摘要:
A technique is provided for one or more network nodes to deterministically select data flows to preempt. In particular, each node employs a set of predefined rules which instructs the node as to which existing data flow should be preempted in order to admit a new high-priority data flow. The rules are precisely defined and are common to all nodes configured in accordance with the present invention. Illustratively, a network node not only selects a data flow to preempt, but additionally may identify other “fate sharing” data flows that may be preempted. As used herein, a group of data flows has a fate-sharing relationship if the application instance(s) containing the data flows functions adequately only when all the fate-shared flows are operational. In a first illustrative embodiment, after a data flow in a fate-sharing group is preempted, network nodes may safely tear down the group's remaining data flows. In a second illustrative embodiment, when a data flow is preempted, all its fate-shared data flows are marked as being “at risk.” Because the at-risk flows are not immediately torn down, it is less likely that resources allocated for the at-risk flows may be freed then subsequently used to establish relatively lower-priority data flows instead of relatively higher-priority data flows.
摘要:
In one embodiment, an intermediate network device includes a communication facility configured to receive a reservation request message that includes a flow spec object. The flow spec object specifies one or more flow parameters that describe a given traffic flow that desires to pass through the intermediate network device. A flow is configured to compare the one or more flow parameters specified in the flow spec object to one or more constants stored in a memory, to determine a type of traffic of the given traffic flow. The flow analyzer determines the type of traffic independent of any differentiated services codepoint (DSCP) values in packets of the given traffic flow. A traffic scheduler is configured to assign the given traffic flow to a particular per hop behavior (PHB) based on the determined type of traffic for the given traffic flow.
摘要:
An MPLS router operable for labeled switch path (LSP) operation defines a compression index for identifying a decompression context between other MPLS LSP routers. The compression index allows a multipoint-to-point link between MPLS routers, thereby avoiding an exhaustive mesh of point-to-point links between each of the MPLS routers. The originator ID identifies each of the multipoint originating endpoints at a common destination, and maintains the context of each compressed header to match incoming compressed headers to the corresponding header values. The originator ID, typically the IP address of the originator, operates as the compression index on the multipoint-to-point connection, operable to distinguish multiple originators of the multipoint-to-point connection and provide header compression for each.
摘要:
A two phase reservation mechanism for use with computer networks carrying voice or other time or bandwidth sensitive traffic. During the first or “resource allocation” phase, network resources sufficient to support the anticipated voice traffic are set aside within the computer network along the route between the sourcing entity and receiving entity. Although the network resources have been set aside, they are specifically not made available to the voice traffic, until the second phase of the reservation mechanism, called the “resource available” phase. During the resource available phase, the network resources that were previously set aside are now made available to the voice traffic.
摘要:
A merging ATM switch forwards to a common downstream ATM switch ATM data cells that it receives from a plurality of upstream ATM switches. The merging ATM switch may have different upstream ATM switches employ different respective virtual-circuit identifiers for those commonly destined data cells, but it uses a common virtual-circuit identifier in forwarding them to the downstream switch. The upstream ATM switches intersperse among the data cells forward-directed resource-management cells bearing the virtual-channel identifiers that those upstream switches use on the commonly destined data cells. The merging ATM switch responds to such forward-directed resource-management cells by sending corresponding forward-directed resource-management cells to the downstream ATM switch and reverse-directed resource-management cells to the upstream ATM switches. The forward-directed resource-management cells sent to the downstream ATM switch bear the common virtual-circuit identifier, whereas the reverse-directed resource-management cells sent to the upstream ATM switches bear the respective virtual-circuit identifiers that those upstream ATM switches placed in the forward-directed resource-management cells. By so doing, the merging switch avoids the ambiguity that resource-management-cell forwarding other-wise entails.