[c-nsp] High latency/jitter at times on a NPE-G1?

[Mark Rogaski]

An entity claiming to be Rodney Dunn (rodunn at cisco.com) wrote:
: If you want to prove it put a sniffer/analyzer on both sides
: of the box and time sync the traces.
: 
: That is just about the ONLY way you can 100% prove the box
: is introducing latency/jitter in the forwarding stream.

Actually, jitter can be measured with a single analyzer by measuring the
variation of the timestamp deltas.  Latency is trickier to measure, but as
long as you have bidirectional traffic you can measure the the net round
trip delay encountered.  Just pick a packet A travelling to the destination
and a packet B returning from the destination.

Let A_1 and B_1 be the timestamps for the respective packets on analyzer #1
and A_2 and B_2 be the timestamps for the respective packets on analyzer
#2.

Net RTD = abs(abs(A_1 - B_1) - abs(A_2 - B_2))

This doesn't tell you how much latency is being introduced, but it does
tell you if the box is introducing latency.  The nice part is you don't
need to sync clocks on the analyzers.


: 
: I'd be very amazed if it's the 72xx because it's software forwarding
: and once the packet goes in the switching vector it can't be
: held up other than if there is congestion for the most part.
: 

Actually, if the device is doing the forwarding under load I'd be surprised
if it wasn't introducing latency.

Mark

-- 
[]                    |  A computer lets you make more mistakes faster than
[] Mark Rogaski       |  any invention in human history - with the possible
[] wendigo at pobox.com  |  exceptions of handguns and tequila.
[] mrogaski at cpan.org  |  -- Mitch Ratliffe, Technology Review April, 1992
[]                    |
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : https://puck.nether.net/pipermail/cisco-nsp/attachments/20060601/a286a9f6/attachment.bin