|
|
|
This feature extends bandwidth pooling to optimize load on multiple links from a single carrier in different sites. All CNA decision-making remains local; to fully optimize the collective bandwidth in all locations, an instance of ANS6 must be running at every location. Likewise, each CNA instance must be able to collect SNMP samples for the full set of local and remote links.
The first requirement is to aggregate the local and remote links as physical links under the engine link associated with the carrier that provides the remote and local links. In the example below, two sites have two physical links of 100 Mbps each, and the load threshold table was written with respect to the multi-site pooled bandwidth of 400 Mbps, with a start avoidance at 300 Mbps. The problem is that the engine link “pooled-att” will be considered “cold” until the pooled bandwidth reaches 300 Mbps, but the two local links have a physical capacity of only 200 Mbps.
To address this issue, ANS6 adds a bandwidth command to describe the physical capacity of the local link(s) and a passive command to identify remote links. The bandwidth command should be specified only for local links that dmaker is optimizing, and it should be set below the full capacity of the physical link. For example, the following configuration pools bandwidth for two sites, each with two local links:
module engine link pooled-att # in Mbps - note slightly lower than 100 Mbps - this also # applies only to the "default" physical link (to be friendly # to sites who have only one physical link per engine link) polling physical-link local-att1 snmp-agent local-rtr interface 2 bandwidth 95 polling physical-link local-att2 snmp-agent local-rtr interface 2 bandwidth 95 polling physical-link att-remote1 snmp-agent r1 interface 3 passive polling physical-link att-remote2 snmp-agent r2 interface 3 passive end end
|
|
|