ICMP. Outline ICMP. ICMP oicmp is provided within IP which generates error. Internet Control Message Protocol. Ping Traceroute

Transcription

1 Outline ICMP Internet Control Message Protocol oicmp Basic o ICMP Packet oicmp Types & Code Ping Traceroute 2 3 ICMP oicmp is provided within IP which generates error messages to help the IP layers (best effort delivery) ofunction of ICMP a node recognizing a transmission problem.(ttl exceed, destination unreachable, etc.) generates ICMP messages ICMP provides some useful diagnostics about network operation (ping, traceroute) o ICMP error messages never generates due to: ICMP error messages selves Broadcast/Multicast 4 Application Transport Network Data Link Physical ICMP SMTP FTP HTTP DNS SNMP BOOTP IGMP ICMP TCP IP Underlying LAN / WAN Technology ARP RARP

2 ICMP encapsulation ICMP header type-relevant ICMP message code- more detail information checksum- covers ICMP header/data (not IP hdr) 5 6 ICMP types & code orfc 792 osamples ICMP type 0/8 echo request/reply ping sends icmp type 8 echo request to a node and expects an icmp type 0 echo reply identifier and sequence number are used to identify datagrams 7 8

3 ICMP type 3 destination unreachable ICMP type 4 source quench 9 router is unable to deliver datagram, it can return the ICMP type 3 with failure code Internet header plus 64 bits of original datagram are used to identify the datagram caused the problem. 10 router detected hosts were overloaded would send this message to hosts that were the major cause. The hosts would then reduce the rate at which subsequence messages are sent RFC recommends that router must not generate source quench, host must still accept the message but need take no action ICMP type 5 route change request ping - ICMP echo request/reply ICMP echo request ICMP echo reply use only by router to suggest a more suitable route to the originator (also called ICMP redirect) ping sends an ICMP echo request to a remote host, which then return an ICMP echo reply to the sender All TCP/IP node is supposed to implement ICMP and respond to ICMP echo 11 12

4 13 Ping Command o variants #1 send a single echo request message and wait for a reply another request is sent if the reply is not received within one seconds continue until at least one reply is received or stop after time out Router#ping Type escape sequence to abort. Sending 5, 100-byte ICMP Echos to , timeout is 2 seconds:!!!!! Success rate is 100 percent (5/5), round-trip min/avg/max = 33/44/57 ms Ping Command o variants #2 send an echo request message every seconds and records the time it takes for each reply every echo request contains a unique sequence number to match replies and request also record round-trip timing also do packet lost statistics 14 C:\>ping Pinging www-01-svr.nu.ac.th [ ] with 32 bytes of data: Request timed out. Reply from : bytes=32 time=134ms TTL=125 Reply from : bytes=32 time=5ms TTL=125 Reply from : bytes=32 time=5ms TTL=125 Ping statistics for : Packets: Sent = 4, Received = 3, Lost = 1 (25% loss), Approximate round trip times in milli-seconds: Minimum = 5ms, Maximum = 134ms, Average = 48ms 15 Ping as debugging tools oping can offer Timing information Connection reliability Destination is reachable (routable) Layer is functional, but not guaranteed telnet! 16 Ping results o no response no end node, no connection o lost packet (significant when >2-3%) transmission error on WAN/LAN, overloading bridges/routers o time acknowledge vary host/network overloading, >100 ms make telnet less acceptable ono lost and echo time is reasonably constant Connection is OK.

5 17 Traceroute ocommand to determine the active route to a destination address ohow to? send a messages to an unused porton the target host with ttl=1 router decrease ttl to 0, it has to return an ICMP time exceed message traceroute sets ttl =2 and retransmits, this time go one more hop ttl++ until messages reach the destination. the target returns an ICMP service unavailable because there is no port service. 18 Traceroute C:\>tracert 19 Tracing route to [ ] over a maximum of 30 hops: 1 <1 ms <1 ms <1 ms ms 23 ms 23 ms ms 6 ms 5 ms ms 5 ms 5 ms ms 14 ms 7 ms ms 5 ms 5 ms ms 11 ms 12 ms ms 15 ms 15 ms ms 34 ms 33 ms * * * Request timed out ms 77 ms 69 ms ms 108 ms 184 ms ms 101 ms 109 ms Trace complete. usually probes each hop 3 times a lost message or a router that doesn t respond is denote with an * User Datagram Protocol

