NPAD (Network Path and Application Diagnosis) is designed
to diagnose network performance problems in your end-system
(the machine your browser is running on) or the network between it and
your nearest NPAD server. For each
diagnosed problem, the server prescribes corrective actions
with instructions suitable for non-experts.
Brief instructions
-
The test results are most accurate over a short network path.
If this NPAD server
is not near you, look for a closer server from the list of
Current NPAD Diagnostic Servers.
-
Have an end-to-end application performance goal
(target round-trip time
and
target data rate)
in mind. Enter the parameters on the form below and click
Start Test.
Messages will appear in the log window as the test runs, followed by a diagnostic report.
-
In the diagnostic report, failed tests (in red) indicate
problems that will prevent the application from meeting the end-to-end performance
goal. For each message, a question-mark link ([?]) leads to
additional detailed information about the results.
-
Every test is fully logged (including your IP address) and test results are
public.
We use the logs and results to further refine the software.
For more information, see the
NPAD Documentation, especially the sections:
Note that tests take 2-5 minutes, depending on the parameters that
you provide and the network path.
If there is a queue, waiting times might be long.
If the Java applet above exhibits errors or
the form is blank, try the command line
diagnostic client. Download it
(diag-client.c)
and compile it:
cc diag-client.c -o diag-client
Run it:
./diag-client <server_name> <port> <target_RTT> <target_data_rate>
Where server_name is the hostname of
a diagnostic server (e.g., this server), and
port is the port number the diagnostic
service runs on (8001 for this server).
This software (NPAD/pathdiag version 1.5.6) is being developed under a
collaboration
between the Pittsburgh Supercomputer Center (PSC)
and the National Center for Atmospheric Research (NCAR),
funded under NSF grant ANI-0334061. The project is focused on using
Web100 and other methods to extend
fairly standard diagnostic techniques to compensate for
"symptom scaling" that leads to false positive diagnostic results
on short paths. It is still experimental software and may have bugs.
Please help us improve this service by providing feedback about the
results. Send suggestions, comments and questions to nettune@psc.edu.
Matt Mathis and John Heffner,
Pittsburgh Supercomputing Center, 2005.
|