Download to read offline
![Router output port
Q[1]
Q[2]
Q[3]
Q[N]
Flow identification
Input links
Output link
Flow identification
Identifies the TCP/UDP flow
to which the arriving packet
belongs
Buffer acceptance
accepts or rejects
incoming packets
Queuing strategy
Logical organization of the
router's buffers
Scheduler
Chooses the packet to
be transmitted first on
the output link](https://image.slidesharecdn.com/4-transport-sharing-210507084522/85/Part-4-reliable-transport-and-sharing-resources-50-320.jpg)
![Router output port
Q[1]
Q[2]
Q[3]
Q[N]
Flow identification
Input links
Output link
Flow identification
Identifies the TCP/UDP flow
to which the arriving packet
belongs
Buffer acceptance
accepts or rejects
incoming packets
Queuing strategy
Logical organization of the
router's buffers
Scheduler
Chooses the packet to
be transmitted first on
the output link](https://image.slidesharecdn.com/4-transport-sharing-210507084522/75/Part-4-reliable-transport-and-sharing-resources-50-2048.jpg)
Uploaded byOlivier Bonaventure
Part 4 : reliable transport and sharing resources
The document discusses concepts of reliable transport, including multiplexing, connection establishment, data transfer, connection release, and resource sharing. It outlines the importance of managing connections through a connection table and describes techniques for dealing with delayed segments and congestion control. The document also emphasizes max-min fairness in bandwidth allocation and methods to detect and manage congestion in network systems.
More Related Content
Similar to Part 4 : reliable transport and sharing resources
Part 4 : reliable transport and sharing resources
- 1. Week 4 Reliable transport Sharingresources
- 2. Agenda • Reliable transport •Multiplexing • Connection establishment • Data transfer • Connection release • Sharing resources
- 3. Multiplexing Server Client Source port :1234 Destination port: 5678 Request Response Source port : 5678 Destination port: 1234 Source Addr: Client Dest. Addr: Server Source Addr: Server Dest. Addr: Client
- 4. Connection table • Aserver interacts with many different clients and maintains a connection table • This table contains one entry per active connection with its state • Each connection is identified by 4 infos • Client address and port • Server address and port
- 5. Agenda • Reliable transport •Multiplexing • Connection establishment • Data transfer • Connection release • Sharing resources • Security principles
- 6. Coping with delayed segments •How to deal with delayed segments ? • Network level guarantee • No packet will survive more than MSL seconds inside the network • Transport entities use on a local clock to detect duplicated connection establishment requests
- 7. Three way handshake CR (seq=x) CA(seq=y, ack=x) CA (seq=x, ack=y) Sequence number x read from local transport clock Local state : Connection to B : - Wait for ack for CR (x) - Start retransmission timer Sequence number y read from local transport clock CA sent to ack CR Local state : Connection to A : - Wait for ack for CA(y) Received CA acknowledges CR Send CA to ack received CA Local state : Connection to B : - established - current_seq = x The sequence numbers used for the data segments will start from x The sequence numbers used for the data segments will start from y D(x) D(y) Local state : Connection to A : - established - current_seq=y Connection established Connection established Host A Host B
- 8. Client FSM Connect.req Send CR Timerexpires Send CR Recvd REJECT Disconnect.ind Recvd(CA) Connect.conf / Send CA Recvd REJECT Disconnect.ind In any state, a Disconnect.ind (abrupt) -> Send REJECT and transition to red state
- 9. Server FSM Recvd (CR) Connect.ind Connect.resp SendCA Recvd REJECT Disconnect.ind Recvd REJECT Disconnect.ind Recvd CA In any state, a Disconnect.ind (abrupt) -> Send REJECT and transition to red state Recvd CR Recvd CR Send CA
- 10. Three way handshake (2) CA(seq=y, ack=z) CR (seq=z) REJECT (ack=y) Connection cancelled No connection is established Host A Host B Sequence number y read from local transport clock Acknowledges CR segment Local state : Connection to A : - Wait for ack for CA(y) Local state : No connection to B Send REJECT to cancel connection establishment
