Source code for pyNTM.flex_model

"""
This is the flexible, more feature-rich model class.  It supports
more features than the PerformanceModel class.  For example, the
FlexModel class supports multiple Circuits (edges) between layer 3
Nodes.  This class will tend to support more topology features than
the PerformanceModel class.

If you are not sure whether to use the PerformanceModel or FlexModel object,
it's best to use the FlexModel object.

This Class is the same as the legacy (version 1.6 and earlier)
Parallel_Link_Model class.

This model type allows multiple links/parallel links between 2 nodes.

There will be a performance impact in this model variant.
"""

from datetime import datetime
from pprint import pprint

import itertools
import networkx as nx
import random

from .circuit import Circuit
from .interface import Interface
from .exceptions import ModelException
from .master_model import _MasterModel
from .rsvp import RSVP_LSP
from .utilities import find_end_index
from .node import Node


# TODO - call to analyze model for Unrouted LSPs and LSPs not on shortest path
# TODO - add support for SRLGs in load_model_file
# TODO - add attribute for Node/Interface whereby an object can be failed by itself
#  and not unfail when a parent SRLG unfails


[docs]class FlexModel(_MasterModel): """ A network model object consisting of the following base components: - Interface objects (set): layer 3 Node interfaces. Interfaces have a 'capacity' attribute that determines how much traffic it can carry. Note: Interfaces are matched into Circuit objects based on the interface circuit_ids --> A pair of Interfaces with the same circuit_id value get matched into a Circuit - Node objects (set): vertices on the network (aka 'layer 3 devices') that contain Interface objects. Nodes are connected to each other via a pair of matched Interfaces (Circuits) - Demand objects (set): traffic loads on the network. Each demand starts from a source node and transits the network to a destination node. A demand also has a magnitude, representing how much traffic it is carrying. The demand's magnitude will apply against each interface's available capacity - RSVP LSP objects (set): RSVP LSPs in the Model - Circuit objects are created by matching Interface objects using common circuit_id """ def __init__( self, interface_objects=set(), node_objects=set(), demand_objects=set(), rsvp_lsp_objects=set(), ): self.interface_objects = interface_objects self.node_objects = node_objects self.demand_objects = demand_objects self.circuit_objects = set() self.rsvp_lsp_objects = rsvp_lsp_objects self.srlg_objects = set() self._parallel_lsp_groups = {} super().__init__( interface_objects, node_objects, demand_objects, rsvp_lsp_objects ) def __repr__(self): return "FlexModel(Interfaces: %s, Nodes: %s, " "Demands: %s, RSVP_LSPs: %s)" % ( len(self.interface_objects), len(self.node_objects), len(self.demand_objects), len(self.rsvp_lsp_objects), )
[docs] def add_network_interfaces_from_list(self, network_interfaces): """ A tool that reads network interface info and updates an *existing* model. Intended to be used from CLI/interactive environment Interface info must be a list of dicts and in format like below example. Example:: network_interfaces = [ {'name':'A-to-B', 'cost':4,'capacity':100, 'node':'A', 'remote_node': 'B', 'circuit_id': 1, 'failed': False}, {'name':'A-to-Bv2', 'cost':40,'capacity':150, 'node':'A', 'remote_node': 'B', 'circuit_id': 2, 'failed': False}, {'name':'A-to-C', 'cost':1,'capacity':200, 'node':'A', 'remote_node': 'C', 'circuit_id': 3, 'failed': False},] :param network_interfaces: python list of attributes for Interface objects :return: self with new Interface objects """ new_interface_objects, new_node_objects = self._make_network_interfaces( network_interfaces ) self.node_objects = self.node_objects.union(new_node_objects) self.interface_objects = self.interface_objects.union(new_interface_objects) self.validate_model()
[docs] def validate_model(self): """ Validates that data fed into the model creates a valid network model """ # create circuits table, flags ints that are not part of a circuit circuits = self._make_circuits_multidigraph(return_exception=True) # Make dict to hold interface data, each entry has the following # format: # {'lsps': [], 'reserved_bandwidth': 0} int_info = self._make_int_info_dict() # Interface reserved bandwidth error sets int_res_bw_too_high = set([]) int_res_bw_sum_error = set([]) error_data = [] # list of all errored checks for interface in iter(self.interface_objects): # pragma: no cover self._reserved_bw_error_checks( int_info, int_res_bw_sum_error, int_res_bw_too_high, interface ) # If creation of circuits returns a dict, there are problems if isinstance(circuits, dict): # pragma: no cover error_data.append({"ints_w_no_remote_int": circuits["data"]}) # Append any failed checks to error_data if int_res_bw_too_high: # pragma: no cover error_data.append({"int_res_bw_too_high": int_res_bw_too_high}) if int_res_bw_sum_error: # pragma: no cover error_data.append({"int_res_bw_sum_error": int_res_bw_sum_error}) # Validate there are no duplicate interfaces unique_interfaces_per_node = self._unique_interface_per_node() # Log any duplicate interfaces on a node if not unique_interfaces_per_node: # pragma: no cover error_data.append(unique_interfaces_per_node) # Make validate_model() check for matching failed statuses # on the interfaces and matching interface capacity circuits_with_mismatched_interface_capacity = [] for ckt in iter(self.circuit_objects): self._validate_circuit_interface_capacity( circuits_with_mismatched_interface_capacity, ckt ) if circuits_with_mismatched_interface_capacity: int_status_error_dict = { "circuits_with_mismatched_interface_capacity": circuits_with_mismatched_interface_capacity } error_data.append(int_status_error_dict) # Validate Nodes in each SRLG have the SRLG in their srlgs set. # srlg_errors is a dict of node names as keys and a list of SRLGs that node is # a member of in the model but that the SRLG is not in node.srlgs srlg_errors = {} for ( srlg ) in ( self.srlg_objects ): # pragma: no cover # noqa # TODO - perhaps cover this later in unit testing nodes_in_srlg_but_srlg_not_in_node_srlgs = [ node for node in srlg.node_objects if srlg not in node.srlgs ] for node in nodes_in_srlg_but_srlg_not_in_node_srlgs: try: srlg_errors[node.name].append(srlg.name) except KeyError: srlg_errors[node.name] = [] if srlg_errors: error_data.append(srlg_errors) # Verify no duplicate nodes node_names = {node.name for node in self.node_objects} if (len(self.node_objects)) != (len(node_names)): # pragma: no cover node_dict = { "len_node_objects": len(self.node_objects), "len_node_names": len(node_names), } error_data.append(node_dict) # Read error_data if error_data: message = "network interface validation failed, see returned data" pprint(message) pprint(error_data) raise ModelException((message, error_data)) else: return self
