1 | """Module Graph of kytos/pathfinder Kytos Network Application.""" |
||
2 | |||
3 | 1 | # pylint: disable=too-many-arguments,too-many-locals |
|
4 | from itertools import combinations, islice |
||
5 | 1 | ||
6 | 1 | from kytos.core import log |
|
7 | 1 | from kytos.core.common import EntityStatus |
|
8 | from napps.kytos.pathfinder.utils import (filter_ge, filter_in, filter_le, |
||
9 | lazy_filter, nx_edge_data_delay, |
||
10 | nx_edge_data_priority, |
||
11 | nx_edge_data_weight) |
||
12 | |||
13 | 1 | try: |
|
14 | import networkx as nx |
||
15 | from networkx.exception import NetworkXNoPath, NodeNotFound |
||
16 | 1 | except ImportError: |
|
17 | 1 | PACKAGE = "networkx==2.5.1" |
|
18 | log.error(f"Package {PACKAGE} not found. Please 'pip install {PACKAGE}'") |
||
19 | 1 | ||
20 | |||
21 | 1 | class KytosGraph: |
|
22 | """Class responsible for the graph generation.""" |
||
23 | 1 | ||
24 | def __init__(self): |
||
25 | 1 | self.graph = nx.Graph() |
|
26 | 1 | self._filter_functions = { |
|
27 | 1 | "ownership": lazy_filter(str, filter_in("ownership")), |
|
28 | "bandwidth": lazy_filter((int, float), filter_ge("bandwidth")), |
||
29 | 1 | "reliability": lazy_filter((int, float), filter_ge("reliability")), |
|
30 | "priority": lazy_filter((int, float), filter_le("priority")), |
||
31 | 1 | "utilization": lazy_filter((int, float), filter_le("utilization")), |
|
32 | 1 | "delay": lazy_filter((int, float), filter_le("delay")), |
|
33 | 1 | } |
|
34 | self.spf_edge_data_cbs = { |
||
35 | 1 | "hop": nx_edge_data_weight, |
|
36 | 1 | "delay": nx_edge_data_delay, |
|
37 | 1 | "priority": nx_edge_data_priority, |
|
38 | } |
||
39 | |||
40 | def clear(self): |
||
41 | """Remove all nodes and links registered.""" |
||
42 | 1 | self.graph.clear() |
|
43 | |||
44 | 1 | def update_topology(self, topology): |
|
45 | 1 | """Update all nodes and links inside the graph.""" |
|
46 | 1 | self.graph.clear() |
|
47 | 1 | self.update_nodes(topology.switches) |
|
48 | 1 | self.update_links(topology.links) |
|
49 | 1 | ||
50 | 1 | def update_nodes(self, nodes): |
|
51 | 1 | """Update all nodes inside the graph.""" |
|
52 | 1 | for node in nodes.values(): |
|
53 | try: |
||
54 | 1 | if node.status != EntityStatus.UP: |
|
55 | continue |
||
56 | 1 | self.graph.add_node(node.id) |
|
57 | |||
58 | for interface in node.interfaces.values(): |
||
59 | if interface.status == EntityStatus.UP: |
||
60 | self.graph.add_node(interface.id) |
||
61 | self.graph.add_edge(node.id, interface.id) |
||
62 | 1 | ||
63 | 1 | except AttributeError as err: |
|
64 | 1 | raise TypeError( |
|
65 | 1 | f"Error when updating nodes inside the graph: {str(err)}" |
|
66 | ) |
||
67 | 1 | ||
68 | def update_links(self, links): |
||
69 | """Update all links inside the graph.""" |
||
70 | 1 | for link in links.values(): |
|
71 | 1 | if link.status == EntityStatus.UP: |
|
72 | 1 | self.graph.add_edge(link.endpoint_a.id, link.endpoint_b.id) |
|
73 | self.update_link_metadata(link) |
||
74 | 1 | ||
75 | def update_link_metadata(self, link): |
||
76 | 1 | """Update link metadata.""" |
|
77 | 1 | for key, value in link.metadata.items(): |
|
78 | if key not in self._filter_functions: |
||
79 | continue |
||
80 | endpoint_a = link.endpoint_a.id |
||
81 | endpoint_b = link.endpoint_b.id |
||
82 | 1 | self.graph[endpoint_a][endpoint_b][key] = value |
|
83 | |||
84 | def get_link_metadata(self, endpoint_a, endpoint_b): |
||
85 | """Return the metadata of a link.""" |
||
86 | return self.graph.get_edge_data(endpoint_a, endpoint_b) |
||
87 | |||
88 | @staticmethod |
||
89 | def _remove_switch_hops(circuit): |
||
90 | """Remove switch hops from a circuit hops list.""" |
||
91 | for hop in circuit["hops"]: |
||
92 | if len(hop.split(":")) == 8: |
||
93 | circuit["hops"].remove(hop) |
||
94 | |||
95 | def _path_cost(self, path, weight="hop", default_cost=1): |
||
96 | """Compute the path cost given an attribute.""" |
||
97 | cost = 0 |
||
98 | for node, nbr in nx.utils.pairwise(path): |
||
99 | cost += self.graph[node][nbr].get(weight, default_cost) |
||
100 | return cost |
