Week9 lec1
- 1. Chapter 3 Transport Layer Computer Networking: A Top Down Approach 4th edition. Jim Kurose, Keith Ross Addison-Wesley, July 2007.
- 2. Last Lecture TCP Congestion Control
- 3. Today’s Lecture Retransmission Timer in TCP Chapter 4 Network Layer
- 4. TCP Congestion Control: Details Congestion Window Imposes a constraint on the rate at which a TCP sender can send traffic into the network The amount of unacknowledged data at a sender may not exceed the minimum of CongWin and RcvWindow LastByteSent-LastByteAcked min{CongWin, RcvWindow} CongWin is dynamic, function of perceived network congestion TCP takes arrival of ACKs as successful delivery Increases Congestion Window Self Clocking TCP uses ACKs to trigger its increase in congestion window size
- 5. TCP Congestion Control Algorithm: Additive Increase Multiplicative Decrease (AIMD) Additive Increase Increase CongWin by 1 MSS until loss detected Multiplicative Decrease Cut CongWin in half after loss CongWin is 20 Kbytes, cut it to half after loss Continue to drop but not allowed to drop below 1 MSS congestion window 24 Kbytes 16 Kbytes 8 Kbytes time
- 6. TCP Slow Start AIMD can take a long time Host A Host B RTT to ramp up to full capacity from scratch When connection begins, CongWin = 1 MSS Increase rate exponentially until first loss event: Double CongWin every RTT Done by incrementing CongWin for every ACK received Summary: Initial rate is slow but ramps up exponentially fast time
- 7. Refinement: Inferring Loss 3 Duplicate ACKs Indicates network capable of delivering some segments. Timeout Indicates a “more alarming” congestion scenario.
- 8. Refinement: Inferring Loss After timeout event CongWin instead set to 1 MSS Window then grows exponentially (enters slow start) Till it reaches one half of the value it had before the timeout Window then grows linearly (congestion avoidance) Threshold Value Determines the window size at which slow start will end After three duplicate ACKs: CongWin is cut in half Window then grows linearly Fast Retransmit and Fast Recovery
- 9. Refinement: Inferring Loss TCP Tahoe For either events Cuts congestion window size to 1 MSS and enters slow start TCP Reno Uses Fast Retransmit and Fast Recovery for three Dupicate ACKs. Same as Tahoe for Timeout events.
- 11. TCP Round Trip Time and Timeout Retransmission Timer larger than the RTT How much larger? How would RTT be calculated? Sample RTT Sample RTT fluctuates from segment to segment Congestion in routers and varying load on end systems TCP never computes a Sample RTT for retransmitted segments. (Why?) EstimatedRTT TCP maintains an average of Sample RTT (RFC 2988) EstimatedRTT = (1- )*EstimatedRTT + *SampleRTT The new value of Estimated RTT is a weighted combination of the previous value of Estimated RTT plus new value of Sample RTT.
- 12. TCP Round Trip Time and Timeout The recommended value of becomes is 0.125 in which case the formula EstimatedRTT = 0.875*EstimatedRTT + 0.125*SampleRTT RTT: gaia.cs.umass.edu to fantasia.eurecom.fr 350 RTT (milliseconds) 300 250 200 150 100 1 8 15 22 29 36 43 50 57 64 71 time (seconnds) SampleRTT Estimated RTT 78 85 92 99 106
- 13. Setting the Timeout Interval • The timeout interval should be greater than or equal to EstimatedRTT • Not too large, delay to retransmit packet • Desirable to set the timeout interval plus some “safety margin.” • Estimate of how much SampleRTT deviates from EstimatedRTT DevRTT = (1- )*DevRTT + *|SampleRTT-EstimatedRTT| ( = 0.25) Then set timeout interval: TimeoutInterval = EstimatedRTT + 4*DevRTT For more details see RFC 2988 V. Jacobson, “Congestion Avoidance and Control”, in Proc. ACM SIGCOMM, pp. 314329, 1988.
