The Tcpdump Group git mirrors - tcpdump/blob - README

   1 @(#) $Header: /tcpdump/master/tcpdump/README,v 1.55 2000-01-28 00:10:22 fenner Exp $ (LBL)
   2
   3 TCPDUMP 3.5
   4 tcpdump.org
   5 tcpdump-workers@tcpdump.org
   6 http://www.tcpdump.org/
   7
   8 This directory contains source code for tcpdump, a tool for network
   9 monitoring and data acquisition.  This software was originally
  10 developed by the Network Research Group at the Lawrence Berkeley
  11 National Laboratory.  The original distribution is available via
  12 anonymous ftp to ftp.ee.lbl.gov, in tcpdump.tar.Z.  More recent
  13 development is performed at tcpdump.org, http://www.tcpdump.org/
  14
  15 Tcpdump uses libpcap, a system-independent interface for user-level
  16 packet capture.  Before building tcpdump, you must first retrieve and
  17 build libpcap, also originally from LBL and now being maintained by
  18 tcpdump.org; see http://www.tcpdump.org/ .
  19
  20 Once libpcap is built (either install it or make sure it's in
  21 ../libpcap), you can build tcpdump using the procedure in the INSTALL
  22 file.
  23
  24 The program is loosely based on SMI's "etherfind" although none of the
  25 etherfind code remains.  It was originally written by Van Jacobson as
  26 part of an ongoing research project to investigate and improve tcp and
  27 internet gateway performance.  The parts of the program originally
  28 taken from Sun's etherfind were later re-written by Steven McCanne of
  29 LBL.  To insure that there would be no vestige of proprietary code in
  30 tcpdump, Steve wrote these pieces from the specification given by the
  31 manual entry, with no access to the source of tcpdump or etherfind.
  32
  33 Over the past few years, tcpdump has been steadily improved by the
  34 excellent contributions from the Internet community (just browse
  35 through the CHANGES file).  We are grateful for all the input.
  36
  37 Richard Stevens gives an excellent treatment of the Internet protocols
  38 in his book ``TCP/IP Illustrated, Volume 1''. If you want to learn more
  39 about tcpdump and how to interpret its output, pick up this book.
  40
  41 Some tools for viewing and analyzing tcpdump trace files are available
  42 from the Internet Traffic Archive:
  43
  44         http://www.acm.org/sigcomm/ITA/
  45
  46 Another tool that tcpdump users might find useful is tcpslice:
  47
  48         ftp://ftp.ee.lbl.gov/tcpslice.tar.Z
  49
  50 It is a program that can be used to extract portions of tcpdump binary
  51 trace files. See the above distribution for further details and
  52 documentation.
  53
  54 Problems, bugs, questions, desirable enhancements, source code
  55 contributions, etc., should be sent to the email address
  56 "tcpdump-workers@tcpdump.org".
  57
  58 -------------------------------------
  59 This directory also contains some short awk programs intended as
  60 examples of ways to reduce tcpdump data when you're tracking
  61 particular network problems:
  62
  63 send-ack.awk
  64         Simplifies the tcpdump trace for an ftp (or other unidirectional
  65         tcp transfer).  Since we assume that one host only sends and
  66         the other only acks, all address information is left off and
  67         we just note if the packet is a "send" or an "ack".
  68
  69         There is one output line per line of the original trace.
  70         Field 1 is the packet time in decimal seconds, relative
  71         to the start of the conversation.  Field 2 is delta-time
  72         from last packet.  Field 3 is packet type/direction.
  73         "Send" means data going from sender to receiver, "ack"
  74         means an ack going from the receiver to the sender.  A
  75         preceding "*" indicates that the data is a retransmission.
  76         A preceding "-" indicates a hole in the sequence space
  77         (i.e., missing packet(s)), a "#" means an odd-size (not max
  78         seg size) packet.  Field 4 has the packet flags
  79         (same format as raw trace).  Field 5 is the sequence
  80         number (start seq. num for sender, next expected seq number
  81         for acks).  The number in parens following an ack is
  82         the delta-time from the first send of the packet to the
  83         ack.  A number in parens following a send is the
  84         delta-time from the first send of the packet to the
  85         current send (on duplicate packets only).  Duplicate
  86         sends or acks have a number in square brackets showing
  87         the number of duplicates so far.
  88
  89         Here is a short sample from near the start of an ftp:
  90                 3.00    0.20   send . 512
  91                 3.20    0.20    ack . 1024  (0.20)
  92                 3.20    0.00   send P 1024
  93                 3.40    0.20    ack . 1536  (0.20)
  94                 3.80    0.40 * send . 0  (3.80) [2]
  95                 3.82    0.02 *  ack . 1536  (0.62) [2]
  96         Three seconds into the conversation, bytes 512 through 1023
  97         were sent.  200ms later they were acked.  Shortly thereafter
  98         bytes 1024-1535 were sent and again acked after 200ms.
