1 | """Module Graph of kytos/pathfinder Kytos Network Application.""" |
||
2 | |||
3 | # pylint: disable=too-many-arguments,too-many-locals |
||
4 | 1 | from itertools import combinations, islice |
|
5 | |||
6 | 1 | from kytos.core import log |
|
7 | 1 | from napps.kytos.pathfinder.utils import (filter_ge, filter_in, filter_le, |
|
8 | lazy_filter, nx_edge_data_delay, |
||
9 | nx_edge_data_priority, |
||
10 | nx_edge_data_weight) |
||
11 | |||
12 | 1 | try: |
|
13 | 1 | import networkx as nx |
|
14 | 1 | from networkx.exception import NetworkXNoPath, NodeNotFound |
|
15 | except ImportError: |
||
16 | PACKAGE = "networkx==2.5.1" |
||
17 | log.error(f"Package {PACKAGE} not found. Please 'pip install {PACKAGE}'") |
||
18 | |||
19 | |||
20 | 1 | class KytosGraph: |
|
21 | """Class responsible for the graph generation.""" |
||
22 | |||
23 | 1 | def __init__(self): |
|
24 | 1 | self.graph = nx.Graph() |
|
25 | 1 | self._filter_functions = { |
|
26 | "ownership": lazy_filter(str, filter_in("ownership")), |
||
27 | "bandwidth": lazy_filter((int, float), filter_ge("bandwidth")), |
||
28 | "reliability": lazy_filter((int, float), filter_ge("reliability")), |
||
29 | "priority": lazy_filter((int, float), filter_le("priority")), |
||
30 | "utilization": lazy_filter((int, float), filter_le("utilization")), |
||
31 | "delay": lazy_filter((int, float), filter_le("delay")), |
||
32 | } |
||
33 | 1 | self.spf_edge_data_cbs = { |
|
34 | "hop": nx_edge_data_weight, |
||
35 | "delay": nx_edge_data_delay, |
||
36 | "priority": nx_edge_data_priority, |
||
37 | } |
||
38 | |||
39 | 1 | def clear(self): |
|
40 | """Remove all nodes and links registered.""" |
||
41 | 1 | self.graph.clear() |
|
42 | |||
43 | 1 | def update_topology(self, topology): |
|
44 | """Update all nodes and links inside the graph.""" |
||
45 | 1 | self.graph.clear() |
|
46 | 1 | self.update_nodes(topology.switches) |
|
47 | 1 | self.update_links(topology.links) |
|
48 | |||
49 | 1 | def update_nodes(self, nodes): |
|
50 | """Update all nodes inside the graph.""" |
||
51 | 1 | for node in nodes.values(): |
|
52 | 1 | try: |
|
53 | 1 | self.graph.add_node(node.id) |
|
54 | |||
55 | 1 | for interface in node.interfaces.values(): |
|
56 | 1 | self.graph.add_node(interface.id) |
|
57 | 1 | self.graph.add_edge(node.id, interface.id) |
|
58 | |||
59 | 1 | except AttributeError as err: |
|
60 | 1 | raise TypeError( |
|
61 | f"Error when updating nodes inside the graph: {str(err)}" |
||
62 | ) |
||
63 | |||
64 | 1 | def update_links(self, links): |
|
65 | """Update all links inside the graph.""" |
||
66 | 1 | for link in links.values(): |
|
67 | 1 | if link.is_active(): |
|
68 | 1 | self.graph.add_edge(link.endpoint_a.id, link.endpoint_b.id) |
|
69 | 1 | self.update_link_metadata(link) |
|
70 | |||
71 | 1 | def update_link_metadata(self, link): |
|
72 | """Update link metadata.""" |
||
73 | 1 | for key, value in link.metadata.items(): |
|
74 | 1 | if key not in self._filter_functions: |
|
75 | 1 | continue |
|
76 | 1 | endpoint_a = link.endpoint_a.id |
|
77 | 1 | endpoint_b = link.endpoint_b.id |
|
78 | 1 | self.graph[endpoint_a][endpoint_b][key] = value |
|
79 | |||
80 | 1 | def get_link_metadata(self, endpoint_a, endpoint_b): |
|
81 | """Return the metadata of a link.""" |
||
82 | 1 | return self.graph.get_edge_data(endpoint_a, endpoint_b) |
|
83 | |||
84 | 1 | @staticmethod |
|
85 | def _remove_switch_hops(circuit): |
||
86 | """Remove switch hops from a circuit hops list.""" |
||
87 | 1 | for hop in circuit["hops"]: |
|
88 | 1 | if len(hop.split(":")) == 8: |
|
89 | 1 | circuit["hops"].remove(hop) |
|
90 | |||
91 | 1 | def _path_cost(self, path, weight="hop", default_cost=1): |
|
92 | """Compute the path cost given an attribute.""" |
||
93 | 1 | cost = 0 |
|
94 | 1 | for node, nbr in nx.utils.pairwise(path): |
|
95 | 1 | cost += self.graph[node][nbr].get(weight, default_cost) |
|
96 | 1 | return cost |
|
97 | |||
98 | 1 | def path_cost_builder(self, paths, weight="hop", default_weight=1): |
|
