Communication Protocols and Internet Architectures Harvard University. Lecture #7. Instructor: Len Evenchik Lecture Agenda

Transcription

1 Communication Protocols and Internet Architectures Harvard University Lecture #7 Instructor: Len Evenchik Lecture Agenda Course Logistics Q&A and Topics from Last Week VLANs and Wireless UDP TCP TCP Congestion Control Connection Management One Minute Wrap-Up (C) L. Evenchik 1

2 Course Logistics Q&A and Some Things from Previous Lectures (C) L. Evenchik 2

3 IPv6 Fundamentals IPv6 Addressing RFC 4291 (Feb. 2006) but we ll check IANA for the Details, Section 2.4 The type of an IPv6 address is identified by the high-order bits of the address, as follows: Address type Binary prefix IPv6 notation Unspecified (128 bits) ::/128 Loopback (128 bits) ::1/128 Multicast FF00::/8 Link-Local unicast FE80::/10 Global Unicast (everything else) Global Unicast Addresses The general format for IPv6 Global Unicast addresses is: n bits m bits 128-n-m bits global routing prefix subnet ID interface ID (C) L. Evenchik 3

4 Example of IPv6 Address fas% ifconfig eth0 Link encap:ethernet HWaddr 00:0b:cd:82:51:ea inet addr: Bcast: Mask: inet6 addr: fe80::20b:cdff:fe82:51ea/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets: err:3025 dropp:0 overruns:0 TX packets: err:0 dropp:0 overruns:0 carrier:0. lo Link encap:local Loopback inet addr: Mask: inet6 addr: ::1/128 Scope:Host UP LOOPBACK RUNNING MTU:16436 Metric:1 RX packets:8708 errors:0 dropped:0 overruns:0 frame:0 TX packets:8708 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes: TX bytes: IPv6 RFC 4291 and RFC3581 RFC 4291, Global Unicast Addresses The general format for IPv6 Global Unicast addresses is: n bits m bits 128-n-m bits global routing prefix subnet ID interface ID IANA Assigned Unicast 3 45 bits 16 bits 64 bits global routing subnet ID interface ID IANA Global Routing 3 13bits 16bits 16 bits 16 bits 64 bits RIR AS Customer Subnet ID interface ID (C) L. Evenchik 4

5 VLANs (part 2) Ethernet Switch and VLAN Topology is for /23 Router is for /23 is for /23 Switch Switch Switch Note that endpoints rarely know if they are on a VLAN (C) L. Evenchik 5

6 Simplistic Router and VLAN Topology Router One router port per subnet. is for /23 is for /23 is for /23 Switch Switch Switch Note that endpoints rarely know if they are on a VLAN Router Support for VLAN Tags is for /23 Router is for /23 is for /23 Switch Switch Switch (C) L. Evenchik 6

7 Wireless LANs and VLANs Distribution System and Network Router Access Point Access Point Infrastructure Network (C) L. Evenchik 7

8 Supporting Wireless Access Points (WAP) with VLANs Router WAP is for /23 is for /23 is for /23 is for /23 for all wireless users WAP WAP WAP Manager Note that endpoints rarely know if they are on a VLAN Transport Layer Protocols (C) L. Evenchik 8

9 Transport Service (Source: unknown, but this comes from the days of the 7-layer model) (Source: unknown) SP3: Service, Purpose, Packet and Procedure A Framework for Understanding and Describing Protocols (C) L. Evenchik 9

10 Service SP3 Framework for Understanding Protocols The Service is a description of what the protocol does, not how it is done. This should be a few sentences long. Purpose The Purpose describes the specific functionality that the protocol provides and how it is accomplished. Examples are flow control, error detection, error correction, etc. Packets The Packet layout determines how the various bits and fields within the packet are defined, assembled and used. Procedures The Procedures describe the various packet exchanges and the reason for each exchange. SP3 - Service What type of Service does the protocol provide? The Service is a short description of what the protocol does, not how it is done. For example, a protocol can provide any of the following services Reliable service, including sequenced delivery. This is commonly known as connection-oriented service. (TCP) Reliable service, but not with sequenced delivery. Unreliable service. This is known as connectionless or datagram service. (IP, UDP) Unreliable service, but with the sequenced delivery of messages. (RTP is an example of this.) Are there more? (C) L. Evenchik 10

