Hi hr, can you be a little more specific. Are you saying that I should (on a per site basis) create class maps for all traffic (obviously I can use class default) and then a policy map with explicit bandwidth statements that equals the total bandwidth of the remote site? I only have a single connection into the MPLS cloud so I can only apply one "service policy output" to that interface and with 30 sites I am not going to be able to specify enough bandwidth for each site.

Best regards,

Dan

Dan,

What I meant is that you need to identify the traffic destined to each spoke site, map it to a separate class-map and perform traffic shaping for each and every one of these class-maps (similar to per DLCI traffic shaping with FrameRelay). This is the only way you can guarantee that you will not overwhelm the remote site link.

policy-map MPLS_Sites

class Site_1

match access-group Sites1

shape average 1540000

class Site_2

match access-group Sites2

shape average 512000

...

...

...

class Site_n

match access-group Sitesn

shape average 512000

Regards

Thanks hr, since I am also using LLQ would the following config be correct.

class-map match-all VIDEO

match dscp af31

class-map match-all VOICE

match dscp ef

!

class-map match-all SITE_1

match access-group name SITE-1-SUBNETS

class-map match-all SITE_2

match access-group name SITE-2-SUBNETS

class-map match-all SITE_3

match access-group name SITE-3-SUBNETS

!

!

policy-map SITE_LLQ

class VOICE

priority percent 20

class VIDEO

bandwidth percent 30

class class-default

fair-queue

random-detect

!

policy-map MPLS_SITES

class SITE_1

shape average 2000000

service-policy SITE_LLQ

class SITE_2

shape average 512000

service-policy SITE_LLQ

class SITE_3

shape average 15440000

service-policy SITE_LLQ

class class-default

Dan

Hi Dan,

This would be the correct approach indeed.

Regards