||
101 | |||
102 | def path_cost_builder(self, paths, weight="hop", default_weight=1): |
||
103 | """Build the cost of a path given a list of paths.""" |
||
104 | paths_acc = [] |
||
105 | for path in paths: |
||
106 | if isinstance(path, list): |
||
107 | paths_acc.append( |
||
108 | { |
||
109 | "hops": path, |
||
110 | "cost": self._path_cost( |
||
111 | path, weight=weight, default_cost=default_weight |
||
112 | ), |
||
113 | } |
||
114 | ) |
||
115 | elif isinstance(path, dict): |
||
116 | path["cost"] = self._path_cost( |
||
117 | path["hops"], weight=weight, default_cost=default_weight |
||
118 | ) |
||
119 | paths_acc.append(path) |
||
120 | else: |
||
121 | raise TypeError( |
||
122 | f"type: '{type(path)}' must be be either list or dict. " |
||
123 | f"path: {path}" |
||
124 | ) |
||
125 | return paths_acc |
||
126 | |||
127 | def k_shortest_paths( |
||
128 | self, source, destination, weight=None, k=1, graph=None |
||
129 | ): |
||
130 | """ |
||
131 | Compute up to k shortest paths and return them. |
||
132 | |||
133 | This procedure is based on algorithm by Jin Y. Yen [1]. |
||
134 | Since Yen's algorithm calls Dijkstra's up to k times, the time |
||
135 | complexity will be proportional to K * Dijkstra's, average |
||
136 | O(K(|V| + |E|)logV), assuming it's using a heap, where V is the |
||
137 | number of vertices and E number of egdes. |
||
138 | |||
139 | References |
||
140 | ---------- |
||
141 | .. [1] Jin Y. Yen, "Finding the K Shortest Loopless Paths in a |
||
142 | Network", Management Science, Vol. 17, No. 11, Theory Series |
||
143 | (Jul., 1971), pp. 712-716. |
||
144 | """ |
||
145 | try: |
||
146 | return list( |
||
147 | islice( |
||
148 | nx.shortest_simple_paths( |
||
149 | graph or self.graph, |
||
150 | source, |
||
151 | destination, |
||
152 | weight=weight, |
||
153 | ), |
||
154 | k, |
||
155 | ) |
||
156 | ) |
||
157 | except (NodeNotFound, NetworkXNoPath): |
||
158 | return [] |
||
159 | |||
160 | def constrained_k_shortest_paths( |
||
161 | self, |
||
162 | source, |
||
163 | destination, |
||
164 | weight=None, |
||
165 | k=1, |
||
166 | graph=None, |
||
167 | minimum_hits=None, |
||
168 | **metrics, |
||
169 | ): |
||
170 | """Calculate the constrained shortest paths with flexibility.""" |
||
171 | graph = graph or self.graph |
||
172 | mandatory_metrics = metrics.get("mandatory_metrics", {}) |
||
173 | flexible_metrics = metrics.get("flexible_metrics", {}) |
||
174 | first_pass_links = list( |
||
175 | self._filter_links( |
||
176 | graph.edges(data=True), **mandatory_metrics |
||
177 | ) |
||
178 | ) |
||
179 | length = len(flexible_metrics) |
||
180 | if minimum_hits is None: |
||
181 | minimum_hits = 0 |
||
182 | minimum_hits = min(length, max(0, minimum_hits)) |
||
183 | |||
184 | paths = [] |
||
185 | for i in range(length, minimum_hits - 1, -1): |
||
186 | for combo in combinations(flexible_metrics.items(), i): |
||
187 | additional = dict(combo) |
||
188 | filtered_links = self._filter_links( |
||
189 | first_pass_links, **additional |
||
190 | ) |
||
191 | filtered_links = ((u, v) for u, v, d in filtered_links) |
||
0 ignored issues
–
show
introduced
by
![]() |
|||
192 | for path in self.k_shortest_paths( |
||
193 | source, |
||
194 | destination, |
||
195 | weight=weight, |
||
196 | k=k, |
||
197 | graph=graph.edge_subgraph(filtered_links), |
||
198 | ): |
||
199 | paths.append( |
||
200 | { |
||
201 | "hops": path, |
||
202 | "metrics": {**mandatory_metrics, **additional}, |
||
203 | } |
||
204 | ) |
||
205 | if len(paths) == k: |
||
206 | return paths |
||
207 | if paths: |
||
208 | return paths |
||
209 | return paths |
||
210 | |||
211 | def _filter_links(self, links, **metrics): |
||
212 | for metric, value in metrics.items(): |
||
213 | filter_func = self._filter_functions.get(metric, None) |
||
214 | if filter_func is not None: |
||
215 | try: |
||
216 | links = filter_func(value, links) |
||
217 | except TypeError as err: |
||
218 | raise TypeError( |
||
219 | f"Error in {metric} value: {value} err: {err}" |
||
220 | ) |
||
221 | return links |
||
222 |