6 Application Transport Network Data Link Physical SMTP FTP HTTP DNS SNMP BOOTP IGMP ICMP TCP IP Underlying LAN / WAN Technology ARP RARP owhat do? Connectionless communication Unreliable transport protocol / Best Effort Create Process to Process communication : Port Error control detect by checksum and silently drop A very simple protocol No flow control No acknowledgement o Process-to-process communication o communication Port / IP Process Process Process Process Application Transport Network port IP port PORT IP Application Transport Network port IP port Data link Data link IP protocol domain protocol domain Physical Physical 23 24

7 25 o Port Number IANA s range Well Known Port : 0 1,023 Registered Port : 1,024 49,151 Dynamic Port : 49, o Port Number Examples (Well Known Port) Port Protocol Description 7 Echo Echoes a received datagram back to the sender 9 Discard Discards any datagram that is received 11 Users Active users 13 Daytime Returns the date and the time 17 Quote Returns a quote of the day 19 Chargen Returns a string of characters 53 Nameserver Domain Name Service 69 TFTP Trivial File transfer Protocol 80 HTTP Hypertext Transfer Protocol 111 RPC Remote Procedure Call 123 NTP Network Time Protocol 161 SNMP Simple Network Management Protocol 162 SNMP Simple Network Management Protocol (trap) o Socket Address Address combined with IP Address and Port Number Process use to create connection Encapsulation o Ethernet Frame Datagram Ethernet header IP header header data Protocol Type in IP header :

8 Datagram Source port :16 Destination port: 16 length :16 Checksum : 16 data Source / Destination port: 16 bits port address Length: Datagram Length : Data + Header Length Checksum: one s complement of header and data including pseudo data (IP Header) Pseudo Header Checksum ouse to detect error odivided data to 16-bits segment and use 1 s complement sum Sender Calculate and Insert in Checksum field Receiver Recalculate and use to detect error wraparound sum checksum Checksum -Sender o Checksum calculation at Sender Add the pseudo header to user datagram Fill the checksum field with 0s Divide the total bits into 16-bit (2 bytes) sections If the total number of bytes is not even, add 1 byte of padding (all 0s). The padding is only for the purpose of calculating the checksum and will be discarded afterwards. Add all 16-bit sections using one s complement arithmetic Complement the result, which is a 16-bit number, and insert in the checksum field Drop the pseudo header and any added padding Deliver the user datagram to the IP layer for encapsulation

9 Checksum - Sender Checksum -Receiver o Checksum calculation at Sender Add the pseudo header to user datagram Add padding if needed Divide the total bits into 16-bit (2 bytes) sections Add all 16-bit sections using one s complement arithmetic Complement the result If the result is all 0s, drop the pseudo header and any added padding and accept the user datagram. If the result anything else, discard the user datagram 36 Operation Connectionless services No flow control and a simple error check Encapsulation and Decapsulation

10 Operation oqueuing of A client site, when a process starts, some implements create both an incoming and outgoing queueassociated with each process identified by ephemeral port number. When process terminates, the queues are destroyed. If an outgoing queue is overflow, the OS ask the client process to wait before sending any more messages. When message arrived for a client, checks created incoming queue for the port number of arrived user datagram. If there is no such incoming queue, discard the user datagram, ask the ICMPto send a port unreachable message to the server. Operation o Queuing of (example) Operation o Multiplex and Demultiplex Use of A simple request-response communication with little concern for flow and error control (not to send bulk data: ftp ) A process with internal flow and error control mechanism. Transport protocol for multicasting and broadcasting. For management process such as SNMP For some route updating protocols such as RIP (Routing Information Protocol) 39 40

11 Use of Example Application TFTP: Trivial File Transfer Protocol DNS : Domain Name System RPC, NFS: Remote procedure call, Network File System SNMP: Simple Network Management Protocol TCP Transmission Control Protocol 41 Outline TCP o TCP protocol o Sliding window otcp in action otcp Timer o Delay Acknowledgment o Nagel s Algorithm o Congestion Control Application Transport Network Data Link Physical SMTP FTP HTTP DNS SNMP BOOTP IGMP ICMP TCP IP Underlying LAN / WAN Technology ARP RARP 43 44