11 The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the SP3 Purpose What specific functionality does the protocol provide and how does it do it? Addressing Multiplexing Sequencing Error control (two parts to this) Flow control Option negotiation Encryption Fragmentation and reassembly plus others... SP3 Packet What do the fields in the Packet header do? Vers IHL IP Precedence DSCP Identification D M F F Total length Fragment offset Time to live Protocol Header checksum Source address Destination address Options (C) L. Evenchik 11

12 SP3 Procedures What Procedures does a Protocol Use? Connection Establishment or Initialization Data Transfer Connection Release or Disconnect Error Handling (of many different types) User Datagram Protocol (C) L. Evenchik 12

13 UDP Protocol User Datagram Protocol is a simple transport layer protocol UDP provides a Datagram delivery service UDP checksum provides end-to-end error detection and this is very important, but it is also important to note that it violates the concept of strict protocol layering. Multiplexing feature is provided via abstract destination points known as protocol Ports Port assignment mechanisms include use of well-known ports as well as dynamic binding Where would you look to find the listing of port number assignments? UDP Packet Format 32 bits Source Port number UDP Length Destination Port number UDP Checksum Data (Payload) (C) L. Evenchik 13

14 The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, Combined IP/UDP Packet Layout 32 bits Version IHL DSCP Total length Identification D F M F Fragment offset Time to live Protocol Header checksum IP Source address IP Destination address IP Options (if present) UDP Header - first four bytes UDP Header - second four bytes UDP Payload Transmission Control Protocol (TCP) (C) L. Evenchik 14

15 Description of TCP (from RFC 793) TCP provides a reliable securable logical circuit or connection service between pairs of processes.. To do this using less reliable communication systems requires... Basic Data Transfer Connections Reliability Flow Control Multiplexing Precedence and Security Remember, TCP is a communication protocol, not a specific piece of software Source, RFC 793 RFC 4614 (2006) TCP is Complicated and Implementation Specific RFC 4614 was written to provide a roadmap to TCP. It describes the most important RFCs and just as important, it explains what prior work is no longer relevant. It is important to note that it was written in See RFC This roadmap is divided into four main section: Basic (core) Functionality Recommended Enhancements Experimental Extensions Historic Extensions (C) L. Evenchik 15

16 TCP Packet Header (RFC 793, 1981) 32 bits Source port Destination port Sequence number TCP header length U R G A C K Acknowledgement P S H R S T S Y N F I N Window TCP header Checksum Urgent pointer Options (0 or more 32 bit words) Data (payload) TCP Packet Header (Today, 2011) 32 bits Source port Destination port Sequence number Per RFC 3168 TCP header length C W C E R E U R G A C K Acknowledgement P S H R S T S Y N F I N Window TCP header Checksum Urgent pointer Options (0 or more 32 bit words) Data (payload) (C) L. Evenchik 16

17 The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, Combined IP/TCP Packet Layout 32 bits Version IHL Type of service Total length Identification D F M F Fragment offset Time to live Protocol Header checksum IP Source address IP Destination address IP Options (if present) TCP Header - first four bytes TCP Header - other header bytes and option fields (if present) TCP Payload Meaning of Flag Bits in TCP Header URG - urgent pointer field is valid ACK - acknowledgement field is valid PSH - this segment requests a push RST - reset the connection SYN - synchronize sequence numbers FIN - sender has reached end of its byte stream CWR and ECE - provides info on congestion (C) L. Evenchik 17

18 TCP Connection Establishment Send SYN, seq=x Network Receive SYN segment Send SYN seq=y, ACK x+1 Receive SYN + ACK segment Send ACK y+1 Receive ACK segment Events at site 1 Events at site 2 TCP State Transition Diagram, Client/Server - abridged CLOSED SYN / SYN + ACK Passive Open LISTEN Active Open / SYN SYN RCVD Active Close / FIN FIN WAIT 1 (Active Close State) ACK / nothing ESTAB LISHED (Data Transfer State) FIN / ACK SYN SENT SYN + ACK / ACK CLOSE WAIT (Passive Close State) (C) L. Evenchik 18

19 State Transition Diagram State 1 Event / Action State 2 TCP State Transition Diagram, Client Side - abridged CLOSED Active Open / SYN SYN SENT (C) L. Evenchik 19

20 TCP State Transition Diagram, Client Side - abridged CLOSED LISTEN Active Open / SYN SYN RCVD SYN SENT SYN + ACK / ACK Active Close / FIN FIN WAIT 1 (Active Close State) ESTAB LISHED (Data Transfer State) FIN / ACK CLOSE WAIT (Passive Close State) TCP State Transition Diagram- Server Side (abridged) CLOSED Passive Open LISTEN (C) L. Evenchik 20

