Protocol Independent Multicast/Sparse Mode

Slide 15

Tree development in CBT The Join Process for another hub Receiver Contacts Local Router sends JOIN_REQUEST to the center switch When msg comes to on-tree switch, a JOIN_ACK is sent back each switch accepting JOIN_ACK overhauls state data Periodically send resound demand to parent switch. On the off chance that reverberate not got in time, then switch sends quit-notice upstream and erases state data.

Slide 16

Inter-Domain Routing Probs w/multicast depicted Large level topology - > intricacy and unsteadiness since no BGP-like convention No system to fabricate various leveled mcast directing Solution – Immediate Future Introduce Hierarchy – multi-convention augmentations to BGP (MBGP) Each switch just knows topology of its own area & how to achieve different spaces Used to decide next jump for a host

Slide 17

Inter-Domain Routing Cont. Imagine a scenario where you have a src in one space & beneficiaries in others. Multicast Source Discovery Protocol When src registers w/RP - > a source dynamic (SA) msg is sent to MSDP peers Prevent circles w/per-RPF flooding (ie: if msg got on right iface - > surge) If MSDP knows about neighborhood amass individuals (utilize IGMP), then it will send a join to the src

Slide 18

Long-Term Inter-Domain Proposals Border Gateway Multicast Protocol Bidirectional imparted trees between areas to single root. Require strict distribution of locations among areas. Address Allocation Protocols Multi Address Set Claim – Helps apportion addresses powerfully crosswise over spaces GLOP – a "glop" of locations statically designated among areas

Slide 19

Problems Deploying IP Multicast [4] Complexity Can\'t place it in center switches Hardware more hard to oversee (probs w/firewalls) Makes old switches futile upsets ISP switch relocation show (switches by and large move from center to edge) Domain Independence ISPs would prefer not to depend on remote RPs Don\'t have any desire to be RP for non-clients Security – anybody can send/listen Address Allocation – anybody can pick a class D addr.

Slide 20

References [1] "Video Multicast over the Internet." Xue Li et al. IEEE Network. 1999. [2] "The Evolution of Multicast: From the MBone to Interdomain Multicast to Internet2 Deployment." Kevin Almeroth. IEEE Network. 2000. [3] "Multicast Routing and Its QoS Extension: Problems, Algorithms, and Protocols." Bin Wang and Jennifer C. Hou. IEEE Network. 2000. [4] "Arrangement Issues for the IP Multicast Service and Architecture." Christophe Diot et Al . IEEE Network. 2000.

Recommended
View more...