[AusNOG] MLPPP experiences
td_miles at yahoo.com
Wed Mar 27 09:25:27 EST 2013
I would start by using iperf in UDP mode to see just how many out of order packets you are getting (probably half by the sound of it).
Can the OS/application you are testing with re-order packets, or will it simply ditch anything that is out of sequence and ask for a re-transmit ?
Personally I am not suprised with the result although I have no first hand experience with MLPPP over links that are so disparate.
> From: Joseph Goldman <joe at apcs.com.au>
>To: ausnog at lists.ausnog.net
>Sent: Wednesday, 27 March 2013 7:54 AM
>Subject: [AusNOG] MLPPP experiences
>Recently found the need to start experimenting with MLPPP for some upcoming deployments.
>Our LAC has been capable for some time we've just never really had a need for it.
>In our office, we have a Telstra ADSL and an Optus ULL line. Ultimately, there is a 10ms difference between them back to our LAC (one's at ~30 the other at ~40). We achieved a MLPPP connection to our LAC but what we found were less then stellar results. The 2 lines, by themselves, sync and run at around 11mbit/800kbit, but when being used in MLPPP we found we couldn't achieve above 10mbit.
>Monitoring the interfaces we could see that it was successfully sharing that 10mbit across both links equally, we just couldn't max out the lot.
>In your experiences, does the difference in latency that large cause big problems (i.e. packets ultimately transmitting out of order and a lot of corrections to be done) or are there other caveats of MLPPP i should be aware of that could cause similar issues?
>For reference sake, our test LAC is a Cisco 7206 and the test CPE is a Mikrotik RouterOS device (using 2 ADSL Modems in bridge into an ethernet port each).
>AusNOG mailing list
>AusNOG at lists.ausnog.net
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the AusNOG