[docs] def update_simulation(self): """ Updates the simulation state; this needs to be run any time there is a change to the state of the Model, such as failing an interface, adding a Demand, adding/removing and LSP, etc. This call does not carry forward any state from the previous simulation results. """ self._parallel_lsp_groups = {} # Reset the attribute # This set of interfaces can be used to route traffic non_failed_interfaces = set() # This set of nodes can be used to route traffic available_nodes = set() # Find all the non-failed interfaces in the model and # add them to non_failed_interfaces. # If the interface is not failed, then by definition, the nodes are # not failed for interface_object in ( interface_object for interface_object in self.interface_objects if interface_object.failed is not True ): non_failed_interfaces.add(interface_object) available_nodes.add(interface_object.node_object) available_nodes.add(interface_object.remote_node_object) # Create a model consisting only of the non-failed interfaces and # corresponding non-failed (available) nodes non_failed_interfaces_model = FlexModel( non_failed_interfaces, available_nodes, self.demand_objects, self.rsvp_lsp_objects, ) # Reset the reserved_bandwidth, traffic on each interface for interface in iter(self.interface_objects): interface.reserved_bandwidth = 0 interface.traffic = 0 for lsp in iter(self.rsvp_lsp_objects): lsp.path = "Unrouted" for demand in iter(self.demand_objects): demand.path = "Unrouted" time_before_lsp_load = datetime.now() print("Routing the LSPs . . . ") # Route the RSVP LSPs self = self._route_lsps() lsp_load_time = datetime.now() - time_before_lsp_load print( "LSPs routed (if present) in {}; routing demands now . . .".format( lsp_load_time ) ) # Route the demands demand_load_start_time = datetime.now() self = self._route_demands(non_failed_interfaces_model) demand_load_time = datetime.now() - demand_load_start_time print("Demands routed in {}; validating model . . . ".format(demand_load_time)) self.validate_model()
# TODO - for some reason this is getting called 2x when the model is being updated # initially. Troubleshoot that. def _route_demands(self, model): """ Routes demands in input 'model' :param model: input 'model' parameter object (may be different from self) :return: model with routed demands """ G = self._make_weighted_network_graph_mdg(include_failed_circuits=False) for demand in model.demand_objects: demand.path = [] # Find all LSPs that can carry the demand from source to dest: key = "{}-{}".format( demand.source_node_object.name, demand.dest_node_object.name ) try: lsp_list = [ lsp for lsp in self.parallel_lsp_groups()[key] if "Unrouted" not in lsp.path ] except KeyError: lsp_list = [] # Check for manually assigned metrics if len(lsp_list) > 0: min_lsp_metric = min([lsp.effective_metric(self) for lsp in lsp_list]) for lsp in lsp_list: if lsp.effective_metric(self) == min_lsp_metric: demand.path.append([lsp]) if demand.path == []: # There are no end to end LSPs for the demand src = demand.source_node_object.name dest = demand.dest_node_object.name # Shortest path in networkx multidigraph try: nx_sp = list(nx.all_shortest_paths(G, src, dest, weight="cost")) except nx.exception.NetworkXNoPath: # There is no path, demand.path = 'Unrouted' demand.path = "Unrouted" continue # all_paths is list of shortest paths from source to destination; these paths # may include paths that have multiple links between nodes all_paths = self._get_all_paths_mdg(G, nx_sp) # Make sure that each path in all_paths only has a single link # between each node. This is path normalization path_list = self._normalize_multidigraph_paths(all_paths) # Check for IGP shortcuts path_list = self.find_igp_shortcuts(path_list, nx_sp) demand.path = path_list self._update_interface_utilization() return self
[docs] def find_igp_shortcuts(self, paths, node_paths): """ Check for LSPs along the shortest path; find the first LSP the demand can take with a source and destination that is on the LSP's IGP path 1. examine each IGP path 2. If none of the nodes on the path have IGP shortcuts, continue to next path 3. If some nodes have IGP shortcuts enabled, note the hop number (1, 2, 3, etc) 4. For nodes that have IGP shortcuts, is there an LSP from that node to a downstream node on the path? - if yes, compare the IGP metric of the path to the LSP remote node to that of the LSP metric to that node - if no, look at next node downstream with IGP shortcuts 5. Look for manually set RSVP LSP metrics that may alter the path calculations :param paths: List of lists; each list contains egress Interfaces along the path from source to destination (ordered from source to destination) # noqa E501 :param node_paths: List of lists; each list contains node names along the path from source to destination (ordered from source to destination) :return: List of lists; each list contains Interfaces and/or RSVP LSPs along each path from source to destination # noqa E501 """ # Check node_paths for igp_shortcuts_enabled nodes # TODO - this can likely be optimized all_nodes_in_paths = [] for node_path in node_paths: all_nodes_in_paths = all_nodes_in_paths + node_path all_nodes_in_paths = set(all_nodes_in_paths) shortcut_enabled_nodes = [ node for node in all_nodes_in_paths if self.get_node_object(node).igp_shortcuts_enabled is True ] if len(shortcut_enabled_nodes) == 0: return paths # ## Find LSPs on shortcut_enabled_nodes that connect to downstream nodes in paths ## # # Substitute IGP enabled LSPs for Interfaces in paths paths_with_lsps = self._insert_lsp_shortcuts(node_paths, paths) if len(paths_with_lsps) == 1: return paths_with_lsps # Inspect paths to determine if manually set LSP metrics affect path selection finalized_paths = self._inspect_for_lsp_metrics(paths_with_lsps) return finalized_paths
