Hi Carlos,

Let me be clearer.  I did not do this:

# ovs-vsctl add-bond ovsbr1 bond0 wlan0 wlan1 lacp=active
# ovs-appctl bond/show bond0

The "bonding" that n-casting used is *not* standard.  We are receiving
traffic from both interfaces, while sending on only one.  So, what you
need is flow entries that does not.  You should try to make sure you
can communicate using the OVS with one wireless interface first.

Regards
KK


On 13 February 2013 08:39, Carlos Patricio Amigo Haering <cam...@udec.cl> wrote:
> Hi KK,
>
> I already have the switch OF but I can't achieve the bonding. I'll try
> another solutions or methods.
>
> The flow entry I suppose consists in when receives traffic from the video
> sender this traffic has to go through 2 ports to the two access points with
> dst-mac of both wireless interfaces(that has the same address), right? I've
> understand well?
>
>
> Thank you again.
>
> Regards.
>
> El 2013-02-13 13:21, kk yap escribió:
>
>> Hi Carlos,
>>
>> I did not use the bonding feature in OVS (and it did not exist back
>> when we were doing the demo).  You just need an OpenFlow switch,
>> nothing fancy.  Focus on the correct flow entry to do what you want.
>>
>> Regards
>> KK
>>
>> On 13 February 2013 07:41, Carlos Patricio Amigo Haering
>> <cam...@udec.cl> wrote:
>>>
>>> Dear KK,
>>>
>>> Thank you for your reply. I'm doing the bonding using ovs and following
>>> your
>>> tips but the value lacp_negotiated continues as false.  Seems like the
>>> bonding is doesn't working. What i'm missing?
>>>
>>> # ifconfig wlan0 down
>>> # ifconfig wlan0 hw ether 1c:af:f7:72:ca:fe (now the wlan0 has the same
>>> mac
>>> address than wlan1)
>>> # ifconfig wlan0 up
>>> # ovs-vsctl add-br ovsbr1
>>> # ovs-vsctl add-bond ovsbr1 bond0 wlan0 wlan1 lacp=active
>>> # ovs-appctl bond/show bond0
>>>
>>> bond_mode: balance-slb
>>> bond-hash-algorithm: balance-slb
>>> bond-hash-basis: 0
>>> updelay: 0 ms
>>> downdelay: 0 ms
>>> next rebalance: 1483 ms
>>> lacp_negotiated: false
>>>
>>> slave wlan1: disabled
>>>         may_enable: false
>>>
>>> slave wlan0: disabled
>>>         may_enable: false
>>>
>>>
>>> Thanks in advance.
>>>
>>> Regards.
>>>
>>>
>>> El 2013-02-08 17:41, kk yap escribió:
>>>
>>>> Hi Carlos,
>>>>
>>>> Some comments (and hopefully helpful ones) inline.
>>>>
>>>> Regards
>>>> KK
>>>>
>>>> On 8 February 2013 11:24, Carlos Patricio Amigo Haering
>>>> <cam...@udec.cl> wrote:
>>>>>
>>>>>
>>>>> I want to replicate the n-casting demo with 1 Openflow switch (A PC
>>>>> with
>>>>> 4
>>>>> NIC running the Stanford software reference design), 2 Linksys WRT54GL
>>>>> as
>>>>> access points and my laptop (Ubuntu 12.10) with two wireless NICs (a
>>>>> internal Intel and another one USB D-link) as a mobile client. Looking
>>>>> the
>>>>> documentation seems a hard work. If I'm well understanding:  The switch
>>>>> has
>>>>> to output traffic(from the sender) for two ports (to both AP) and the
>>>>> laptop
>>>>> has to receive the info from each AP using each wireless interfaces and
>>>>> then
>>>>> do in the laptop the bonding between these interfaces. I'm missing
>>>>> something
>>>>> else?
>>>>
>>>>
>>>>
>>>> No.. you got it mostly right here.  I am guessing you will figure out
>>>> more of the details as you go.
>>>>
>>>>
>>>>> The bonding apparently is the most work, right? About this I've been
>>>>> reading
>>>>> this mail list and I've saw that can I use ovs for do it, so the long
>>>>> method
>>>>> of the documentation is deprecated?
>>>>
>>>>
>>>>
>>>> Bingo.  Yes.  The bonding driver is quite hard to use.  I would use OVS.
>>>
>>>
>>>
>>>
>>>
>>>>
>>>>> Also I've read that using ovs doesn't
>>>>> work the bonding between two wireless interfaces, this is correct? I'm
>>>>> a
>>>>> little bit confused about the usage of ovs, Always I associate ovs with
>>>>> a
>>>>> switch and no with a laptop as mobile client. Or I have to do bonding
>>>>> in
>>>>> the
>>>>> switch too?
>>>>
>>>>
>>>>
>>>> Unfortunately, bonding means different thing to different people.  For
>>>> the purpose of n-casting, you want to get traffic from both
>>>> interfaces.  This can be easily done by inserting the correct flow
>>>> rules.  The secret sauce here is to change the mac and ip addresses so
>>>> that the packets would not be dropped.  Give that a shot and let me
>>>> know how it works.
>>>>
>>>>>
>>>>> Internet shows several methods to do bonding(between wired interfaces)
>>>>> like
>>>>> using ifenslave, these methods are valid too?
>>>>>
>>>>> I have a lot of questions, sorry about that but I've never done
>>>>> something
>>>>> like link aggregation.
>>>>
>>>>
>>>>
>>>> Don't sweat over it.  It sounds like you have done your due diligence.
>>>>
>>>>>
>>>>> Thanks in advance.
>>>>>
>>>>> Regards.
>>>>>
>>>>> _______________________________________________
>>>>> openflow-discuss mailing list
>>>>> openflow-discuss@lists.stanford.edu
>>>>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
>>>
>>>
>>>
>
_______________________________________________
openflow-discuss mailing list
openflow-discuss@lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss

Reply via email to