Avaya 555-245-600 Benutzerhandbuch

Seite von 378
Converged network design
Issue 6 January 2008
307
 
Figure 83: Layered Server Cluster Topology
Assume that each G650 is equipped with redundant TN2602AP Media Resource 320 circuit 
packs optionally configured for load balancing or IP bearer duplication. Each G650 is also 
assumed to contain duplicated TN2312BP IP Server Interfaces (IPSI). The number of TN799DP 
Control Lan (C-LAN) socket termination boards would be sized to accommodate the devices in 
the wider network and the call capacity of the cluster, but for this small configuration assume a 
C-LAN for each G650. Also assume the Layer 3 devices use hardware switching for Layer 3 
forwarding and that they are also capable of Layer 2 switching between ports. It is important to 
remember that IP Telephony LAN traffic consists primarily of smaller (approximately 218 octet) 
packets and it is the per packet overhead that impacts software based routing. Consider that 
telephony traffic is roughly an order of magnitude more packets per unit bandwidth than typical 
web page transfers.
Layers
The obvious question is why there is a separate Layer 2 access level if the devices at the Layer 
3 distribution layer are capable of Layer 2 switching? In general, the access layer reduces the 
complexity of the network block by separating the functions of the devices, and provides 
scalability when more ports are required as the network grows. To ensure network modularity, 
the routers serving this cluster should be dedicated to the cluster and sized to the task. 
Simplification argues for the reduction of subnets and therefore routed interfaces in the cluster 
since the service is common. If remote IPSIs or multiple server clusters are implemented across 
the network, using a single subnet within the cluster simplifies the configuration of the entire 
network. The addition of static subnets in the direction of the cluster [8] increases the