Information Technology Reference
In-Depth Information
■
bandwidth interzone:
Used to specify the maximum amount of bandwidth for calls between a local zone and other zones.
■
bandwidth session:
Controls the maximum bandwidth for a call within a local zone.
■
bandwidth total:
Specifies the maximum total bandwidth for all calls within or into/out of a local zone.
■
bandwidth remote:
Controls the total bandwidth for calls between the local gatekeeper and all remote zones.
■
arq reject-unknown-prefix:
Configures a gatekeeper to reject admission requests for zone prefixes that are not configured.
This command can be used to ensure that routing loops do not develop across redundant CUCM trunks.
Traffic Policing and Shaping
Policing
is the process by which traffic is dropped, re-marked, or simply transmitted according to whether the traffic conforms to a
contract. Policing typically takes place at the edge of a network and can be configured in an input or output direction on an interface.
Shaping
involves the buffering of excess traffic during transient periods of network congestion. The buffered traffic is then
transmitted when congestion abates. Shaping usually occurs at the edge of the network and can be applied to an interface in an output
direction.
Both traffic policers and shapers rely on token-bucket algorithms. Tokens are periodically added (to a token bucket), and these tokens
directly influence how much traffic can be sent, with each token allowing either a bit or byte to be sent. When a policer or shaper
wants to decide whether traffic conforms to or exceeds configured contract rates, it checks the number of tokens remaining in the
token bucket.
There are three types of class-based policer (configured using the MQC):
■
Single-rate, two-color policer:
In this case, a single token bucket is used, and traffic simply either conforms to or exceeds
the configured contract rate. Actions can be specified for traffic that conforms or exceeds the contract rate.
■
Single-rate, three-color policer (RFC 2697):
Two token buckets are used, with tokens being periodically added to the first
bucket and any “overspilled” tokens filling the second bucket.