- 11. Three way handshake (3) CR(seq=z) Current state does not contain a CR with seq=x REJECT (ack=y) Connection established CR (seq=z) Retransmission timer expires CA (seq=w, ack=z) CA (seq=z, ack=w) CA (seq=y, ack=x) Sequence number z read from local transport clock Local state : Connection to B : - Wait for ack for CR (z) - Start retransmission timer Host A Host B Current state does not contain a segment with seq=y REJECT ignored Sequence number w read from local transport clock CA sent to ack CR Local state : Connection to A : - Wait for ack for CA(w) Received CA acknowledges CR Send CA to ack received CA Local state : Connection to B : - established - current_seq = z
- 12. Three way handshake (4) InvalidCA received from A Send REJECT CA (seq=w, ack=z) CR (seq=z) CA (seq=z, ack=y) REJECT (ack=w) REJECT (ack=z) Sequence number w read from local transport clock Acknowledges CR segment Local state : Connection to A : - Wait for ack for CA(w) Current state does not contain a CR with seq=z Host A Host B No connection is established
- 13. Agenda • Reliable transport •Multiplexing • Connection establishment • Data transfer • Connection release • Sharing resources • Security principles
- 14. Reliable data transfer •What are the main differences with the reliable protocols of the datalink layer ? • Segments can be reordered • Buffers can change dynamically • Bytestream service
- 15. Buffer management • Atransport entity serves a variable number of applications with a limited buffer • The buffer/window allocated to a given connection may need to change dynamically as connections start and stop
- 16. Buffer management A B Data.req(a) Data.ind(a) D(0,a) C(OK,0,w=1) C(OK,0,w=3) Data.req(c) D(2,c) 2 new buffers become available Data.req(b) Data.ind(b) D(1,b) 0 1 2 3 0 1 2 3 0 1 2 3 C(OK,1,w=3) Rwin=1 0 1 2 3 Swin=3, rwin=1 0 1 2 3 Swin=3, rwin=1 0 1 2 3 Swin=3, rwin=1 0 1 2 3 Swin=3, rwin=1 Swin=3, rwin=3 0 1 2 3 Data.req(d) D(3,d) 0 1 2 3
- 17. Buffer management A B Data.req(b) Data.ind(a) Data.req(a) D(0,a) C(OK,0,w=0) 2 new buffers are available 0 1 2 3 Rwin=1 0 1 2 3 Swin=3, rwin=1 0 1 2 3 Swin=3, rwin=1 0 1 2 3 Receiver cannot handle segment immediately C(OK,0,w=3) Lost segment 0 1 2 3 Swin=3, rwin=0 Window blocked No transmission possible Waits for control segment Waits for data segment How to recover from deadlock ? Persistence timer on receiver, resend control segment after timer expiration
- 18. Delayed segments A B D(1,b) Timerexpiration Retransmission D(1,b) D(3,d) C(OK,0) C(OK,0) C(OK,3) C(OK,0) C(OK,1) D(0,e) Data.ind(e) Data.ind(b) !!!!!!!!!!!! D(0,a) Data.req(a) Data.ind(a) Data.ind(b) Data.ind(e)
- 19. Delayed segments • Howto deal with them ? • Packets cannot live more than MSL seconds inside the network • Only one segment carrying sequence number x can be transmitted during MSL seconds • upper bound on maximum throughput
- 20. Bidirectional transfer • Howto efficiently carry data in both directions ?
- 21. Piggybacking A B Data.req(a) Data.ind(a) D(0,0,a) Data.req(b) D(1,0,b) Error Discarded Data.req(c) D(2,0,c) Segment ->buffer Retransmission Data.ind(b) D(1,5,b) Data.ind(c) Data.req(d) D(3,6,d) Data.ind(d) C(OK,2) C(OK,3) Data.req(x) D(5,0,w) Data.req(w) Data.ind(w) D(6,0,x) Data.ind(x) D(5,0,w) acks D(0,0,a)
- 22. Bytestream • How toprovide a bytestream service ?
- 23. Byte stream service(2) A B Data.req(ijkl) Data.req(mnop) Data.req(abcdef) Data.ind(ab) D(0,ab) C(OK,1) C(OK,1) D(2,cd) Lost segment D(4,ef) Placed in buffer Data.ind(cdef) D(2,cd) Expiration timer Retransmission D(6,ijklmnop) Data.ind(ijklmnop) C(OK,5) C(OK,13)
- 24. Agenda • Reliable transport •Multiplexing • Connection establishment • Data transfer • Connection release • Sharing resources • Security principles
- 25. Connection release • Gracefulrelease • Data transfer finished and connection must be terminated • Abrupt release • Something went wrong and the connection must be closed immediately • Data can be lost !
- 26. Graceful release D(‘a’,1233) DISCONNECT.req (A-B) DISCONNECT.ind(A-B) ACK,1234 DISCONNECT.conf(A-B) ACK,4567 DISCONNECT.conf(A-B) DISCONNECT.req(B-A) DISCONNECT.ind(B-A) DR(B-A,4567) Outgoingconnection (A->B) closed Incoming connection (A->B) closed Incoming connection (B->A) closed Outgoing connection (B->A) closed DR(A-B,1234) DATA.ind(‘a’)
- 27. Graceful release (2) D(‘a’,1230) DISCONNECT.req(A-B) DISCONNECT.ind(A-B) ACK(1230) DISCONNECT.conf(A-B) Outgoing connection (A->B) closed Incoming connection (A->B) closed DR(A-B,1234) DATA.ind(‘a’) D(‘bcd’,1231) ACK(1230) DATA.ind(‘bcd’) ACK(1234)
- 28. Abrupt release CR (seq=z) CA(seq=w, ack=z) CA (seq=z, ack=w) D Data.req() Data.ind() Disc.req() D Data.req() DR Disc.req() Connection closed Connection closed This segment will not be delivered !
