Image Processing Reference
In-Depth Information
the MIB is a database that can be used to maintain OpenFlow table [
22
] and key
control information to assist packet forwarding and the nodes
reaction to failures.
Besides, the CPE component assures interaction with Resource Management Func-
tions (RMF) [
96
] to properly configure schedulers on the nodes [
80
] so that each
CoS receives the amount of bandwidth assigned to it. The nodes execute these
functions based on the instructions that they receive from the ARC. To this end, the
CPE enables UDP port recognition (routers are permanently listening on a specific
UDP port) or IP Router Alert Option (RAO) [
95
] on nodes to properly intercept,
interpret and process control messages. The CPE also assists the SDNC by filling in
relevant control messages with appropriate information such as the IDs (e.g., IP or
MAC addresses) of outgoing interfaces and their capacities as
Record Route Object
(RRO) [
97
,
98
] on paths. When the MIB and the CPE are deployed at the network
ERs, they include additional functions for inter-domain forwarding and routing
operations [
99
]. Traffic control and traffic conditioning (e.g., traffic shaping and
policing) must also be enforced at the ERs to force admitted traffic flows to comply
with the SLAs [
14
].
'
8.4.2 Network Configurations and Operations
The SDNC can rely on the underlying link state routing protocols [
86
] to discover
and maintain the network topology dynamically. Hence, the network topology is
fed to appropriate algorithms (e.g., Dijkstra [
86
]) to compute all possible paths,
especially the ER-to-ER paths inside the core network under the SDNC
s control.
One may want to combine the ER-to-ER paths to create all possible branched paths
for multicasting purposes upon need [
17
]. Among the possible paths, the best ones
that can be used for efficient service delivery can be filtered and maintained based
on appropriate criteria such as the number of hops, bottleneck link bandwidth
and/or delay. It is worth recalling that the use of deterministic paths, such as the
Label Switched Paths (LSPs) in Multiprotocol Label Switching (MPLS), is very
important to improve network resource control and the traffic engineering under
resource over-reservation. Besides the LSPs, deterministic paths can be maintained
also by properly configuring a unique multicast channel (Source, Group) on each
path, as in [
19
]; the Source may be the IP address of the ingress ER where the path
originates from and the Group may be a standard multicast address. Further, the
ARC computes initial over-reservations parameters for the CoSs at each interface
on each path and sends the information to the nodes on the path using the OpenFlow
protocol.
As a result, every visited node on the path intercepts the message and configures
its local interfaces and the MIB accordingly (e.g., OpenFlow tables, forwarding/
routing tables and resource over-reservation parameters). In this way, the network is
initialized and set to run. As the network is operating, every service or cooperative
communication request must be directed to the SDNC for the processing. This is to
assure service Authentication, Authorization, and Accounting (AAA), QoS and
'
