Avaya 555-245-600 Manual Do Utilizador
Network design
298 Avaya Application Solutions IP Telephony Deployment Guide
shows, traffic up to the CIR is
guaranteed, whereas traffic beyond the CIR usually is not. This is how Frame Relay is intended
to work. CIR is a committed and reliable rate, whereas burst is a bonus when network
conditions permit it without infringing upon the CIR of any user. For this reason, burst frames
are marked as discard eligible (DE), and are queued or discarded when network congestion
exists. Although experience has shown that customers can achieve significant burst throughput,
it is unreliable and unpredictable, and not suitable for real-time applications like IP Telephony.
Therefore, the objective is to prevent voice traffic from entering the burst range and being
marked DE. One way to accomplish this is to prohibit bursting by shaping the traffic to the CIR
and setting the excess burst size (B
to work. CIR is a committed and reliable rate, whereas burst is a bonus when network
conditions permit it without infringing upon the CIR of any user. For this reason, burst frames
are marked as discard eligible (DE), and are queued or discarded when network congestion
exists. Although experience has shown that customers can achieve significant burst throughput,
it is unreliable and unpredictable, and not suitable for real-time applications like IP Telephony.
Therefore, the objective is to prevent voice traffic from entering the burst range and being
marked DE. One way to accomplish this is to prohibit bursting by shaping the traffic to the CIR
and setting the excess burst size (B
e
– determines the burst range) to zero. However, this also
prevents data traffic from using the burst range.
Additional frame relay information
One interesting piece of knowledge is that most IXCs convert the long-haul delivery of Frame
Relay into ATM. That is, the Frame Relay PVC is converted to an ATM PVC at the first Frame
Relay switch after leaving the customer premise. It is not converted back to Frame Relay until
the last Frame Relay switch before entering the customer premise. This is significant because
ATM has built- in Class of Service (CoS). A customer can contract with a carrier to convert the
Frame Relay PVC into a constant bit rate (CBR) ATM PVC. ATM CBR cells are delivered with
lower latency and higher reliability.
Finally, under the best circumstances, Frame Relay is still inherently more susceptible to delay
than ATM or TDM. Therefore, after applying the best possible queuing mechanism, one should
still expect more delay over Frame Relay than is present over ATM or TDM.
Relay into ATM. That is, the Frame Relay PVC is converted to an ATM PVC at the first Frame
Relay switch after leaving the customer premise. It is not converted back to Frame Relay until
the last Frame Relay switch before entering the customer premise. This is significant because
ATM has built- in Class of Service (CoS). A customer can contract with a carrier to convert the
Frame Relay PVC into a constant bit rate (CBR) ATM PVC. ATM CBR cells are delivered with
lower latency and higher reliability.
Finally, under the best circumstances, Frame Relay is still inherently more susceptible to delay
than ATM or TDM. Therefore, after applying the best possible queuing mechanism, one should
still expect more delay over Frame Relay than is present over ATM or TDM.
MPLS
MultiProtocol Label Switching (MPLS) VPN service from service providers is commonly used by
enterprises for WAN connectivity. The service is often available over different types of access
links, and usually offers multiple classes of service. MPLS service is typically expected to
provide good QoS and therefore to satisfy VoIP requirements, though this often depends on the
Service Layer Agreement (SLA) and the actual quality delivered by the service provider.
With MPLS service, unlike private WAN, the enterprise controls QoS explicitly only on the
access link —- that is, on the connection from each enterprise site to the MPLS network. Within
the MPLS network QoS is controlled by the service provider. The enterprise affects the service
given to its traffic by assigning the traffic to appropriate classes of service in the service
provider’s network. This is done with DiffServ Code Point (DSCP) marking in the packet’s IP
header. DSCP remarking by the enterprise edge routers may be required, mapping the DSCPs
of enterprise traffic to the DSCP values designated by the MPLS service provider for the
different classes of service in their service offering.
enterprises for WAN connectivity. The service is often available over different types of access
links, and usually offers multiple classes of service. MPLS service is typically expected to
provide good QoS and therefore to satisfy VoIP requirements, though this often depends on the
Service Layer Agreement (SLA) and the actual quality delivered by the service provider.
With MPLS service, unlike private WAN, the enterprise controls QoS explicitly only on the
access link —- that is, on the connection from each enterprise site to the MPLS network. Within
the MPLS network QoS is controlled by the service provider. The enterprise affects the service
given to its traffic by assigning the traffic to appropriate classes of service in the service
provider’s network. This is done with DiffServ Code Point (DSCP) marking in the packet’s IP
header. DSCP remarking by the enterprise edge routers may be required, mapping the DSCPs
of enterprise traffic to the DSCP values designated by the MPLS service provider for the
different classes of service in their service offering.