preview

Analysis Of Local Search Algorithm For STP

Decent Essays

From the tree SP we presented in the algorithm that we have obtained via Local Search Algorithm for STP, we have generated the matrix of cost. This is done by assigning a cost to all the edges of tree SP and by assigning a cost on “n” no. of nodes to all the other edges in graph. This assignment of cost helps in recognizing the cost of the longest possible path between a pair of nodes in any spanning tree is n−1 (i.e. it passes n−1 edges) while the cost of the shortest path between any pair of nodes without using of SPT edges is at least “n” (i.e. passes one edge). Consequently, the 802.1d protocol will produce the intended spanning tree “SP”.
3.5 DATA GENERATION
In this section we progress by generating network topologies and traffic …show more content…

root = 1; in_tree = {root}; considered = ∅; while #in_tree< n do select (u ∈in_tree) and (u !∈ considered); selectnum_branch∈ [min..max] ; foreach i ∈ [1..num_branch] do if #in_tree< n then select (v ∈ [1..n]) and (u /∈in_tree); creatEdge(u, v); in_tree = in_tree + {v} end end considered = considered + u; end To the obtained spanning tree from above algorithm we add two types of edges so that we can get a bi-connected graph. The bi-connected graph has a significance that if any of the edge becomes down then also the network will be connected via another edge. This gives us assurance of always up time for a network. This means in case of link failure alternate link will always be present to ensure the network connectivity.

In this type1 edge connect a leaf with the higher level node while the type 2 edge connect a non- leaf node (not the root) with the no-leaf node or lower level node of different branch. For each tree new “n-1” edges are added while the generation of bi-connected graph.
To pretend a network in which a switch has many ports, we define a ratio “r”. This means each node in the tree is connected to at least “r” edges. In each test graph, from the generated bi-connected graph, we create three more trees with ratio r15 = n/15, r10 = n/10 and r5 = n/5 (where n = no. of nodes).

3.5.4 The FAT Tree:
Figure shown below depicts the Fat Tree - another topology for DCNs proposed in [35] It is called Fat Tree because it is not a

Get Access