Cisco Cisco ONS 15454 SONET Multiservice Provisioning Platform (MSPP) Guía De Diseño

OL-4556-01

 

154

If using the native VLAN (VLAN 1) in the service provider network as a metro tag, it is important 
that this tag always be added to the customer traffic, even though the native VLAN ID is not 
normally added to transmitted frames.  If the VLAN 1 metro tag were not added on frames 
entering the service provider network, then the customer VLAN tag would appear to be the metro 
tag, with disastrous results.  The global configuration command vlan dot1q tag native must be 
used to prevent this by forcing a tag to be added to VLAN 1.  Avoiding the use of VLAN 1 as a 
metro tag transporting customer traffic is recommended to reduce the risk of misconfiguration.  A 
best practice is to use VLAN 1 as a private management VLAN in the service provider network. 
 
The IEEE 802.1Q class of service (COS) priority field on the added metro tag is set to zero by 
default, but may be modified by input or output policy maps. 
 

 
The ML-Series card incorporates QoS features to provide control over access to network 
bandwidth resources.  This control enables providers to implement priorities specified in Service 
Level Agreements (SLAs) and offers tools to enable traffic engineering. 
 
The ML-Series QoS provides the ability to classify each packet in the network based on its 
interface of arrival, bridge group, class of service (CoS), IP precedence, and IP differentiated 
services code points.  When classified, further QoS functions can be applied to each packet as it 
traverses the network. 
 
Policing is also provided by the ML-Series card to ensure that no attached equipment submits 
more than a pre-defined amount of bandwidth into the network.  This feature limits the bandwidth 
available to a customer, and provides a mechanism to support traffic engineering. 
 
Priority marking allows Ethernet IEEE 802.1P CoS bits to be marked, as they exit the ML-Series 
card.  This feature operates on the outer IEEE 802.1P tag when coupled with QinQ. 
 
Per class flow queuing is provided to enable fair access to excess network bandwidth, and low 
latency queuing is supported for voice traffic.  It allows allocation of bandwidth to support service-
level agreements and ensure applications with high network resource requirements are 
adequately served.  Buffers are allocated to queues dynamically from a shared resource pool.   
The allocation process incorporates the instantaneous system load as well as the allocated 
bandwidth to each queue to optimize buffer allocation to each queue. 
 
The ML-Series card uses an advanced Weighted Deficit Round Robin (WDRR) scheduling 
process to provide fair access to excess bandwidth as well as guaranteed throughput to each 
class flow. 
 
Admission control is a process that is invoked each time that service is configured on the ML-
Series card to ensure that the card’s available QoS resources are not overcommitted.  In 
particular, admission control ensures that no configurations are accepted where a sum of the 
committed bandwidths on an interface exceed the total bandwidth of that interface. 
 
The QoS bandwidth allocation of Multicast and Broadcast traffic is handled separately and 
differently than Unicast traffic.  Aggregate Multicast and Broadcast traffic are given a fixed 
bandwidth commit of 10% on each interface, and treated as best effort for traffic exceeding 10%.   
Multicast and Broadcast are supported at line-rate.