def _insert_lsp_shortcuts(self, node_paths, paths): """ Given node_paths and paths, finds and inserts LSPs from shortcut-enabled nodes along the path :param paths: List of lists; each list contains egress Interfaces along the path from source to destination (ordered from source to destination) # noqa E501 :param node_paths: List of lists; each list contains node names along the path from source to destination (ordered from source to destination) :return: List of lists; each list is a path with any possible LSP shortcuts inserted in place of the any applicable Interfaces """ # Substitute IGP enabled LSPs for Interfaces in paths for node_path in node_paths: # Find Nodes along the path that have igp_shortcuts_enabled and have # LSPs to downstream Nodes in the path path_lsps = [] # List of LSPs to substitute into path next_node_to_check = [] # Next node name in path to check for LSPs for node_name in node_path: # Make sure the next node checked is downstream from the end of any LSPs # the traffic has taken thusfar if len(next_node_to_check) > 0: if node_path.index(node_name) < node_path.index( next_node_to_check[-1] ): continue if self.get_node_object(node_name).igp_shortcuts_enabled is True: # Get the source node object source_node = self.get_node_object(node_name) # Get the source node object index in node_path source_node_index = node_path.index(node_name) # Check for LSPs from present node in path (source_node) to downstream nodes in path; # look for the LSPs that go furthest downstream first destinations = node_path[source_node_index + 1 :] destinations.reverse() for destination in destinations: # Take the LSPs whose source node matches source_node and whose dest node matches # the destination we are iterating through and whose effective metric matches the # shortest path from source_node to destination key = "{}-{}".format(source_node.name, destination) try: candidate_lsps_for_demand = self.parallel_lsp_groups()[key] min_metric = min( lsp.effective_metric(self) for lsp in candidate_lsps_for_demand if "Unrouted" not in lsp.path ) lsps = [ lsp for lsp in candidate_lsps_for_demand if lsp.effective_metric(self) == min_metric and "Unrouted" not in lsp.path ] except (KeyError, ValueError): # If there is no LSP group that matches the demand source/dest (KeyError) or # there are no routed LSPs for the demand (ValueError), then set lsps # to empty list lsps = [] if lsps: # Break out of the destinations iteration as traffic will want to take # the first LSP(s) available the traffic farthest along the path path_lsps.append(lsps) lsp_end_node = lsps[0].dest_node_object.name next_node_to_check.append(lsp_end_node) break # Now that path_lsps is known, substitute those into path finalized_paths = [] if len(path_lsps) > 0: for interface_path in paths: finalized_path = self._insert_lsps_into_path(path_lsps, interface_path) if finalized_path != -1: for path in finalized_path: # finalized_path may be a list of lists, so add each component path if path not in finalized_paths: finalized_paths.append(path) else: finalized_paths.append(interface_path) else: # No LSPs available for shortcuts finalized_paths = paths return finalized_paths def _inspect_for_lsp_metrics(self, paths_with_lsps): """ Checks for manually set LSP metrics and how they affect best path """ if len(paths_with_lsps) <= 1: return # Metrics for each path path_metrics = set() # List with each path and its metric paths_with_metrics = [] for path in paths_with_lsps: path_metric = 0 for item in path: if isinstance(item, Interface): path_metric += item.cost elif isinstance(item, RSVP_LSP): path_metric += item.effective_metric(self) path_metrics.add(path_metric) paths_with_metrics.append([path, path_metric]) # See if all paths have same metric if len(path_metrics) == 1: return paths_with_lsps else: lowest_metric = min(path_metrics) return [ path_and_metric[0] for path_and_metric in paths_with_metrics if path_and_metric[1] == lowest_metric ] def _update_interface_utilization(self): """Updates each interface's utilization; returns Model object with updated interface utilization.""" # In the model, in an interface is failed, set the traffic attribute # to 'Down', otherwise, initialize the traffic to zero for interface_object in self.interface_objects: interface_object.traffic = "Down" if interface_object.failed else 0.0 routed_demand_object_generator = ( demand_object for demand_object in self.demand_objects if "Unrouted" not in demand_object.path ) # For each demand that is not Unrouted, add its traffic value to each # interface object in the path for demand_object in routed_demand_object_generator: # Expand each LSP into its interfaces and add that the traffic per LSP # to the LSP's path interfaces. # Can demand take LSP? # Is there a parallel_lsp_group that matches the source and dest for the demand_object? key = "{}-{}".format( demand_object.source_node_object.name, demand_object.dest_node_object.name, ) # Find the routed LSPs that can carry the demand try: candidate_lsps_for_demand = self.parallel_lsp_groups()[key] min_metric = min( lsp.effective_metric(self) for lsp in candidate_lsps_for_demand if "Unrouted" not in lsp.path ) lsps_for_demand = [ lsp for lsp in candidate_lsps_for_demand if lsp.effective_metric(self) == min_metric and "Unrouted" not in lsp.path ] except (KeyError, ValueError): # If there is no LSP group that matches the demand source/dest (KeyError) or there are no routed # LSPs for the demand (ValueError), then set lsps_for_demand to empty list lsps_for_demand = [] if lsps_for_demand != []: self._update_int_traffic_for_end_to_end_lsps( demand_object, lsps_for_demand ) # If demand_object is not taking LSPs end to end, IGP route it, using hop by hop ECMP else: # demand_traffic_per_int will be dict of # ('source_node_name-dest_node_name': <traffic from demand>) k,v pairs # # Example: The interface from node G to node D has 2.5 units of traffic from 'demand' # {'G-D': 2.5, 'A-B': 10.0, 'B-D': 2.5, 'A-D': 5.0, 'D-F': 10.0, 'B-G': 2.5} demand_traffic_per_item = self._demand_traffic_per_item(demand_object) for item, traffic in demand_traffic_per_item.items(): if isinstance(item, Interface): for path, path_info in demand_object._path_detail.items(): if item in path_info["items"]: item.traffic += path_info["path_traffic"] elif isinstance(item, RSVP_LSP): # Get LSP interfaces interfaces = item.path["interfaces"] for interface in interfaces: # Add traffic to the Interface interface.traffic += traffic return self def _update_int_traffic_for_end_to_end_lsps(self, demand_object, lsps_for_demand): """ For a given Demand that is taking a single RSVP LSP end to end, update the traffic on the LSP's interfaces within self. :param demand_object: Demand that is traveling end to end on a single LSP :param lsps_for_demand: List of parallel LSPs that transport demand_object end to end. The demand_object splits its traffic evenly across the LSPs :return: None; interface traffic updated within self """ # demand_object takes LSP end to end. # Find each demand's path list, determine the ECMP split across the # routed LSPs, and find the traffic per path (LSP) num_routed_lsps_for_demand = len(lsps_for_demand) traffic_per_demand_path = demand_object.traffic / num_routed_lsps_for_demand path_detail = {} for lsp in lsps_for_demand: # Build the path detail: lsp and path traffic path_detail["path_{}".format(lsps_for_demand.index(lsp))] = {} path_detail["path_{}".format(lsps_for_demand.index(lsp))]["items"] = [lsp] path_detail["path_{}".format(lsps_for_demand.index(lsp))][ "path_traffic" ] = traffic_per_demand_path # Add detailed path info