- 14. TCP Connection Management • How TCP connection is established? – Three Way handshake? • How TCP connection is closed? – Four Way handshake? Reading Assignment
- 15. TCP Connection Management TCP server lifecycle TCP client lifecycle •Reading Assignment
- 16. Chapter 4 Network Layer Computer Networking: A Top Down Approach 4th edition. Jim Kurose, Keith Ross Addison-Wesley, July 2007.
- 17. Network layer Transport segment from sending to receiving host On sending side encapsulates segments into datagrams On receiving side, delivers segments to transport layer Network layer protocols in routers Router examines header fields in all IP datagrams passing through it application transport network data link physical network data link physical network data link physical network data link physical network data link physical network data link physical network network data link data link physical physical network data link physical network data link physical network data link physical network data link physical application transport network data link physical
- 18. Two Key Network-Layer Functions Forwarding: Move packets from router’s input to appropriate router’s output Local action of transferring a packet Routing: Determine route taken by packets from source to destination. Network wide process that determines the end to end paths The algorithms that calculate these paths are referred to as routing algorithms
- 19. Routing and Forwarding routing algorithm Forwarding Table o Router Forwards a packet by examining the destination address of the packet local forwarding table header value output link o Routing Algorithm value in arriving determines the values packet’s header that are inserted in the routers forwarding tables 0100 0101 0111 1001 0111 3 2 2 1 1 3 2
- 20. Network Layer Host, router network layer functions: Transport layer: TCP, UDP Network layer IP protocol •addressing conventions •datagram format •packet handling conventions Routing protocols •path selection •RIP, OSPF, BGP forwarding table ICMP protocol •error reporting •router “signaling” Link layer physical layer
- 21. IP Address • An IP address is a 32-bit address. • The IP addresses are unique. • The address space of IPv4 is 232 or 4,294,967,296.(more than 4 billion) • Represented in binary or decimal
- 22. CLASSFUL ADDRESSING • IP addressing when started used the concept of classes. • This architecture is called classful addressing. • Organizations can be of different sizes • Require varying numbers of IP addresses on the Internet. • A system was devised whereby the IP address space would be divided into classes. • Each class occupies some part of the whole address space • Some classes are for • Large networks on the Internet • Smaller organizations • Some classes are reserved for special purposes.
- 23. IP Address Classes • There are five classes in the “classful” system, which are given letters A through E. • The first few bits in binary notation can immediately tell us the class of the address
- 25. Netid and Hostid •An IP address is divided into Netid and Hostid •Network Identifier (Netid): Few bits starting from the left-most bit are used to identify the network where the host is located. •Host Identifier (Hostid): The remainder of the bits are used to identify the host on the network. •These parts are of varying lengths depending on the class of the address.
- 26. Netid and Hostid
- 27. Class A • Class A is divided into 128 blocks (27) – Each block having different Netid • First block covers addresses from – 0.0.0.0 to 0.255.255.255 (netid 0) • Second block covers addresses from – 1.0.0.0 to 1.255.255.255 (netid 1) • The last block covers addresses from – 127.0.0.0 to 127.255.255.255 (netid 127) • Each block in this class contains 16,777,216 addresses • Class A addresses were designed for large organizations with large number of hosts attached to their network – 16,777,216 too large, millions of class A addresses are wasted
- 28. Class B • Class B is divided into 16384 blocks (214) – Each block having different netid • First block covers addresses from – 128.0.0.0 to 128.0.255.255 (netid 128.0) • The last block covers addresses from – 191.255.0.0 to 191.255.255.255 (netid 191.255) • Each block in this class contains 65,536 addresses • Class B addresses were designed for medium size organizations with thousands of hosts attached to their network – 65,536 is also too large for medium size organizations, many of class B addresses are wasted
- 29. Class C • Class C is divided into 2097152 blocks (221) – Each block having different netid • First block covers addresses from – 192.0.0.0 to 192.0.0.255 (netid 192.0.0) • The last block covers addresses from – 223.255.255.0 to 223.255.255.255 (netid 223.255.255) • Each block in this class contains 256 addresses • Class C addresses were designed for small organizations – 256 is small for most organizations
- 30. IP Address Classes • Given the network address 17.0.0.0, find the class, the netid, and the range of the addresses?.
