The capacity of a path, a.k.a. bottleneck bandwidth, is the maximum IP-layer throughput that a flow can get in the network path from a sender S to a receiver R. The capacity does not depend on the load of the path. The route that the path follows must be unique and should remain constant during the measurements. The capacity of the path is determined by the link with the minimum transmission rate. We refer to that link as the narrow link of the path.
Pathrate is an end-to-end capacity estimation tool. It uses packet-pairs and packet-trains (a.k.a. `packet dispersion' methods), in conjunction with statistical techniques, to estimate the capacity of the narrow link in the path.
Pathrate requires the user to have access at both ends of the path (i.e., you have to be able to run Pathrate at both S and R). It uses UDP packets for the packet-pairs and packet-trains, as well as a TCP connection for exchange of control information. It can be run from user-space, and it does not require superuser privileges.
$gunzip pathrate_x.y.z.tar.gz
$tar xvf pathrate_x.y.z.tar
$cd pathrate_x.y.z
$./configure
$make
The two executables that will be created, are:
$pathrate_snd
By default, pathrate_snd will exit after 1 measurement.
Use switch -i to run sender in iterative mode.
Other options are,
-q : quite mode
-v : verbose mode
-o < file > : print log in user specified file
-H|-h : print help and exit
and then run pathrate_rcv at the destination of the path R:
$pathrate_rcv -s sender_address
Options for receiver are :
-s : hostname/ipaddress of sender
-q : quite mode
-v : verbose mode
-Q : Quick termination mode
-o < file > : print log in user specified file
-O < file > : append log in user specified file [default is pathrate.output]
-N < file > : print output in ULM format to
-H|-h : print help and exit
The argument sender_address is the name or ip address of host S. If, for example, you
want to measure the capacity from a host abc.foo1.com to a host xyz.foo2.com, run
abc.foo1.com$pathrate_snd -i -q
and then run
xyz.foo2.com$pathrate_rcv -s abc.foo1.com
You can let the sender running at the background.
Note, however, that the sender will only accept one receiver at a time.
It is important to run Pathrate from relatively idle hosts. Before running Pathrate , make sure that there are no other CPU or I/O intensive processes running. If there are, it is likely that they will interfere with Pathrate's user-level packet timestamping and render the measurements inaccurate.
Then, Pathrate sends a few packet trains of increasing length (called "preliminary measurements"). The goal here is to detect if the narrow link has parallel sub-channels, or if it performs traffic shaping. This phase also checks whether the path is "easy to measure" (very lightly loaded). In that case, Pathrate reports its final estimate and exits. This is called quick termination mode. An important objective of these preliminary measurements is to measure the bandwidth resolution (think of this as the bin width of a histogram). The final capacity estimate will be a range of this width.
The very final outcome of Pathrate is the capacity estimate for the path.
Pathrate stores all intermediate results in the file "pathrate.output". If you don't want to get all the intermediate measurements at the receiver's stdout, use -q flag. I suggest you do that if you run the receiver remotely and the capacity of the path is just a few kbps. The reason is that the telnet/ssh traffic can interfere with Pathrate 's measurements.
altair.cc.gatech.edu$./pathrate_snd -i -q &
./pathrate_rcv -s altair -o test.out pathrate run from altair.cc.gatech.edu to utah.ron.lcs.mit.edu on Wed Apr 9 22:26:05 2003 --> Average round-trip time: 41.1ms --> Capacity Resolution: 2.1 Mbps -- Phase I: Detect possible capacity modes -- -> Train length: 2 - Packet size: 600B -> 0% completed -> Train length: 3 - Packet size: 898B -> 2% completed -> Train length: 4 - Packet size: 1196B -> 5% completed -> Train length: 4 - Packet size: 1219B -> 7% completed -> Train length: 4 - Packet size: 1242B -> 10% completed .............................................................. .............................................................. -> Train length: 4 - Packet size: 1488B -> 87% completed -> Train length: 4 - Packet size: 1488B -> 90% completed -> Train length: 4 - Packet size: 1488B -> 92% completed -> Train length: 4 - Packet size: 1488B -> 95% completed -> Train length: 4 - Packet size: 1488B -> 97% completed -- Local modes : In Phase I -- * Mode: 103 Mbps to 105 Mbps - 134 measurements Modal bell: 705 measurements - low : 90 Mbps - high : 116 Mbps * Mode: 140 Mbps to 142 Mbps - 20 measurements Modal bell: 50 measurements - low : 136 Mbps - high : 143 Mbps -- Phase II: Estimate Asymptotic Dispersion Rate (ADR) -- -- Number of trains: 500 - Train length: 48 - Packet size: 1488B -- Local modes : In Phase II -- * Mode: 96 Mbps to 98 Mbps - 298 measurements Modal bell: 478 measurements - low : 91 Mbps - high : 102 Mbps ------------------------------------------------- Final capacity estimate : 96 Mbps to 98 Mbps -------------------------------------------------For more verbose output use -v flag.
./pathrate_rcv -s altair -v -o test.out pathrate run from altair to jawks on Wed Apr 9 18:07:11 2003 --> Average round-trip time: 0.3ms --> Minimum acceptable packet pair dispersion: 15 usec -- Maximum train length discovery -- Train length: 2 -> 98 Mbps Train length: 3 -> 65 Mbps Train length: 4 -> 98 Mbps Train length: 5 -> 98 Mbps Train length: 6 -> 98 Mbps .............................................................. .............................................................. Train length: 36 -> 98 Mbps Train length: 40 -> 94 Mbps Train length: 44 -> 97 Mbps Train length: 48 -> 98 Mbps --> Maximum train length: 48 packets --Preliminary measurements with increasing packet train lengths-- Train length: 2 -> 98 Mbps 90 Mbps 98 Mbps 99 Mbps 98 Mbps 87 Mbps 98 Mbps Train length: 3 -> 98 Mbps 97 Mbps 88 Mbps 108 Mbps 98 Mbps 98 Mbps 46 Mbps Train length: 4 -> 98 Mbps 86 Mbps 98 Mbps 98 Mbps 98 Mbps 96 Mbps 98 Mbps Train length: 5 -> 98 Mbps 98 Mbps 98 Mbps 99 Mbps 99 Mbps 98 Mbps 98 Mbps Train length: 6 -> 91 Mbps 98 Mbps 98 Mbps 100 Mbps 94 Mbps 98 Mbps 98 Mbps Train length: 7 -> 97 Mbps 81 Mbps 98 Mbps 98 Mbps 98 Mbps 83 Mbps 84 Mbps Train length: 8 -> 98 Mbps 98 Mbps 97 Mbps 98 Mbps 97 Mbps 98 Mbps 95 Mbps Train length: 9 -> 98 Mbps 98 Mbps 98 Mbps 98 Mbps 98 Mbps 98 Mbps 98 Mbps Train length: 10 -> 98 Mbps 98 Mbps 97 Mbps 96 Mbps 98 Mbps 97 Mbps 98 Mbps --> Capacity Resolution: 718 kbps `Quick-termination' - Sufficiently low measurement noise --> Coefficient of variation: 0.030 ------------------------------------------------- Final capacity estimate : 96 Mbps to 97 Mbps -------------------------------------------------Receiver can be force to exit at this point by use of -Q flag.