to demand demand_object._path_detail = path_detail # Get the interfaces for each LSP in the demand's path for lsp in lsps_for_demand: try: lsp_path_interfaces = lsp.path["interfaces"] except TypeError: # Will error if lsp.path == 'Unrouted' pass # Now that all interfaces are known, # update traffic on interfaces demand touches for interface in lsp_path_interfaces: # Get the interface's existing traffic and add the # portion of the demand's traffic interface.traffic += traffic_per_demand_path def _demand_traffic_per_item(self, demand): """ Given a Demand object, return the (key, value) pairs for how much traffic each Interface/LSP gets from the routing of the traffic load over Model Interfaces. : demand: Demand object : return: dict of (Interface: <traffic from demand> ) k, v pairs Example:: The interface from node B to E (name='B-to-E') below has 8.0 units of traffic from 'demand'; the interface from A to B has 12.0, etc. {Interface(name = 'A-to-B', cost = 4, capacity = 100, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = '1'): 12.0, Interface(name = 'A-to-B_2', cost = 4, capacity = 50, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = '2'): 12.0, Interface(name = 'B-to-E', cost = 3, capacity = 200, node_object = Node('B'), remote_node_object = Node('E'), circuit_id = '7'): 8.0, Interface(name = 'B-to-E_3', cost = 3, capacity = 200, node_object = Node('B'), remote_node_object = Node('E'), circuit_id = '27'): 8.0, Interface(name = 'B-to-E_2', cost = 3, capacity = 200, node_object = Node('B'), remote_node_object = Node('E'), circuit_id = '17'): 8.0} """ shortest_path_item_list = [] for path in demand.path: shortest_path_item_list += path # Unique interfaces across all shortest paths # shortest_path_int_set = set(shortest_path_int_list) shortest_path_item_set = set(shortest_path_item_list) unique_next_hops = self._find_unique_next_hops(shortest_path_item_set) # shortest_path_info will be a dict with the following info for each path: # - an ordered list of interfaces in the path # - a dict of cumulative splits for each interface at that point in the path # - the amount of traffic on the path # Example: # shortest_path_info = # {'path_0': {'interfaces': [ # Interface(name='A-to-B_2', cost=4, capacity=50, node_object=Node('A'), remote_node_object=Node('B'), # circuit_id='2'), # Interface(name='B-to-E_2', cost=3, capacity=200, node_object=Node('B'), remote_node_object=Node('E'), # circuit_id='17')], # 'path_traffic': 4.0, # 'splits': {Interface(name='A-to-B_2', cost=4, capacity=50, node_object=Node('A'), # remote_node_object=Node('B'), circuit_id='2'): 2, # Interface(name='B-to-E_2', cost=3, capacity=200, node_object=Node('B'), # remote_node_object=Node('E'), circuit_id='17'): 6}}, # 'path_1': {'interfaces': [ # Interface(name='A-to-B_2', cost=4, capacity=50, node_object=Node('A'), remote_node_object=Node('B'), # circuit_id='2'), # Interface(name='B-to-E', cost=3, capacity=200, node_object=Node('B'), remote_node_object=Node('E'), # circuit_id='7')], # 'path_traffic': 4.0, # 'splits': {Interface(name='A-to-B_2', cost=4, capacity=50, node_object=Node('A'), # remote_node_object=Node('B'), circuit_id='2'): 2, # Interface(name='B-to-E', cost=3, capacity=200, node_object=Node('B'), # remote_node_object=Node('E'), circuit_id='7'): 6}}} shortest_path_info = {} path_counter = 0 # Iterate thru each path for the demand to get traffic per interface # and the splits for each demand for path in demand.path: # Dict of cumulative splits per interface traffic_splits_per_interface = {} path_key = "path_" + str(path_counter) shortest_path_info[path_key] = {} # Create cumulative path splits for each item in the path total_splits = 1 # Normalize any paths with LSPs for item in path: # Update the total cumulative splits in the path before # traffic reaches the item in the path if isinstance(item, Interface): total_splits = total_splits * len( unique_next_hops[item.node_object.name] ) elif isinstance(item, RSVP_LSP): total_splits = total_splits * len( unique_next_hops[item.source_node_object.name] ) traffic_splits_per_interface[item] = total_splits # Find path traffic max_split = max([split for split in traffic_splits_per_interface.values()]) path_traffic = float(demand.traffic) / float(max_split) shortest_path_info[path_key]["items"] = path shortest_path_info[path_key]["splits"] = traffic_splits_per_interface shortest_path_info[path_key]["path_traffic"] = path_traffic path_counter += 1 # For each path, determine which interfaces it transits and add # that path's traffic to the interface # Create dict to hold cumulative traffic for each interface for demand traff_per_int = dict.fromkeys(shortest_path_item_set, 0) for path, info in shortest_path_info.items(): for interface in info["items"]: traff_per_int[interface] += info["path_traffic"] # Round all traffic values to 1 decimal place traff_per_int = { interface: round(traffic, 1) for interface, traffic in traff_per_int.items() } demand._path_detail = shortest_path_info return traff_per_int def _find_unique_next_hops(self, shortest_path_item_set): """ From a set of items from all the shortest paths, determine how many unique next hops there are from a given node. :param shortest_path_item_set: a set of items (Interfaces, RSVP_LSPs) from all the shortest paths :return: a dict with keys for each Node name and values being a list of each unique next hop from that Node In the example output below, Node B has 2 unique next hops, buth of which are RSVP LSPs Example output:: {'A': [Interface(name = 'A-G', cost = 25, capacity = 100, node_object = Node('A'), remote_node_object = Node('G'), circuit_id = '6'), Interface(name = 'A-B', cost = 10, capacity = 100, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = '1')], 'B': [RSVP_LSP(source = B, dest = D, lsp_name = 'lsp_b_d_1'), RSVP_LSP(source = B, dest = D, lsp_name = 'lsp_b_d_2')], 'D': [RSVP_LSP(source = D, dest = F, lsp_name = 'lsp_d_f_1')], 'G': [Interface(name = 'G-F', cost = 25, capacity = 100, node_object = Node('G'), remote_node_object = Node('F'), circuit_id = '7')]} """ # Dict to store how many unique next hops each node has in the shortest paths unique_next_hops = {} # Iterate through all the items for item in shortest_path_item_set: if isinstance(item, Interface): unique_next_hops[item.node_object.name] = [] # For a given Interface's node_object, determine how many # Interfaces on that Node are facing next hops for hop in shortest_path_item_set: if ( isinstance(hop, Interface) and hop.node_object.name == item.node_object.name or not isinstance(hop, Interface) and isinstance(hop, RSVP_LSP) and hop.source_node_object.name == item.node_object.name ): unique_next_hops[item.node_object.name].append(hop) elif isinstance(item, RSVP_LSP): unique_next_hops[item.source_node_object.name] = [] # For an LSP's