21 TCP State Transition Diagram- Server Side (abridged) CLOSED Passive Open SYN / SYN ACK LISTEN SYN RCVD ACK / nothing SYN SENT ESTAB LISHED (Data Transfer State) FIN / ACK CLOSE WAIT TCP State Transition Diagram, Client/Server - abridged CLOSED SYN / SYN + ACK Passive Open LISTEN Active Open / SYN SYN RCVD Active Close / FIN FIN WAIT 1 (Active Close State) ACK / nothing ESTAB LISHED (Data Transfer State) FIN / ACK SYN SENT SYN + ACK / ACK CLOSE WAIT (Passive Close State) (C) L. Evenchik 21

22 TCP Sequencing, Flow Control and Segmentation Some General Characteristics A TCP data stream is a sequence of octets (bytes). This is different than the link layer protocols we have studied. TCP uses a sliding window and this window identifies bytes not packets TCP does not know anything about the bytes it is sending. (As you would expect with transport protocols.) The size of the Flow Control window changes dynamically over time Retransmission timeout also changes dynamically (based on RTT) over time TCP Byte Stream Functionality A TCP data stream is a sequence of octets (bytes.) Data is acknowledged at the byte level, not the segment, packet or frame level. TCP connections are full duplex, point-to-point. Data flows simultaneously but independently in each direction. When data is actually sent on the wire is at the discretion of the sending TCP software. When data is given to the application is at the discretion of the receiving TCP. The intent of the PUSH bit is to modify when TCP sends or processes the received data (as we will see.) The receiving application can be told about important data located in the byte stream via the use of the URGENT bit. Note this is a hint not a command. (C) L. Evenchik 22

23 TCP Segmentation TCP sends data in segments. The default segment size is 536 bytes of data (for IPv4.) TCP maximum segment size (MSS) is negotiated during connection setup. TCP decides how big the segments can be, not the application software. The maximum segment size (MSS) is dependent upon the size of MTU (Maximum Transmission Unit.) What does this violate? TCP segments can arrive out of order. Lets think about this for a minute. TCP Flow Control TCP uses a byte oriented, variable size, sliding window. Window Advertisements and Acknowledgements are independent. Sending an ACK = X means that X-1 has been received correctly. Window size changes dynamically. How is this done? A window advertisement of zero is legitimate (but a 1 byte segment can always be sent.) Remember, TCP is full duplex so flow control must be done in both directions simultaneously. In TCP, congestion control is not the same a flow control. More on this in a minute. (C) L. Evenchik 23

24 Meaning of Flag Bits in TCP Header URG - urgent pointer field is valid ACK - acknowledgement field is valid PSH - this segment requests a push RST - reset the connection SYN - synchronize sequence numbers FIN - sender has reached end of its byte stream CWR and ECE - provides info on congestion An Example of TCP Sequencing, Flow Control and Segmentation seq=x, ack=y, win=a data=z bytes seq=y, ack=x+z win=b, data=zz Parameters of note: Sequence number (seq) Acknowledgement (ack) Number of bytes sent (data) Window Advertisement (win) (C) L. Evenchik 24

25 Wrap Around for 32 bit Sequence Number This table shows for example that the 32-bit sequence number wraps around in 6 minutes for a 100BaseT link A TCP option can extend the sequence number Source and Copyright of table is Computer Networks by Davie and Peterson Keeping the Pipe Full This table assumes a 100 msec RTT time The Delay X Bandwidth product dictates the necessary size for the Advertised Window A TCP option allows TCP to increase the advertised window size Source and Copyright of table is Computer Networks by Davie and Peterson (C) L. Evenchik 25

26 TCP Congestion Contol TCP Congestion Contol Goodput Offered Load (C) L. Evenchik 26

27 TCP Congestion Control (1) The assumption today is that networks are reliable (but of course, not perfect) and therefore packets are dropped due to network congestion. A second assumption is that network congestion occurs at routers (or other network devices) due to bursts of traffic. Congestion control and flow control are very different. (A classic drawing represents this with pipes and overflowing buckets of water.) Congestion control and Slow Start were not part of RFC 793. Van Jacobson designed them in TCP Slow Start addresses congestion control, not flow control. TCP Congestion Control (2) Note that Slow Start is a misnomer. It is actually exponential growth. TCP uses four intertwined Congestion Control algorithms and mechanisms (RFC 5681) slow start congestion avoidance fast retransmit fast recovery. TCP Congestion Avoidance is additive-increase, multiplicative-decrease (AIMD) Finally, the first assumption we make for congestion control is that networks are reliable, and this does not apply to wireless networks. What does this mean for real world implementations? (C) L. Evenchik 27