12 45 TCP owhat TCP do? RFC 793, 1122, 1323, 2018, 2581 Connection Oriented Service Full Duplex per connection Reliable by add more overhead Acknowledgement Flow control Timer 46 TCP otcp Process-to-process communication (Typical transport layer function as ) IP protocol domain TCP protocol domain TCP otcp communication Port / IP : Socket Application Process port Process port Application Process port TCP PORT TCP Transport Transport IP Network IP Network IP Process port TCP:Socket osocket Address: A connection is identified by the socket address at its to ends client socket: ,3000; ,21 server socket: ,21; ,3000; Data link Physical Data link Physical 47 48

13 TCP:Socket osocket Multiple Connection:server s unique socket address can be accessed simultaneously by clients TCP : Protocol o TCP passed block of data to IP, consisting of the TCP header and application layer data, called segment o adding reliability in TCP by Errordetection and correction (due to segments corrupted) Flowcontrol (prevent a transmitter overrunning a receiver owing a resource limitations) Resequencing(IP can deliver datagrams in any order) Removing duplicatesegments (due to error-recovery mechanisms used by TCP) TCP : Reliability o Using sequence numbers to identify data opositive acknowledgmentsof data received in the correct sequence oretransmission of segmentswhich have not been acknowledged within a (variable) time limit TCP Encapsulation o Ethernet Frame Protocol Type in IP header : 6 TCP Datagram Ethernet header IP header TCP header data 51 52

14 TCP Header TCP Header osource, destination port:16,16-identify applications at ends of the connection osequence:32-indicates 1st data octet in this segment oacknowledgment:32-next expected sequence number, valid only when the ACK bit (reside in flag) is set TCP Header odata offset:4-32 bit words offset tells the receiver where user data begins o reserved:6 -not used oflag:6 URG: validity of urgent pointer field ACK: validity of acknowledge field PSH: push request (pass segment to app. layer immediately) RST: reset the connection SYN: initial synchronization 55 FIN: sender at end of byte stream 56 TCP Header owindow:16-advertise amount of buffer space this node has allocated ochecksum:16-16 bits 1 s complement of pseudo header, TCP header and data ourgent pointer:16-byte position of data that should be processed first ooptions-variable length option e.g. MSS (max segment size) tells destination node

15 Sliding Window o send and wait for acknowledgment o no ACK in a certain time, retransmit the packet o use for flow control : prevent sender from overloading receiver with data, e.g. high-performance server to slow PC congestion inside network, e.g. router performance, slow link speed o How to provide flow control? set the appropriate size of sliding window size Sliding Window : Flow Control oreceiver advertises it s windows size in acknowledgments osender will adjusts its allowed to send pointer as receiver s advertisement Sliding Window : small window size Sliding Window : larger window size 59 1 byte window size utilizes efficiency of channel in half (half-duplex transmission) why not send many packets and get back cumulative ACK? 60 A larger window size allows more data to be transmitted pending acknowledgment Window size specifies how many bytes the receiver is willing to accept

16 Sequence number of segment Retransmittion Data continuously sent more than segment need not to wait for acknowledgment every segment Retransmit begin with loss segment Error recovery Error recovery 63 receiver has to send ACK with sequence number sender reset timer when receives ACK 64 on time out, sender will retransmit the segment this mechanism is used for error recovery

17 Sliding window buffer sender groups its packet to be transmitted with window indication Sliding window example movement of the right and left edges of the window TCP : Action obefore data could be transferred, a connection must be opened servers do passive open (listen) clients do active open (connect) o when it finished, the connection is closed otcp has general 3 phases connection setup phase data phase connection close phase 68 TCP Connection Setup TCP uses 3-way handshake to establish a connection exchange the sequence number ensures that both ends are ready and sync sequence number

18 TCP Transfer Phase TCP Connection Close simple example with terminal connection such as Telnet. Host echoes back each received character use FIN flag to close connection TCP Open/Close mechanisms ohalf open-one end has closed, aborted without the knowledge of the other end host may be crashed, power off no detection if no data transfer reset segment (RST bit) is sent when detected ohalf close-one end of connection terminated its output, but still receiving data from the other end osimultaneous open-both end perform an active open to each other osimultaneous close-both end perform an passive open to each other 72 TCP State diagram : open -close