source_node_object, for hop in shortest_path_item_set: if ( isinstance(hop, Interface) and hop.node_object.name == item.source_node_object.name or not isinstance(hop, Interface) and isinstance(hop, RSVP_LSP) and hop.source_node_object.name == item.source_node_object.name ): unique_next_hops[item.source_node_object.name].append(hop) return unique_next_hops def _insert_lsps_into_path(self, path_lsps, path): """ Substitutes the path_lsps into the path. Although the path argument is a single path, multiple paths will be returned if there are multiple parallel LSPs (in path_lsps component lists) between any hops in the path. :param path_lsps: List of lists. Each component list holds the parallel LSPs from a common source to a common destination Example:: [[RSVP_LSP(source = B, dest = D, lsp_name = 'lsp_b_d_2'), RSVP_LSP(source = B, dest = D, lsp_name = 'lsp_b_d_1')], [RSVP_LSP(source = D, dest = F, lsp_name = 'lsp_d_f_1')]] :param path: List of Interfaces in path from source to destination :return: List of path permutations with the LSPs substituted in """ # path_slices is a list of lists; each component list is a tuple # for an LSP substitution in the path: # [(start_index, end_index, parallel_lsp_1), . . # . . ((start_index, end_index, parallel_lsp_x)] # Example: 2 LSPs from B to D and 1 LSP from D to F # [[[1, 3, RSVP_LSP(source = B, dest = D, lsp_name = 'lsp_b_d_2')], # [1, 3, RSVP_LSP(source = B, dest = D, lsp_name = 'lsp_b_d_1')]], # [[3, 5, RSVP_LSP(source = D, dest = F, lsp_name = 'lsp_d_f_1')]]] path_slices = [] for lsp_group in path_lsps: # lsp_group_slices is a list of tuples for each parallel LSP in lsp_group: # [(start_index, end_index, parallel_lsp_1),. . # . . ((start_index, end_index, parallel_lsp_x)] lsp_group_slices = [] try: start_interface = [ interface for interface in path if isinstance(interface, Interface) and interface.node_object == lsp_group[0].source_node_object ][0] end_interface = [ interface for interface in path if isinstance(interface, Interface) and interface.remote_node_object == lsp_group[0].dest_node_object ][0] except IndexError: # There is no LSP source/dest match for the Interfaces in path; this may happen # with Demands that take ECMP paths, where one path has LSP shortcuts and other # paths do not have LSP shortcuts return -1 # code for no LSPs on path slice_to_sub_start_index = path.index(start_interface) slice_to_sub_end_index = path.index(end_interface) + 1 for lsp in lsp_group: lsp_group_slices.append( [slice_to_sub_start_index, slice_to_sub_end_index, lsp] ) path_slices.append(lsp_group_slices) # Sub in the LSPs, starting from the end of the path (to preserve index values) path_slices.reverse() path_slice_combos = list(itertools.product(*path_slices)) # List to hold the path(s) with the LSP(s) substituted in paths_with_substitutions = [] for path_slice_combo in path_slice_combos: path_prime = path[:] for path_slice in path_slice_combo: start_index = path_slice[0] end_index = path_slice[1] lsp = path_slice[2] path_prime[start_index:end_index] = [lsp] paths_with_substitutions.append(path_prime) return paths_with_substitutions def _get_all_paths_mdg(self, G, nx_sp): """ Examines hop-by-hop paths in G and determines specific edges transited from one hop to the next :param G: networkx multidigraph object containing nx_sp, contains Interface objects in edge data :param nx_sp: List of node paths in G Example:: nx_sp from A to D in graph G:: [['A', 'D'], ['A', 'B', 'D'], ['A', 'B', 'G', 'D']] :return: List of lists of possible specific model paths from source to destination nodes. Each 'hop' in a given path may include multiple possible Interfaces that could be transited from one node to the next adjacent node. Example:: all_paths from 'A' to 'D' is a list of lists; notice that there are two Interfacs that could be transited from Node 'B' to Node 'G' [[[Interface(name = 'A-to-D', cost = 40, capacity = 20.0, node_object = Node('A'), remote_node_object = Node('D'), circuit_id = 1)]], [[Interface(name = 'A-to-B', cost = 20, capacity = 125.0, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = 2)], [Interface(name = 'B-to-D', cost = 20, capacity = 125.0, node_object = Node('B'), remote_node_object = Node('D'), circuit_id = 3)]], [[Interface(name = 'A-to-B', cost = 20, capacity = 125.0, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = 4)], [Interface(name = 'B-to-G', cost = 10, capacity = 100.0, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = 5), Interface(name = 'B-to-G_2', cost = 10, capacity = 50.0, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = 6)], [Interface(name = 'G-to-D', cost = 10, capacity = 100.0, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = 7)]]] """ all_paths = [] for path in nx_sp: current_hop = path[0] this_path = [] for next_hop in path[1:]: this_hop = [] values_source_hop = G[current_hop][next_hop].values() min_weight = min(d["cost"] for d in values_source_hop) ecmp_links = [ interface_index for interface_index, interface_item in G[current_hop][ next_hop ].items() if interface_item["cost"] == min_weight ] # Add Interface(s) to this_hop list and add traffic to Interfaces for link_index in ecmp_links: this_hop.append(G[current_hop][next_hop][link_index]["interface"]) this_path.append(this_hop) current_hop = next_hop all_paths.append(this_path) return all_paths def _make_weighted_network_graph_mdg( self, include_failed_circuits=True, needed_bw=0, rsvp_required=False ): """ Returns a networkx weighted networkx multidigraph object from the input Model object :param include_failed_circuits: include interfaces from currently failed circuits in the graph? :param needed_bw: how much reservable_bandwidth is required? :param rsvp_required: True|False; only consider rsvp_enabled interfaces? :return: networkx multidigraph with edges that conform to the needed_bw and rsvp_required parameters """ G = nx.MultiDiGraph() # Get all the edges that meet 'failed' and 'reservable_bw' criteria if include_failed_circuits is False: considered_interfaces = ( interface for interface in self.interface_objects if ( interface.failed is False and interface.reservable_bandwidth >= needed_bw ) ) elif include_failed_circuits is True: considered_interfaces = ( interface for interface in self.interface_objects if interface.reservable_bandwidth >= needed_bw ) if rsvp_required is True: edge_names = ( ( interface.node_object.name, interface.remote_node_object.name, { "cost": interface.cost, "interface": interface, "circuit_id": interface.circuit_id, }, ) for interface in considered_interfaces if interface.rsvp_enabled is True ) else: edge_names = ( ( interface.node_object.name, interface.remote_node_object.name, { "cost": interface.cost, "interface": interface, "circuit_id": interface.circuit_id, }, ) for interface in considered_interfaces ) # Add edges to networkx DiGraph G.add_edges_from(edge_names) # Add all the nodes node_name_iterator = (node.name for node in self.node_objects) G.add_nodes_from(node_name_iterator) return G def _normalize_multidigraph_paths(self, path_info): # TODO - static? """ Takes the multidigraph_path_info and normalizes it to create all the path combos that only have one link between each node. :param path_info: List of of interface hops from a source node to a destination node. Each hop in the path is a list of all the interfaces from the current node to the next node. :return: List of lists. Each component list is a list with a unique Interface combination for the egress Interfaces from source to destination path_info example from source node 'B' to destination node 'D'. Example:: [ [[Interface(name = 'B-to-D', cost = 20, capacity = 125, node_object = Node('B'), remote_node_object = Node('D'), circuit_id = '3')]], # there is 1 interface from B to D and a complete path [[Interface(name = 'B-to-G_3', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = '28'), Interface(name = 'B-to-G', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = '8'), Interface(name = 'B-to-G_2', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = '18')], # there are 3 interfaces from B to G [Interface(name = 'G-to-D', cost = 10, capacity = 100, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = '9')]] # there is 1 int from G to D; end of path 2 ] Example:: [ [Interface(name = 'B-to-D', cost = 20, capacity = 125, node_object = Node('B'), remote_node_object = Node('D'), circuit_id = '3')], # this is a path with one hop [Interface(name = 'B-to-G_3', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = '28'), Interface(name = 'G-to-D', cost = 10, capacity = 100, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = '9')], # this is a path with 2 hops [Interface(name = 'B-to-G_2', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = '18'), Interface(name = 'G-to-D', cost = 10, capacity = 100, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = '9')], # this is a path with 2 hops [Interface(name = 'B-to-G', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = '8'), Interface(name = 'G-to-D', cost = 10, capacity = 100, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = '9')] # this is a path with 2 hops ] """ # List to hold unique path(s) path_list = [] for path in path_info: path = list(itertools.product(*path)) for path_option in path: path_list.append(list(path_option)) return path_list def _make_circuits_multidigraph( self, return_exception=True, include_failed_circuits=True ): """ Matches interface objects into circuits and returns the circuits list :param return_exception: Should an exception be returned if not all the interfaces can be matched into a circuit? :param include_failed_circuits: Should circuits that will be in a failed state be created? :return: a set of Circuit objects in the Model, each Circuit comprised of two Interface objects """ G = self._make_weighted_network_graph_mdg( include_failed_circuits=include_failed_circuits ) # Determine which interfaces pair up into good circuits in G graph_interfaces = ( (local_node_name, remote_node_name, data) for (local_node_name, remote_node_name, data) in G.edges(data=True) if G.has_edge(remote_node_name, local_node_name) ) # Set interface object in_ckt = False for interface in iter(self.interface_objects): interface.in_ckt = False circuits = set([]) # Using the paired interfaces (source_node, dest_node) pairs from G, # get the corresponding interface objects from the model to create # the Circuit object for interface in iter(graph_interfaces): # Get each interface from model for each try: int1 = self.get_interface_object_from_nodes( interface[0], interface[1], circuit_id=interface[2]["circuit_id"] )[0] except ( TypeError, IndexError, ): # TODO - are the exception catches necessary? msg = ( "No matching Interface Object found: source node {}, dest node {} " "circuit_id {} ".format( interface[0], interface[1], interface[2]["circuit_id"] ) ) raise ModelException(msg) try: int2 = self.get_interface_object_from_nodes( interface[1], interface[0], circuit_id=interface[2]["circuit_id"] )[0] except (TypeError, IndexError): msg = ( "No matching Interface Object found: source node {}, dest node {} " "circuit_id {} ".format( interface[1], interface[0], interface[2]["circuit_id"] ) ) raise ModelException(msg) # Mark the interfaces as in ckt if int1.in_ckt is False and int2.in_ckt is False: # Mark interface objects as in_ckt = True int1.in_ckt = True int2.in_ckt = True ckt = Circuit(int1, int2) circuits.add(ckt) # Find any interfaces that don't have counterpart exception_ints_not_in_ckt = [ (local_node_name, remote_node_name, data) for (local_node_name, remote_node_name, data) in G.edges(data=True) if not (G.has_edge(remote_node_name, local_node_name)) ] if exception_ints_not_in_ckt: exception_msg = ( "WARNING: These interfaces were not matched " "into a circuit {}".format(exception_ints_not_in_ckt) ) if return_exception: raise ModelException(exception_msg) else: return {"data": exception_ints_not_in_ckt} self.circuit_objects = circuits
[docs] def get_interface_object_from_nodes( self, local_node_name, remote_node_name, circuit_id=None ): """ Returns a list of Interface objects with the specified local and remote node names. If 'circuit_id' is not specified, may return a list of len > 1, as multiple/parallel interfaces are allowed in Parallel_Link_Model objects. If 'circuit_id' is specified, will return a list of len == 1, as specifying the 'circuit_id' will narrow down any list of multiple interfaces to a single interface because circuit_ids bond interfaces on different nodes into a Circuit object. :param local_node_name: Name of local node Interface resides on :param remote_node_name: Name of Interface's remote Node :param circuit_id: circuit_id of Interface (optional) :return: list of Interface objects with common local node and remote node """ interface_gen = iter(self.interface_objects) if circuit_id is None: interface_list = [ interface for interface in interface_gen if interface.node_object.name == local_node_name and interface.remote_node_object.name == remote_node_name ] else: interface_list = [ interface for interface in interface_gen if interface.node_object.name == local_node_name and interface.remote_node_object.name == remote_node_name and interface.circuit_id == circuit_id ] if len(interface_list) > 1: msg = ( "There is an internal error with circuit_iding; Interface circuit_ids must be unique" " per Node and the same circuit_id can only appear in a Parallel_Link_Model object " "twice and on separate Nodes" ) return ModelException(msg) return interface_list