  99         Then, for no apparent reason, 0-511 is retransmitted, 3.8
 100         seconds after its initial send (the round trip time for this
 101         ftp was 1sec, +-500ms).  Since the receiver is expecting
 102         1536, 1536 is re-acked when 0 arrives.
 103
 104 packetdat.awk
 105         Computes chunk summary data for an ftp (or similar
 106         unidirectional tcp transfer). [A "chunk" refers to
 107         a chunk of the sequence space -- essentially the packet
 108         sequence number divided by the max segment size.]
 109
 110         A summary line is printed showing the number of chunks,
 111         the number of packets it took to send that many chunks
 112         (if there are no lost or duplicated packets, the number
 113         of packets should equal the number of chunks) and the
 114         number of acks.
 115
 116         Following the summary line is one line of information
 117         per chunk.  The line contains eight fields:
 118            1 - the chunk number
 119            2 - the start sequence number for this chunk
 120            3 - time of first send
 121            4 - time of last send
 122            5 - time of first ack
 123            6 - time of last ack
 124            7 - number of times chunk was sent
 125            8 - number of times chunk was acked
 126         (all times are in decimal seconds, relative to the start
 127         of the conversation.)
 128
 129         As an example, here is the first part of the output for
 130         an ftp trace:
 131
 132         # 134 chunks.  536 packets sent.  508 acks.
 133         1       1       0.00    5.80    0.20    0.20    4       1
 134         2       513     0.28    6.20    0.40    0.40    4       1
 135         3       1025    1.16    6.32    1.20    1.20    4       1
 136         4       1561    1.86    15.00   2.00    2.00    6       1
 137         5       2049    2.16    15.44   2.20    2.20    5       1
 138         6       2585    2.64    16.44   2.80    2.80    5       1
 139         7       3073    3.00    16.66   3.20    3.20    4       1
 140         8       3609    3.20    17.24   3.40    5.82    4       11
 141         9       4097    6.02    6.58    6.20    6.80    2       5
 142
 143         This says that 134 chunks were transferred (about 70K
 144         since the average packet size was 512 bytes).  It took
 145         536 packets to transfer the data (i.e., on the average
 146         each chunk was transmitted four times).  Looking at,
 147         say, chunk 4, we see it represents the 512 bytes of
 148         sequence space from 1561 to 2048.  It was first sent
 149         1.86 seconds into the conversation.  It was last
 150         sent 15 seconds into the conversation and was sent
 151         a total of 6 times (i.e., it was retransmitted every
 152         2 seconds on the average).  It was acked once, 140ms
 153         after it first arrived.
 154
 155 stime.awk
 156 atime.awk
 157         Output one line per send or ack, respectively, in the form
 158                 <time> <seq. number>
 159         where <time> is the time in seconds since the start of the
 160         transfer and <seq. number> is the sequence number being sent
 161         or acked.  I typically plot this data looking for suspicious
 162         patterns.
 163
 164
 165 The problem I was looking at was the bulk-data-transfer
 166 throughput of medium delay network paths (1-6 sec.  round trip
 167 time) under typical DARPA Internet conditions.  The trace of the
 168 ftp transfer of a large file was used as the raw data source.
 169 The method was:
 170
 171   - On a local host (but not the Sun running tcpdump), connect to
 172     the remote ftp.
 173
 174   - On the monitor Sun, start the trace going.  E.g.,
 175       tcpdump host local-host and remote-host and port ftp-data >tracefile
 176
 177   - On local, do either a get or put of a large file (~500KB),
 178     preferably to the null device (to minimize effects like
 179     closing the receive window while waiting for a disk write).
 180
 181   - When transfer is finished, stop tcpdump.  Use awk to make up
 182     two files of summary data (maxsize is the maximum packet size,
 183     tracedata is the file of tcpdump tracedata):
 184       awk -f send-ack.awk packetsize=avgsize tracedata >sa
 185       awk -f packetdat.awk packetsize=avgsize tracedata >pd
 186
 187   - While the summary data files are printing, take a look at
 188     how the transfer behaved:
 189       awk -f stime.awk tracedata | xgraph
 190     (90% of what you learn seems to happen in this step).
 191
 192   - Do all of the above steps several times, both directions,
 193     at different times of day, with different protocol
 194     implementations on the other end.
 195
 196   - Using one of the Unix data analysis packages (in my case,
 197     S and Gary Perlman's Unix|Stat), spend a few months staring
 198     at the data.
 199
 200   - Change something in the local protocol implementation and
 201     redo the steps above.
 202
 203   - Once a week, tell your funding agent that you're discovering
 204     wonderful things and you'll write up that research report
 205     "real soon now".
 206