Mailing List Archive

Mark wrote:
> Hi ,
>
> i'm getting crazy about the problem below.
>
> We aggregate xDSL customers using PPP (PPPoA or PPPoE) over PA-A3 ATM
> adapters.
> Now, some customers are experiencing very high tcp packet loss but when
> we test ATM (with ATM OAM pings) we don't see any packet loss. It is
> possible that it is a problem on the router? Only some customers are
> experiencing the problem. We tried anything on the router (delete pcv,
> change PPP type etc) but problem persist.
> I'm pretty sure that problems are on ATM provider, but on ATM side what
> i need do check? ATM Buffers? ATM Errors? My ATM provider say that ATM
> ping are enough to check that ATM network is working well but I'm sure
> that problems are atm related. Maybe a IOS Bug?

G'day Mark,

Check the QoS parameters (Sustained Cell Rate, Peak Cell Rate, Maximum
Burst Size) that your PVC is configured for against your ATM provider's
SLA/technical spec. If you're bursting faster/for longer than what your
provider is configured for, it'll start dropping cells. This will
likely only be noticed with larger packets than your typical 64 byte
ICMP echo, such as your typical ~1000 byte TCP packet. Try pinging the
IP interface on your ATM provider's end of the PVC with various packet
sizes up to the MTU.

Cheers,

--
James Wakefield,
Unix Administrator, Information Technology Services Division
Deakin University, Geelong, Victoria 3217 Australia.

Phone: 03 5227 8690 International: +61 3 5227 8690
Fax: 03 5227 8866 International: +61 3 5227 8866
E-mail: james.wakefield@deakin.edu.au
Website: http://www.deakin.edu.au
_______________________________________________
cisco-bba mailing list
cisco-bba@puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-bba

After a bit of research i found that rx_cell_loss are increasing quickly.
What this means?
rgds
mark

James Wakefield ha scritto:
> Mark wrote:
>
>> Hi ,
>>
>> i'm getting crazy about the problem below.
>>
>> We aggregate xDSL customers using PPP (PPPoA or PPPoE) over PA-A3 ATM
>> adapters.
>> Now, some customers are experiencing very high tcp packet loss but when
>> we test ATM (with ATM OAM pings) we don't see any packet loss. It is
>> possible that it is a problem on the router? Only some customers are
>> experiencing the problem. We tried anything on the router (delete pcv,
>> change PPP type etc) but problem persist.
>> I'm pretty sure that problems are on ATM provider, but on ATM side what
>> i need do check? ATM Buffers? ATM Errors? My ATM provider say that ATM
>> ping are enough to check that ATM network is working well but I'm sure
>> that problems are atm related. Maybe a IOS Bug?
>>
>
> G'day Mark,
>
> Check the QoS parameters (Sustained Cell Rate, Peak Cell Rate, Maximum
> Burst Size) that your PVC is configured for against your ATM provider's
> SLA/technical spec. If you're bursting faster/for longer than what your
> provider is configured for, it'll start dropping cells. This will
> likely only be noticed with larger packets than your typical 64 byte
> ICMP echo, such as your typical ~1000 byte TCP packet. Try pinging the
> IP interface on your ATM provider's end of the PVC with various packet
> sizes up to the MTU.
>
> Cheers,
>
>

Mark wrote:
> After a bit of research i found that rx_cell_loss are increasing quickly.
> What this means?
> rgds
> mark
>

It could mean a few things...can you paste the full output of a show atm
vc and a show atm controller for the VC in question?

My instinct is to suspect a mismatch in ATM QoS parameters between your
end and the ATM provider, probably because I've experienced that before.
If you're getting cell loss on the rx side, your VC's QoS may be
configured for rates lower than your ATM provider has configured,
meaning they're bursting faster and/or for longer than your interface
will accept, and your interface is marking them for discard by setting
CLP. Physical problems such as cable damage or dirty connectors etc.
are probably not going to be it as your OAM pings are working. It's
possible you're running out of buffers/queue slots, you could try
increasing your rx queue in that case. Other than that, it could be a
buggy controller or IOS, which you should be able to search cisco.com for.

--
James Wakefield,
Unix Administrator, Information Technology Services Division
Deakin University, Geelong, Victoria 3217 Australia.

Phone: 03 5227 8690 International: +61 3 5227 8690
Fax: 03 5227 8866 International: +61 3 5227 8866
E-mail: james.wakefield@deakin.edu.au
Website: http://www.deakin.edu.au
_______________________________________________
cisco-bba mailing list
cisco-bba@puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-bba