99 | """Build the cost of a path given a list of paths.""" |
||
100 | 1 | paths_acc = [] |
|
101 | 1 | for path in paths: |
|
102 | 1 | if isinstance(path, list): |
|
103 | 1 | paths_acc.append( |
|
104 | { |
||
105 | "hops": path, |
||
106 | "cost": self._path_cost( |
||
107 | path, weight=weight, default_cost=default_weight |
||
108 | ), |
||
109 | } |
||
110 | ) |
||
111 | 1 | elif isinstance(path, dict): |
|
112 | 1 | path["cost"] = self._path_cost( |
|
113 | path["hops"], weight=weight, default_cost=default_weight |
||
114 | ) |
||
115 | 1 | paths_acc.append(path) |
|
116 | else: |
||
117 | raise TypeError( |
||
118 | f"type: '{type(path)}' must be be either list or dict. " |
||
119 | f"path: {path}" |
||
120 | ) |
||
121 | 1 | return paths_acc |
|
122 | |||
123 | 1 | def k_shortest_paths( |
|
124 | self, source, destination, weight=None, k=1, graph=None |
||
125 | ): |
||
126 | """ |
||
127 | Compute up to k shortest paths and return them. |
||
128 | |||
129 | This procedure is based on algorithm by Jin Y. Yen [1]. |
||
130 | Since Yen's algorithm calls Dijkstra's up to k times, the time |
||
131 | complexity will be proportional to K * Dijkstra's, average |
||
132 | O(K(|V| + |E|)logV), assuming it's using a heap, where V is the |
||
133 | number of vertices and E number of egdes. |
||
134 | |||
135 | References |
||
136 | ---------- |
||
137 | .. [1] Jin Y. Yen, "Finding the K Shortest Loopless Paths in a |
||
138 | Network", Management Science, Vol. 17, No. 11, Theory Series |
||
139 | (Jul., 1971), pp. 712-716. |
||
140 | """ |
||
141 | 1 | try: |
|
142 | 1 | return list( |
|
143 | islice( |
||
144 | nx.shortest_simple_paths( |
||
145 | graph or self.graph, |
||
146 | source, |
||
147 | destination, |
||
148 | weight=weight, |
||
149 | ), |
||
150 | k, |
||
151 | ) |
||
152 | ) |
||
153 | 1 | except (NodeNotFound, NetworkXNoPath): |
|
154 | 1 | return [] |
|
155 | |||
156 | 1 | def constrained_k_shortest_paths( |
|
157 | self, |
||
158 | source, |
||
159 | destination, |
||
160 | weight=None, |
||
161 | k=1, |
||
162 | minimum_hits=None, |
||
163 | **metrics, |
||
164 | ): |
||
165 | """Calculate the constrained shortest paths with flexibility.""" |
||
166 | 1 | mandatory_metrics = metrics.get("mandatory_metrics", {}) |
|
167 | 1 | flexible_metrics = metrics.get("flexible_metrics", {}) |
|
168 | 1 | first_pass_links = list( |
|
169 | self._filter_links( |
||
170 | self.graph.edges(data=True), **mandatory_metrics |
||
171 | ) |
||
172 | ) |
||
173 | 1 | length = len(flexible_metrics) |
|
174 | 1 | if minimum_hits is None: |
|
175 | 1 | minimum_hits = 0 |
|
176 | 1 | minimum_hits = min(length, max(0, minimum_hits)) |
|
177 | |||
178 | 1 | paths = [] |
|
179 | 1 | for i in range(length, minimum_hits - 1, -1): |
|
180 | 1 | for combo in combinations(flexible_metrics.items(), i): |
|
181 | 1 | additional = dict(combo) |
|
182 | 1 | filtered_links = self._filter_links( |
|
183 | first_pass_links, **additional |
||
184 | ) |
||
185 | 1 | filtered_links = ((u, v) for u, v, d in filtered_links) |
|
0 ignored issues
–
show
introduced
by
![]() |
|||
186 | 1 | for path in self.k_shortest_paths( |
|
187 | source, |
||
188 | destination, |
||
189 | weight=weight, |
||
190 | k=k, |
||
191 | graph=self.graph.edge_subgraph(filtered_links), |
||
192 | ): |
||
193 | 1 | paths.append( |
|
194 | { |
||
195 | "hops": path, |
||
196 | "metrics": {**mandatory_metrics, **additional}, |
||
197 | } |
||
198 | ) |
||
199 | 1 | if len(paths) == k: |
|
200 | 1 | return paths |
|
201 | 1 | if paths: |
|
202 | 1 | return paths |
|
203 | 1 | return paths |
|
204 | |||
205 | 1 | def _filter_links(self, links, **metrics): |
|
206 | 1 | for metric, value in metrics.items(): |
|
207 | 1 | filter_func = self._filter_functions.get(metric, None) |
|
208 | 1 | if filter_func is not None: |
|
209 | 1 | try: |
|
210 | 1 | links = filter_func(value, links) |
|
211 | 1 | except TypeError as err: |
|
212 | 1 | raise TypeError( |
|
213 | f"Error in {metric} value: {value} err: {err}" |
||
214 | ) |
||
215 | return links |
||
216 |