Hi, I've a nv2 ap with 3 client registered. The clients have a pppoe client set on cpe and one of them has a 30 megabit/s down profile. I done a bandwidth test and in the photo there's the result.
I don't know why tcp performance are so poor instead of udp.

Maybe the cpu is exhausted by btest in TCP mode?

Oh I just take a lo ok at the cpu and is at 100% during test. Why?
I have 2 load balanced 100/10 connection and locally I've no problem to reach high speed.
Maybe the pppoe encapsulation?

For bandwidth testing use only devices connected behind the tested link. Do not use btest on devices you are actually testing.

For bandwidth testing use only devices connected behind the tested link. Do not use btest on devices you are actually testing.

I used the pppoe server RB (rb951) for test the cpe.

The CPU on most routerboards isn't fast enough for TCP testing. Use a normal PC behind the device at test.

The CPU on most routerboards isn't fast enough for TCP testing. Use a normal PC behind the device at test.

I put in download many file on a pc connected to the cpe but I reach max 15 megabit/s instead of 30.

Hi 'ibm' ,

is the PPPoE concentrator a Mikrotik? , if zo .. check out the queue type 'default-small' .. that's the one MT PPPoE concentrators use.

Unfortunately, the Queue size is 10 as default, which is fine for a couple of Mbits, but not for 8+Mbits

Change it to 50 or 100, your problems will be over

Don't forget karma

Hi 'ibm' ,

is the PPPoE concentrator a Mikrotik? , if zo .. check out the queue type 'default-small' .. that's the one MT PPPoE concentrators use.

Unfortunately, the Queue size is 10 as default, which is fine for a couple of Mbits, but not for 8+Mbits

Change it to 50 or 100, your problems will be over

Don't forget karma

Thanks a lot, it's that the problem. I increased only the default queue size and forgot to increase also the default small, now they are both at 50 and it's all good. I added you karma.

Thanks for getting back to 'us'.!

Glad all went well,

Menno

Did you change pfifo queue size on all APs or only on the PPPoE concentrator?
I wrote a thread about this issue: http://forum.mikrotik.com/viewtopic.php?f=7&t=83477

In your experience, is PFIFO 50-100 better than SFQ?

Thank you!

Did you change pfifo queue size on all APs or only on the PPPoE concentrator?
I wrote a thread about this issue: http://forum.mikrotik.com/viewtopic.php?f=7&t=83477

In your experience, is PFIFO 50-100 better than SFQ?

Thank you!

I changed the queue-size only on concentrator because my aps don't perform routing. For the sfq or pfifo question I don't know because I always used the pfifo and if I remember well the pfifo queues are better for large number of queue.

Thank you for your reply!

So you have all queues on the concentrator?

Thank you for your reply!

So you have all queues on the concentrator?

Yes

Thank you for your reply!

So you have all queues on the concentrator?

Yes

Me too , it's best practice to do that in my opinion.. if using NAT .. it's better to let another routerboard take care of that

Running PPPoE in our entire network for our customers since 1,5 year .. never looked back... before we were running a separate L2 net for all our customers. But when crossing the 10 customer mark, we noticed things were not really that flexibe, running PPPoE together with OSPF on our network has really been an eyeopener.

Menno

Running PPPoE in our entire network for our customers since 1,5 year .. never looked back... before we were running a separate L2 net for all our customers. But when crossing the 10 customer mark, we noticed things were not really that flexibe, running PPPoE together with OSPF on our network has really been an eyeopener.
Menno

Do you make EoIP from every CPE to Concentrator ?

Running PPPoE in our entire network for our customers since 1,5 year .. never looked back... before we were running a separate L2 net for all our customers. But when crossing the 10 customer mark, we noticed things were not really that flexibe, running PPPoE together with OSPF on our network has really been an eyeopener.
Menno

Do you make EoIP from every CPE to Concentrator ?

Nope, not yet .. have L2 connection from CPE to concentrator per location (Seperate PPPoE VLAN bridged to Ethernet on SXT, on which we have the router of customer connected.. that router initiates PPPoE session).

Reason i say 'not yet' is because we have multiple Points of Presence and would like to be able to failover PPPoE sessions in the future, for now we have multiple concentrators per Point of Presence.

Menno

Running PPPoE in our entire network for our customers since 1,5 year .. never looked back... before we were running a separate L2 net for all our customers. But when crossing the 10 customer mark, we noticed things were not really that flexibe, running PPPoE together with OSPF on our network has really been an eyeopener.
Menno

Do you make EoIP from every CPE to Concentrator ?

Nope, not yet .. have L2 connection from CPE to concentrator per location (Seperate PPPoE VLAN bridged to Ethernet on SXT, on which we have the router of customer connected.. that router initiates PPPoE session).

Reason i say 'not yet' is because we have multiple Points of Presence and would like to be able to failover PPPoE sessions in the future, for now we have multiple concentrators per Point of Presence.

Menno

For the future I'm planning to use OSPF with MPLS/VPLS to have L2 connectivity on a L3 network.

Running PPPoE in our entire network for our customers since 1,5 year .. never looked back... before we were running a separate L2 net for all our customers. But when crossing the 10 customer mark, we noticed things were not really that flexibe, running PPPoE together with OSPF on our network has really been an eyeopener.
Menno

Do you make EoIP from every CPE to Concentrator ?

Nope, not yet .. have L2 connection from CPE to concentrator per location (Seperate PPPoE VLAN bridged to Ethernet on SXT, on which we have the router of customer connected.. that router initiates PPPoE session).

Reason i say 'not yet' is because we have multiple Points of Presence and would like to be able to failover PPPoE sessions in the future, for now we have multiple concentrators per Point of Presence.

Menno

For the future I'm planning to use OSPF with MPLS/VPLS to have L2 connectivity on a L3 network.

Mpls/Vpls is the way to go. To move 1500byte Packets without Fragmentation and mtu path problems gives a better transport network. With vpls you can bridge access networks which makes management much easier.

.. With vpls you can bridge access networks which makes management much easier.

bridge access network ..to concentrator/s ? so to vlans? sorry, me newbie

.. With vpls you can bridge access networks which makes management much easier.

bridge access network ..to concentrator/s ? so to vlans? sorry, me newbie

You can build a bridge on a central router an add vpls tunnels to it. The other end of the vpls tunnels are
connected to bridges where the cpe's are connected. So you can build a flat access network over a routed
network.

You can build a bridge on a central router an add vpls tunnels to it. The other end of the vpls tunnels are
connected to bridges where the cpe's are connected. So you can build a flat access network over a routed
network.

Now the 'bridge stuff' it's more clear, thnx!

Maybe I'm wrong but:
1) if CPEs 'wan ports' would be all bridged togheter -> single broadcast domain (if no vlans) -> potential security problem ??
2) we would not lose backhauling bandwith with broadcasts (arp,..) between CPEs ?

I would try to figure out on my own..

(sorry for beeing off topic)

You can build a bridge on a central router an add vpls tunnels to it. The other end of the vpls tunnels are
connected to bridges where the cpe's are connected. So you can build a flat access network over a routed
network.

Now the 'bridge stuff' it's more clear, thnx!

Maybe I'm wrong but:
1) if CPEs 'wan ports' would be all bridged togheter -> single broadcast domain (if no vlans) -> potential security problem ??

We manage the cpe's. So no customer gets direct access. If not you've to do bridge filtering.

2) we would not lose backhauling bandwith with broadcasts (arp,..) between CPEs ?

That's not that much. You might do some filtering to avoid arp from cpe to cpe.