28 TCP Slow Start (1) Timeout (Packet Lost) CWND Number of TCP Segments Sent at One Time Slow Start SSTHRESH Congestion Avoidance Phase Time Parameters of note: Receiver s Advertised Window Size for Flow Control Initial Window (IW) Congestion Window Size (CWND) Slow Start Threshold (SSTHRESH) TCP Slow Start (2) The Slow Start Threshold is decreased after a loss This graph shows one segment sent at start of Slow Start phase Source and Copyright of graph is Computer Networks by Tanenbaum (C) L. Evenchik 28

29 Packet Flow in Slow Start 1 Segment ACK! 2 Segments 4 Segments 8 Segments 16 Segments would be sent! next 8 ACKS would be! sent back! Additive Increase in Congestion Avoidance 4 Segments 5 Segments 6 Segments would be sent next 5 ACKS would be! sent back! (C) L. Evenchik 29

30 TCP Slow Start (3) Fast Recovery and Fast Retransmit Algorithm Fast Retransmit occurs after three (3) duplicate Acks are received Fast Recovery means that CWND is not reduced to IW Source and Copyright of graph is unknown Fast Retransmit via Duplicate ACK Packet 1 Packet 2 Packet 3 Packet 4 Packet 5 X ACK 2 ACK 2 ACK 2 ACK 2 Resend Packet 2 ACK 6 Packet 6 Note that this is a simplified diagram and that TCP Acks and Sequence numbers are based on the byte, not the segment or packet (C) L. Evenchik 30

31 Additional TCP Features TCP has Selective Acknowledgements (SACK) TCP connections can be aborted immediately by RST bit TCP uses a modified three way handshake to close connections TCP ports identify applications at each end of the connection via Port numbers TCP uses both static and dynamic port binding TCP checksum uses the same pseudo header as UDP Some TCP Implementation Details Implementations are known as Tahoe, Reno, Vegas, etc. RTT Calculation (Jacobson 1988) Karn s Algorithm: don t update RTT on retransmitted segments. Nagle s algorithm (1984): addresses problem of sending one byte at a time Silly Window Syndrome (1982) The congestion control RFCs we have talked about The number of segments that are initially sent during Slow Start continues to change (RFC 3390 and recent ID) (C) L. Evenchik 31

32 Connection Management Application Layer Connection Management Network SERVER SERVER How does a system keep track of all of its application layer connections? Can we see the details of these connections? (C) L. Evenchik 32

33 Application Layer Software Schematic Appl. 1 Appl. 2 Appl. 3 (TCP) Appl. 3 (UDP) Appl. 4 TCP IP Layer UDP Ethernet (Layers 1 and 2) TCP Packet Header 32 bits Source port Destination port Sequence number TCP header length U R G A C K Acknowledgement P S H R S T S Y N F I N Window TCP header Checksum Urgent pointer Options (0 or more 32 bit words) Data (payload) (C) L. Evenchik 33

34 Some Well Known TCP Port Numbers 20,21 FTP File transfer 22 SSH Secure Shell 23 Telnet Remote login, not encrypted 25 SMTP 80 HTTP world wide web 110 POP3 Remote access 443 HTTPS Encrypted web traffic 1720 H.323 Video conferencing 5060 SIP Session Initiation Protocol (C) L. Evenchik 34

35 Connection Management Table Network SERVER SERVER Connection ID # netstat -an tcp ESTABLISHED tcp ESTABLISHED tcp ESTABLISHED tcp FIN_WAIT_2 tcp ESTABLISHED tcp ESTABLISHED tcp TIME_WAIT tcp *.80 *.* LISTEN tcp *.443 *.* LISTEN tcp *.22 *.* LISTEN tcp.. (C) L. Evenchik 35

36 One Minute Wrap-Up Please do this Wrap-Up at the end of each lecture. Tear off this page and hand it in as you leave, or fill out the form on the website. Do not sign your name. The form on the website is also anonymous. Please answer three questions: What is your grand Aha for today s class? What concept did you find most confusing in today s class? What questions should I address next time Thank you! (C) L. Evenchik 36