[docs] def add_circuit( self, node_a_object, node_b_object, node_a_interface_name, node_b_interface_name, cost_intf_a=1, cost_intf_b=1, capacity=1000, failed=False, circuit_id=None, ): """ Creates component Interface objects for a new Circuit in the Model. The Circuit object will then be created during the validate_model() call. :param node_a_object: Node object :param node_b_object: Node object :param node_a_interface_name: name of component Interface on node_a :param node_b_interface_name: name of component Interface on node_b :param cost_intf_a: metric/cost of node_a_interface component Interface :param cost_intf_b: metric/cost of node_b_interface component Interface :param capacity: Circuit's capacity :param failed: Should the Circuit be created in a Failed state? :param circuit_id: Optional. Will be auto-assigned unless specified :return: Model with new Circuit comprised of 2 new Interfaces """ if circuit_id is None: raise ModelException("circuit_id must be specified explicitly") circuit_ids = self.all_interface_circuit_ids if circuit_id in circuit_ids: err_msg = "circuit_id value {} is already exists in model".format( circuit_id ) raise ModelException(err_msg) int_a = Interface( node_a_interface_name, cost_intf_a, capacity, node_a_object, node_b_object, circuit_id, ) int_b = Interface( node_b_interface_name, cost_intf_b, capacity, node_b_object, node_a_object, circuit_id, ) existing_int_keys = {interface._key for interface in self.interface_objects} if int_a._key in existing_int_keys: raise ModelException( "interface {} on node {} - " "interface already exists in model".format(int_a, node_a_object) ) elif int_b._key in existing_int_keys: raise ModelException( "interface {} on node {} - " "interface already exists in model".format(int_b, node_b_object) ) self.interface_objects.add(int_a) self.interface_objects.add(int_b) self.validate_model()
[docs] def get_all_paths_reservable_bw( self, source_node_name, dest_node_name, include_failed_circuits=True, cutoff=10, needed_bw=0, ): """ For a source and dest node name pair, find all simple path(s) with at least needed_bw reservable bandwidth available less than or equal to cutoff hops long. The amount of simple paths (paths that don't have repeating nodes) can be very large for larger topologies and so this call can be very expensive. Use the cutoff argument to limit the path length to consider to cut down on the time it takes to run this call. :param source_node_name: name of source node in path :param dest_node_name: name of destination node in path :param include_failed_circuits: include failed circuits in the topology :param needed_bw: the amount of reservable bandwidth required on the path :param cutoff: max amount of path hops :return: Return the path(s) in dictionary form: path = {'path': [list of shortest path routes]} Example:: >>> model.get_all_paths_reservable_bw('A', 'B', False, 5, 10) {'path': [ [Interface(name = 'A-to-D', cost = 40, capacity = 20.0, node_object = Node('A'), remote_node_object = Node('D'), circuit_id = 2), Interface(name = 'D-to-B', cost = 20, capacity = 125.0, node_object = Node('D'), remote_node_object = Node('B'), circuit_id = 7)], [Interface(name = 'A-to-D', cost = 40, capacity = 20.0, node_object = Node('A'), remote_node_object = Node('D'), circuit_id = 2), Interface(name = 'D-to-G', cost = 10, capacity = 100.0, node_object = Node('D'), remote_node_object = Node('G'), circuit_id = 8), Interface(name = 'G-to-B', cost = 10, capacity = 100.0, node_object = Node('G'), remote_node_object = Node('B'), circuit_id = 9)] ]} """ # Define a networkx DiGraph to find the path G = self._make_weighted_network_graph_mdg( include_failed_circuits=include_failed_circuits, needed_bw=needed_bw ) # Define the Model-style path to be built converted_path = {"path": []} # Find the simple paths in G between source and dest digraph_all_paths = nx.all_simple_paths( G, source_node_name, dest_node_name, cutoff=cutoff ) # Remove duplicate paths from digraph_all_paths # (duplicates can be caused by multiple links between nodes) digraph_unique_paths = [ list(path) for path in {tuple(path) for path in digraph_all_paths} ] try: for path in digraph_unique_paths: model_path = self._convert_nx_path_to_model_path(path, needed_bw) converted_path["path"].append(model_path) except BaseException: return converted_path # Normalize the path info to get all combinations of with parallel # interfaces path_info = self._normalize_multidigraph_paths(converted_path["path"]) return {"path": path_info}
[docs] def get_shortest_path(self, source_node_name, dest_node_name, needed_bw=0): """ For a source and dest node name pair, find the shortest path(s) with at least needed_bw available. :param source_node_name: name of source node in path :param dest_node_name: name of destination node in path :param needed_bw: the amount of reservable bandwidth required on the path :return: Return the shortest path in dictionary form: shortest_path = {'path': [list of shortest path routes], 'cost': path_cost} """ # Define a networkx DiGraph to find the path G = self._make_weighted_network_graph_mdg( include_failed_circuits=False, needed_bw=needed_bw ) # Define the Model-style path to be built converted_path = dict() converted_path["path"] = [] converted_path["cost"] = None # Find the shortest paths in G between source and dest multidigraph_shortest_paths = nx.all_shortest_paths( G, source_node_name, dest_node_name, weight="cost" ) # Get shortest path(s) from source to destination; this may include paths # that have multiple links between nodes try: for path in multidigraph_shortest_paths: model_path = self._convert_nx_path_to_model_path(path, needed_bw) converted_path["path"].append(model_path) converted_path["cost"] = nx.shortest_path_length( G, source_node_name, dest_node_name, weight="cost" ) except BaseException: return converted_path # Normalize the path info to get all combinations of with parallel # interfaces path_info = self._normalize_multidigraph_paths(converted_path["path"]) return {"cost": converted_path["cost"], "path": path_info}
[docs] def get_shortest_path_for_routed_lsp( self, source_node_name, dest_node_name, lsp, needed_bw ): """ For a source and dest node name pair, find the shortest path(s) with at least needed_bw available for an LSP that is already routed. Return the shortest path in dictionary form: shortest_path = {'path': [list of shortest path routes], 'cost': path_cost} :param source_node_name: name of source node :param dest_node_name: name of destination node :param lsp: LSP object :param needed_bw: reserved bandwidth for LSPs :return: dict {'path': [list of lists, each list a shortest path route], 'cost': path_cost} """ # Define a networkx DiGraph to find the path G = self._make_weighted_network_graph_routed_lsp(lsp, needed_bw=needed_bw) # Define the Model-style path to be built converted_path = {"path": [], "cost": None} # Find the shortest paths in G between source and dest digraph_shortest_paths = nx.all_shortest_paths( G, source_node_name, dest_node_name, weight="cost" ) try: for path in digraph_shortest_paths: model_path = self._convert_nx_path_to_model_path_routed_lsp( path, needed_bw, lsp ) converted_path["path"].append(model_path) converted_path["cost"] = nx.shortest_path_length( G, source_node_name, dest_node_name, weight="cost" ) except BaseException: return converted_path # Normalize the path info to get all combinations of with parallel # interfaces path_info = self._normalize_multidigraph_paths(converted_path["path"]) return {"cost": converted_path["cost"], "path": path_info}
