A forum search for M3P (Mikrotik Packet Packing Protocol) brings up surprisingly few hits, and none that I saw really commented on its use, or otherwise.

So is it widely used?

Is it worth using?

Who uses what levels of compression?

How much processor use does it generate?

Would you use it at CPE<>AP level or just for backhauls?

Can/would you use it with nstream as well?

Doesn't anybody know?

I've been looking at using it too. While I do not have answers to your exact questions. Have done a bit of testing on a live network. I enable it on the ap and compressed everything. Using just the bandwidith tester on a marginal link, it seem to the throughput seem to jump up to from about 1m to 1.3 -1.5 range. This is just a seat of the pants type test but results seam very good.


Erik

Hello,
im using it regurarly on backbone links where is constant throughput over 10Mbits, it "packs" small packets together so the wireless device can handle more real data. You have to know that average data packet size is around 400 bytes, so when packed more into one 1500 byte packet you save wireless frame bandwidth. But you loose ping , because the RB is waiting some time - around 15ms to pack the data.
Other tha simple packing DOESNT work for me, woking with nstreme flawlessly.
I hope i written it right.

Jan

thanks for the answer!

Simple question unanswerable from the Manual:

I've several subscribers who use VoIP or Skype, and who might benefit from M3P. Their Mikrotik CPEs connect to the Mikrotik Gateway via wireless and an L2TP tunnel across one or two intermediate Mikrotik Routers.

Do I just run M3P on the terminal routers (CPE<>Gateway) or does it have to be enabled on all the intervening routers?

RouterOS won't let me implement M3P on an L2TP interface. If I enable it on the master interfaces what happens to packets not passing via the tunnel, ie intranet traffic?

as the links are tunneled you should only needed it on the end points as the devices in the middle don't care as the packet is never really examined as it is encapsulated via the tunnel.

as the links are tunneled you should only needed it on the end points as the devices in the middle don't care as the packet is never really examined as it is encapsulated via the tunnel.

But M3P won't work on a tunnel interface.

If I implement M3P packing just on a backhaul does it pack packets regardless of source/destination addresses, thus maximising throughput, or does it only pack packets with matching source/destination addresses.

If the latter, can packing be utilised by masquerading at each end of the link?

You can also utilize a larger MTU on the link. For instance, if you have a link capable of 6000MTU, you can pack all the way up to this and 'cheat' your way to jumbo frames.

two notes on packing.

the router will hold packets until it reaches the interface MTU, or 10-20ms has gone by. This means that if you have a very lightly used link, you might see very strange pings. On busy links this will be minimized and only cost the actually CPU packing time which is minuscule.

I have tested this on ptp links and when the links are unused, pings will come through something like 15ms, 12ms, 9ms, 6ms, 3ms. They would have all bee 3ms replies, but packet 1-4 got held and then transmitted with #5.

Do the same test on a 10% utilized link and the are all 3-5ms because packets are flowing at a fast enough rate to send a full size packet every 1-2ms.