- 31. IP Address Classes Solution • The class is A because the first byte is between 0 and 127. The block has a netid of 17. The addresses range from 17.0.0.0 to 17.255.255.255.
- 32. IP Address Classes • Given the network address 220.34.76.0, find the class, the block/netid, and the range of the addresses?.
- 33. IP Address Classes Solution • The class is C because the first byte is between 192 and 223. The block has a netid of 220.34.76. The addresses range from 220.34.76.0 to 220.34.76.255
- 34. Issues with Classful Addressing • Flexibility in Internal addressing: – Thousands of hosts connected to one large network • Inefficient Use of Address Space: – Wastage of limited IP address space. • Router Table Entries: – Increase in router table entries. Example: • Hosts needed by organization are 5000 • Hosts in B 65,536 and in C 256 – Organization with 5,000 hosts is in a dilemma – It can only choose to either waste 90% of a Class B address or use 20 different Class C networks. – Replacing with 20 C networks will increase entries in routers
- 35. Subnetting • To better meet the administrative and technical requirements of larger organizations, the “classful” IP addressing system was enhanced through a technique known as subnet addressing or subnetting. • A three-level hierarchy is created: networks, which contain subnets, each of which then has a number of hosts.
- 36. Subnetting • Subnetting adds an additional level to the hierarchy of structures used in IP addressing. • IP addresses must be broken into three elements instead of two. • Network ID is unchanged • The host ID into a subnet ID and host ID. • These subnet ID bits are used to identify each subnet within the network. • Splitting the host ID into subnet ID and host ID, we reduce the size of the host ID portion of the address • Class A networks have 24 bits to split between the subnet ID and host ID: class B networks have 16, and class C networks only 8.
- 37. IPv4 Datagram Format IP protocol version number header length (bytes) “type” of data max number remaining hops (decremented at each router) upper layer protocol to deliver payload to Type of Service field is now called Differentiated Service? See RFC 2474 Home Assignment type of ver head. len service length fragment 16-bit identifier flgs offset time to header protocol live checksum total datagram length (bytes) for fragmentation/ reassembly 32 bit source IP address 32 bit destination IP address Options (if any) data (variable length, a TCP or UDP segment) IPV5? Does it exists? e.g. timestamp, record route taken etc Find other use?
- 38. IP Fragmentation & Reassembly Network links have MTU (Maximum Transmission Unit) – Maximum length of data that can be encapsulated in a frame. o Different link between sender and destination • Use different MTUs Large IP datagram is divided (“fragmented”) o One datagram becomes several datagrams o Referred as Fragments o “Reassembled” only at final destination o IP header bits used to identify, order related fragments fragmentation: in: one large datagram out: 3 smaller datagrams reassembly
- 39. IP Fragmentation & Reassembly Identification Number, Flag and Offset fields in the datagram header Identification Number o o Sending Host stamps the datagram with an identification number Destination receives datagram o Examines the identification number Flag Bit o o All fragments set to 1 Last fragment set to 0 Offset Field o Use to specify where fragment fits in the within the original IP datagram
- 40. IP Fragmentation & Reassembly Example Datagram of 4000 bytes o 20 bytes of IP header, 3980 bytes of payload o MTU is 1500 bytes Fragment Bytes ID Offset 0 (beginning) Flag 1st Fragment 1480 bytes 777 1 (there is more) 2nd Fragment 1480 bytes 777 185 1 (there is more) 3rd Fragment 1020 bytes 777 370 0 (Last )