search

perfsonar / bwctl

A scheduling and policy framework for measurement tools
Apache License 2.0
16 stars 6 forks source link

Add 'tcpdump' as a tool to BWCTL #5

Closed arlake228 closed 3 years ago

arlake228 commented 9 years ago

From @arlake228 on March 14, 2015 12:17

Original issue 902 created by arlake228 on 2014-05-19T09:54:54.000Z:

This is a 'what if' bug, close if it doesn't make technical sense.

It was mentioned in a workshop yesterday that being able to invoke tcpdump (headers only, specific interface that traffic is traversing) via bwctl would be useful.

This bug is to figure out if it would be possible.

Copied from original issue: perfsonar/project#900

arlake228 commented 9 years ago

Comment #1 originally posted by arlake228 on 2014-05-19T11:52:42.000Z:

Is the idea that, along with the normal iperf/etc results, you'd get a tcpdump showing the traffic for the test?

arlake228 commented 9 years ago

Comment #2 originally posted by arlake228 on 2014-05-19T11:58:41.000Z:

I think that would be the common and constrained use case (e.g. we would know ports involved, and could use that to limit what tcpdump is dumping).

I think just being able to run it absent of a performance test may be valuable, but I am not sure how valuable.

The other nagging question is related to the data this would generate, and how to manage that (e.g. a 10 second 10g iperf may be MBs in size uncompressed).

arlake228 commented 9 years ago

Comment #3 originally posted by arlake228 on 2014-05-19T12:18:06.000Z:

Yeah, the general case would be fraught with security implications, though, given the protocol, probably a bit easier than having bwctl return two different sets of result data.

As to the 10g iperf, i'd be curious how big it will be compressed since we could probably set it up to compress the results before shipping back to the client. Either way, i'd presume they'd be big, but i'd be curious how big.

arlake228 commented 9 years ago

Comment #4 originally posted by arlake228 on 2014-05-20T07:37:39.000Z:

10s iperf3 test on a 10G host:

[rootjz@star-pt1 ~]# ls -hals test1.dmp 825M -rw-r--r-- 1 root root 825M May 20 00:33 test1.dmp
[rootjz@star-pt1 ~]# ls -hals test1.dmp.gz 3.1M -rw-r--r-- 1 root root 3.1M May 20 00:33 test1.dmp.gz

20s iperf3 test on a 10G host:

[rootjz@star-pt1 ~]# ls -hals test2.dmp 1.2G -rw-r--r-- 1 root root 1.2G May 20 00:36 test2.dmp [rootjz@star-pt1 ~]# ls -hals test2.dmp.gz 4.5M -rw-r--r-- 1 root root 4.5M May 20 00:36 test2.dmp.gz

So I guess the good news is that it compresses well.

arlake228 commented 9 years ago

Comment #5 originally posted by arlake228 on 2015-01-27T14:39:53.000Z:

<empty>