Beruflich Dokumente
Kultur Dokumente
Fall 2002
Outline
Introduction to multicasting
Multicast vs. multiple unicast Pros of multicasting Services offered Example applications Requirements for achieving reliability Components of a PGM network Operation PGMPDU types PGMPDU formats
Introduction to Multicasting
Communication can be of three types: unicast,
Unicast: One to one communication between hosts. The destination address is the address of the unique host which is the intended recipient of the message.
Introduction to Multicasting
Broadcast: One to all communication between hosts on a given network. The destination address is the address of the network.
Network
11111111
Internet
Introduction to Multicasting
Multicast: One to many (but not all) communication between hosts. A single packet is transmitted to a group of recipients. The destination address is the address of the group.
Multicast
Multiple Unicast
Slide copied from pg. 407, Forouzan: TCP/IP Protocol Suite, 2nd edition
Multiple Unicasting:
The source transmits
multiple packets.
Pros of Multicasting
Bandwidth:
0.8 0.7 0.6 0.5
No. of Clients
Server Load:
Server (source) only sends one copy of the packet
Network Loading:
Network has fewer packets to transfer
Multicast data packet (A-PDU) delivery from multiple sources to multiple receivers. Reliability:
Data integrity (uses checksums) No-loss OR detectable loss Guarantees that a receiver in a group either receives all data packets (A-PDUs) or is able to detect unrecoverable data packet loss. No-duplicate
Ordered or unordered
Uses the services provided by IP Multicast: Unreliable (best effort) packet delivery.
Example Applications
Financial services Transmitting news, stock quotes, financial indices and other documents to multiple receivers. E-Learning Streaming audio and video to many students at once. Interactive communication between classmates Interactive TV Wireless Entertainment Delivery of data and content to hundreds of millions of cell phones. Bulk Data transfer
T I M E O U T
Used by receivers to indicate the non-receipt of data. In the Case of Loss How do the receivers know that some data that was ACK sent was lost? Loss can be detected since subsequent packets after the loss arrive out of order. No fixed answer: Depends on the application. Certain criteria can be used for deciding on the type of acknowledgement.
Receiver
Sender
Negative ACKs:
Frequency is a function of
buffer management.
and reliability.
For Multicast Protocols: The burden of positive acknowledgements can overwhelm the source. Positive ACK based reliable multicast protocols do not scale well!!
PGM uses negative acknowledgements (NAKs).
Source receivers in the group send a NAK to All the sends a single copy of the first packet.
The receivers detect a lost packet since a different seq. no was expected
the source Packet Gets lost/corrupted at thethe number of receivers in the group The Number of NAKs equals first link. Source sends the of NAKs cause network congestion A large number next packet in the sequence. Packet reaches all the receivers in the group. This condition is known as NAK Implosion
Unlike other transport layer protocols which only deal with end-to-end (machine-to-machine) communication, PGM requires PGM compatible routers.
communication.
Routers
Network Elements
Operation
Normal Data Transfer (i.e. without loss)
Packets are multicast to the group. No reverse traffic to the source if the transmission was successful
Operation
Source Path Messages (SPMs)
If the data establishes a path for router. The router path repeated by each the particular source by process is changes, subsequent SPMs update the path duplicates the SPM. SPMs are multicast by the state in PGM network elements. They establish source pathsource at Address)intervals in frequent as shown. then replaces the path NLA in the first upstream noting the Path NLA (Network address of eachin the with the The router receivers simply note the Layer SPM SPM. exactly the same always contains the address of the (A-PDUs) way that data packets multicast. They provoke own outgoing interface. are always know the The of its prompt Hence the network detection of missing network element. addressPath NLA elements and receiverspackets nearest A the absence at the first transmit. in SPM arrives next upstream or source. of upstream network element network element/source. address of SPM isdata to router. Finally the the transmitted downstream.
Operation
Case 1: A single packet loss
NE forwardsSequence the (i) is last-hop to the Network Receiver NEthe retransmitted data (RDATA) togroup. Element Upstreamunicasts a NAK to themulticast PGMand multicasts a Packet with unicasts no NAK to the source only those NEs and along the exact reverse path from the (NE) to the downstream did not Elements original NCFgroup members thatNetworkreceivethe source. data. Packet with Sequence no. (i + 1) is now multicast. This is achieved by multicasting RDATA packets NCF the NE (+ve) acknowledges the immediately sends aaNCF to the NE Source, on receiving a NAK, NAK by multicasting only on packet. Receiver detects loss of packet with Sequence No. (i). interfaces lost packet NAK was received. NE also forwards the is retransmitted (multicast) to the Finally the on which a NAK to the upstream PGM NE. group.
Operation
Case 2: Multiple packet loss
NE forwardsSequence the (i) is last-hop to the only those NEs Receivers unicast NAKs toNAKmulticastPGM Network Upstream NEthe retransmitted data (RDATA) togroup. Elements Packet with unicasts no the to the source and multicasts a and along the exact reverse path from the (NE) to the downstream did not Elements original NCFgroup members thatNetworkreceivethe source. data. Packet with Sequence no. (i + 1) is now multicast. This is achieved by multicasting RDATA packets only on the NE (+ve) acknowledges the immediately sends a NCF packets. Source, on receiving a NAK, NAKs by multicasting NCF to the NE Receivers detect loss of packet with Sequence No. (i). interfaces lost packet NAK was received. NE also forwards a a is retransmitted (multicast) to the group. Finally the on which single NAK to the upstream PGM NE.
SPMs have two functions: They establish source path state PGM network elements They provoke prompt detection of missing packets in the absence of data to transmit Data packets carry ODATA (original data) and RDATA
(repair data) NAKs are transmitted by receivers to request repairs. NCFs are used to acknowledge the receipt of NAKs
Sent by a source to establish source path state in network elements and to provide transmit window state to receivers. network header destination address of an SPM is The network-header source address of an SPM is the a unicast NLA of address. multicast groupthe entity that originates the SPM.
Source Port Type Options Global Source ID Global Source ID TSDU Length SPM Sequence Number Trailing Edge Sequence Number Leading Edge Sequence Number NLA AFI Path NLA Reserved Destination Port Checksum
Type
Options
Global Source ID
Checksum
TSDU Length
Global Source ID
Global Source ID
NLA AFI Source NLA
TSDU Length
Reserved
NLA AFI
Multicast Group NLA Option Extension when present
Reserved
References
T. Speakman, J. Crowcroft, J. Gemmell, D. Farinacci,
S. Lin, D. Leshchiner, M. Luby, T. Montgomery, L. Rizzo, A. Tweedly, N. Bhaskar, R. Edmonstone, R. Sumanasekera, L. Vicisano: RFC 3208: PGM: Reliable Transport Protocol Specification, (ftp://ftp.isi.edu/in-notes/rfc3208.txt)
Derek Chen-Becker and Manoj Singla: Implementation
Questions?