19 TCP Connection Setup TCP Transfer Phase 73 (1) SYN = send sending sequence of host A (2) SYN + ACK = send sending sequence of host B + ack number of next host A sequence (3) ACK = ack number of next host B sequence 74 SEQ = Sequence of sending Data ACK = Acknowledge by sending next sequence of received data No sending Data send ACK only Send SEQ only SYN TCP Connection Close TCP State diagram : open Timeout send RST 75 If there s no received data wait for ack : host A send SEQ + FIN Host B received FIN : Terminate all sending : send SEQ + ACK Host B Appl. Terminate : send SEQ + FIN Host A Last ack : send ACK 76

20 TCP State diagram : close TCP Timer oretransmission timer- expecting acknowledgment time opersist timer- keeps window size information flowing okeepalive timer- detect idle connection due to crashing or reboot o2msl- duration of TIME_WAIT state (Maximum Segment Lifetime) TCP Timer o Retransmission Timer 79 Fixed time-out is unacceptable because: impossible to support both LAN/WAN if too short, retransmission problem if too long, decrease throughput TCP uses adaptive retransmission timer learning by measurement of round-trip time (RTT) experiences track these changes to adjust its time-out 80 TCP Timer o Retransmission Timer Jacobson s retransmission time-out (RTO): RTO = A + 4D A = A + h(err) D = D + g( Err - D) Err = M - A A= Average RTT D= Mean deviation h= deviation gain normally set to 0.25 g= gain normally set to Err= difference between Last RTT and current A M= Last RTT

21 TCP Timer o Retransmission Timer : Backoff Exponential backoff when retransmission Increase Double in each retransmission Limit to 64 secs Limit to reset at 9 minutes TCP Timer o Retransmission Timer : Karn s algorithm Consider a case : a packet is transmitted, a time out occurs, the packet is retransmitted, an ACK is received. Is ACK for the first or the second? Karn s algorithm specifies when time-out, do not update the RTT estimator (A) new RTO is calculated only for a not-retransmitted segment TCP Timer o Persist Timer one end advertise window=0 to stop transferring later, it send a segment with window advertisement, but a segment is lost! if no any mechanism, transferring is stopped other end set a persist timer ~500 ms to ask for a new window updated send 1 byte of data to probe 84 TCP Timer o Persist Timer : Silly Window Syndrome small amount of data are exchanged, instead of full-sized segments cause: receiver advertise small windows, instead of waiting for a bigger one sender transmit small chunk, instead of waiting for a bigger one solve: receiver must not advertise small windows sender try to transmit a full-sized segment of a half of window advertisement buffer

22 TCP Timer okeep Alive Timer periodically sent TCP segment to confirm the connection if no acknowledge after a number of retries, the connection is reset. keep-alive segment should not be passed to the application layer Delay Acknowledgment not send ACK immediately after receiving data, delayed ACK typically.2 s, then send win size, ACK and echo data together (piggyback) most implementations use a 200 ms delay Host requirements RFC specifies delay ACK should be implemented with max 500 ms delay Nagel s algorithm RFC 896: prevent sending of small segments which cause congestion in WAN TCP can have only one outstanding unack small segment. Can t send more until ack arrives this collects more data before next sending self-clocking :go as fast as the small latency not for applications that send small data chunks e.g. X windows, a mouse click has to be sent as real time as possible 88 Congestion Control ocongestion : outgoing way has less capacity to send data faster LAN to slower WAN multiple input go to router s less capacity output owhat s then? -packet dropped; if more data sent, more bad situation ohow does a host know that lost packets are from congestion or damage packet? No way!, but our assumption is, lost packets cause by damage is very small (<1%) We assume that the loss come from congestion!

23 Congestion Control o How to solve congestion? not easy to direct solve, but we can avoid Use Jacobson s Congestion Avoidance and Control Algorithm o Jacobson s algorithm: 2 parts slow start congestion avoidance oslowstart Congestion Control set congestion window (cwnd) to one segment data sent no more than cwndand receiver s windows advertisement double cwndeach sending until reach receiver s windows advertisement if congestion occurs, perform slow start with congestion avoidance Congestion Control o Slow start with congestion avoidance slow start until cwndreaches a half of old cwnd at congestion point perform congestion avoidance: increasing cwnd by 1/cwndfor each ackor 1for cwnd-full ack lead to linear increasing of cwnd 91 92