| With the development of network communication technologies,many distributed multimedia appliactions are becoming increasingly important in our life. These include audio and video on demand, video conference, distributed storage backups, software updates, and virtual reality games, to name a few. Multicast has always been a hot research topic as the key technology of these distributed real-time applications, because multicast allows reducing source server load and increasing network capacity savings. However, IP multicast is difficult to support for large scale real-world deploymentin the traditional network because the limitations of the current service model and architecture. For example, all underlying multicast routers must support multicast protocols, the shortesr-path based multicast tree cannot optimize network resources, and there is not effective control mechanism for multicast group.Software Defined Networking (SDN) decouples the control plane and the forwarding plane, and makes the deployment of multicast more promising through a centralized control. No distributed multicast tree constructin protocol is needed in SDN-based multicast. The multicast tree can be constructed in the controller and the multicast forwarding devices (OpenFlow switches etc.) just make fast forwarding of multicast packets according to the forwarding roles in the flow table arranged by the controller. As we know, the forwarding devices use expensive resources and technologies (like TCAM), and due to this the size of flow table is limited. In other words, the number of forwarding states is limited. Huge amount of forwarding states will be created when there are a large number of multicast groups. So, there still exists the scalability problem.SDN-based multicast archtecture and some related technologies are introduced in this paper first, and then a new scalable multicast tree model, the Degree-dependent Branch-node Weighted Steiner Tree (DBWST) problem, is proposed in connection with the limited flow table size.The DBWST problem is NP-hard. It aims to minimize the total cost of edges and branch nodes. The cost of a branch node is degree-dependent. The forwarding states are stored only in the branch nodes, which can reduce the number of multicast states effectively. The degree-dependent branch node cost can banlance the work load avoiding the case of bottleneck and restrict the multicast states in the branch nodes. The communication between two branch nodes can use tunnel technology.We use an extended SPH algorithm (ESPH) to get an approximate DBWST firstly, which is time-efficient. In ESPH, destination nodes join the existing multicast tree dynamically and it is helpful to achieve dynamic multicast.Simulation results show that ESPH can get an approximate DBWST efficiently, the solution can reduce multicast flow table effectively, and the degree dependent cost makes the load balance of branch nodes.In order to get a better tree, we design an approximation algorithm, which named Path-Vector based Harmony Search Algorithm (PVHS), to solve this problem. Path vector representation is designed to construct the multicast memory. The path vector means a solution vector in a harmony and denotes the ordered set of nodes from source to a destination in the multicast tree. Multicast memory considering, local adjustment of path vector and generating the path vector randomly is used in the iterative updating. Simulation results on randomly generated topologies indicate that the trees obtained by PVHS are more cost-efficient... |
PDF Full Text Request