- 29. Agenda • Reliable transport •Sharing resources • Which resources need to be shared • Medium Access Control • Congestion Control
- 30. Network resources • Whatare the resources that are shared by multiple users inside a network ?
- 31. Sharing bandwidth • Severalnodes on a single link
- 32.
- 33. Agenda • Reliable transport •Sharing resources • Which resources need to be shared • Medium Access Control • Congestion Control
- 34. How to shareaccess to a link ? • Deterministic solutions • Probabilistic solutions
- 35.
- 36.
- 37. Agenda • Reliable transport •Sharing resources • Which resources need to be shared • Medium Access Control • Congestion Control
- 38. Adapting to different bandwidth Whatis the round-trip-time in milliseconds this network if A sends 10,000 bits segments, all links are 1 m long and B returns 100 bits acks ? If A sends four 10,000 bits segments in a row, after how many msec will it receive the last
- 39.
- 40. The congestion problem • IfR1 has a buffer of five 10,000 bits long packets, how many packets will be dropped if A and B send a burst of ten packets at line rate ?
- 41. Fairness • What isthe final objective of congestion control ? • On a single link • Fair share • In a large network • Max-min fairness
- 42. Max-min fairness • amax-min allocation of bandwidth is an allocation of bandwidth which maximises the allocation of bandwidth to the sources receiving the smallest allocation • a max-min fair allocation is such that in order to increase the bandwidth allocated to one source, it is necessary to decrease the bandwidth allocated to another source which already receives a lower allocation
- 43. Max-min fairness • amax-min allocation maximises the allocation of bandwidth to the sources receiving the smallest allocation • Property • to increase the bandwidth allocated to one source, it is necessary to decrease the bandwidth allocated to another source which already receives a lower allocation
- 44. Example • Max-min fairbandwidth allocation R R R R R 1000 Mbps 100 Mbps S1 S2 S3 S4 S5 S6 S7 S8 D1 D3 D4 D2 D5 D6 D7 D8 Link1 Link2 Link3 Link4
- 45. How to achieve max-minfairness ? • Two possible approaches • Modify the routers to reach max-min fairness • Modify the endhosts to reach max-min fairness
- 46. How to detect congestion? • Packet losses • Increased delay • Routers add information to packets • Forward binary feedback • Backward binary feedback • Rate feedback
- 47.
- 48. Congestion control • AdditiveIncrease / Multiplicative Decrease # Additive Increase Multiplicative Decrease if congestion : rate=rate*betaC # MD, betaC<1 else rate=rate+alphaN # AI
- 49. Simple Router output port Buffer Inputlinks Output link Buffer acceptance accepts or rejects incoming packets
- 50. Router output port Q[1] Q[2] Q[3] Q[N] Flowidentification Input links Output link Flow identification Identifies the TCP/UDP flow to which the arriving packet belongs Buffer acceptance accepts or rejects incoming packets Queuing strategy Logical organization of the router's buffers Scheduler Chooses the packet to be transmitted first on the output link
- 51. Round robin Flow 2 Flow1 Flow 3 Flow 4 Flow 5 Flow 1 Flow 2 Flow 3 Flow N Scheduler : F1 F2 F3 F4 FN
- 52. Round robin • Ifthe RR scheduler starts with queue1 and serves them in numerical order, what is the ordering of the packets on the output link ? A B D E F G Queue 1 Queue 2 Queue 3
- 53. Example F1 F2 F3 Flow2 (L=2) Flow1(L=1) Flow3 (L=1)F1 F2 F3 Flow1 and Flow3 send packets of size 1 Flow 2 sends packets of size 2
- 54. Round-Robin • Advantage • Canprovide fairness independently of the characteristics of the flows • Drawback • Difficult to scale to a very large number of flows
Editor's Notes
- #9 In this example, the duplicate CR is likely to be a previous retransmission of the CR that was delayed in the network.
- #23 MSL means Maximum Segment Lifetime
- #53 How to determine a max-min fair bandwidth allocation for a given network ? Algorithm [Bertsekas & Gallager, Data Networks, 2nd edition, Prentice Hall 1992] First start with an allocation of 0 Mbps for each source Then equally increment the allocation to each source until one link becomes saturated. At this point, each source which uses the saturated link receives an allocation equal to the bandwidth of this saturated link divided by the number of sources using this bottleneck link. Next, the allocation of all the sources which do not use a saturated link is equally incremented until another link becomes saturated. The algorithm continues from step to step, always incrementing the allocation of the sources which do not use a saturated link, until all sources use at least one of the saturated links.