def _convert_nx_path_to_model_path(self, nx_graph_path, needed_bw): """ Given a path from an networkx DiGraph, converts that path to a Model style path and returns that Model style path A networkx path is a list of nodes in order of transit. ex: ['A', 'B', 'G', 'D', 'F'] The corresponding model style path would be:: [Interface(name = 'A-to-B', cost = 20, capacity = 125, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = 9), Interface(name = 'B-to-G', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = 6), Interface(name = 'G-to-D', cost = 10, capacity = 100, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = 2), Interface(name = 'D-to-F', cost = 10, capacity = 300, node_object = Node('D'), remote_node_object = Node('F'), circuit_id = 1)] :param nx_graph_path: list of node names :param needed_bw: needed reservable bandwidth on the requested path :return: List of Model Interfaces from source to destination """ # Define a model-style path to build model_path = [] # look at each hop in the path for hop in nx_graph_path: current_hop_index = nx_graph_path.index(hop) next_hop_index = current_hop_index + 1 if next_hop_index < len(nx_graph_path): next_hop = nx_graph_path[next_hop_index] interface = [ interface for interface in self.get_interface_object_from_nodes(hop, next_hop) if interface.reservable_bandwidth >= needed_bw ] model_path.append(interface) return model_path def _convert_nx_path_to_model_path_routed_lsp(self, nx_graph_path, needed_bw, lsp): """ Given a path from an networkx DiGraph, converts that path to a Model style path and returns that Model style path A networkx path is a list of nodes in order of transit. ex: ['A', 'B', 'G', 'D', 'F'] Because a networkx path does not show the edges used, this def examines the interface(s) from each hop to the next hop and adds them to a hop_interface_list The corresponding model style path could be:: [Interface(name = 'A-to-B', cost = 20, capacity = 125, node_object = Node('A'), remote_node_object = Node('B'), circuit_id = 9), Interface(name = 'B-to-G', cost = 10, capacity = 100, node_object = Node('B'), remote_node_object = Node('G'), circuit_id = 6), Interface(name = 'G-to-D', cost = 10, capacity = 100, node_object = Node('G'), remote_node_object = Node('D'), circuit_id = 2), Interface(name = 'D-to-F', cost = 10, capacity = 300, node_object = Node('D'), remote_node_object = Node('F'), circuit_id = 1)] :param nx_graph_path: list of node names :param needed_bw: needed reservable bandwidth on the requested path :param lsp: RSVP LSP object to be acted on :return: List of Model Interfaces from source to destination """ # Define a model-style path to build model_path = [] # look at each hop in the path for hop in nx_graph_path: current_hop_index = nx_graph_path.index(hop) next_hop_index = current_hop_index + 1 if next_hop_index < len(nx_graph_path): next_hop = nx_graph_path[next_hop_index] for interface in self.get_interface_object_from_nodes(hop, next_hop): # Look at all the interface(s) from (current) hop to next_hop; see if # any of those interfaces are in the current path for lsp; if they are, # see if any of them could handle the additional_needed_bandwidth for lsp hop_interface_list = [] if interface in lsp.path["interfaces"] and ( interface.reservable_bandwidth + lsp.reserved_bandwidth >= needed_bw ): hop_interface_list.append(interface) elif interface.reservable_bandwidth >= needed_bw: # If the interface is not in the current path but can # accommodate the needed_bw, then add that interface # to model_path hop_interface_list.append(interface) if hop_interface_list: model_path.append(hop_interface_list) return model_path def _determine_lsp_state_info(self, lsps, traff_on_each_group_lsp): """ Determine LSP's specific path and reserved bandwidth; also consume reserved bandwidth on transited Interfaces :param lsps: List of parallel LSPs (LSPs with common source/dest nodes) :param traff_on_each_group_lsp: How much traffic each LSP should attempt to carry :return: None; determines path and reserved bandwidth for each LSP in lsps and also consumes reservable bandwidth on each Interface each LSP transits """ for lsp in lsps: # Check to see if configured_setup_bandwidth is set; if so, # set reserved_bandwidth and setup_bandwidth equal to # configured_setup_bandwidth value if lsp.configured_setup_bandwidth is None: lsp.reserved_bandwidth = traff_on_each_group_lsp lsp.setup_bandwidth = traff_on_each_group_lsp else: lsp.reserved_bandwidth = lsp.configured_setup_bandwidth lsp.setup_bandwidth = lsp.configured_setup_bandwidth G = self._make_weighted_network_graph_mdg( include_failed_circuits=False, rsvp_required=True, needed_bw=lsp.setup_bandwidth, ) lsp.path = {} # Get shortest paths in networkx multidigraph try: nx_sp = list( nx.all_shortest_paths( G, lsp.source_node_object.name, lsp.dest_node_object.name, weight="cost", ) ) except nx.exception.NetworkXNoPath: # There is no path; path = 'Unrouted' lsp.path = "Unrouted" lsp.reserved_bandwidth = "Unrouted" continue # Convert node hop by hop paths from G into Interface-based paths all_paths = self._get_all_paths_mdg(G, nx_sp) # all_paths may have hops between nodes that can take different Interfaces; # normalize those hops that could transit any of multiple Interfaces into # distinct, unique possible paths candidate_path_info = self._normalize_multidigraph_paths(all_paths) # Candidate paths with enough reservable bandwidth candidate_path_info_w_reservable_bw = [] # Determine which candidate paths have enough reservable bandwidth for path in candidate_path_info: if ( min(interface.reservable_bandwidth for interface in path) >= lsp.setup_bandwidth ): candidate_path_info_w_reservable_bw.append(path) # If multiple lowest_metric_paths, find those with fewest hops if not candidate_path_info_w_reservable_bw: lsp.path = "Unrouted" lsp.reserved_bandwidth = "Unrouted" continue elif len(candidate_path_info_w_reservable_bw) > 1: fewest_hops = min( len(path) for path in candidate_path_info_w_reservable_bw ) lowest_hop_count_paths = [ path for path in candidate_path_info_w_reservable_bw if len(path) == fewest_hops ] if len(lowest_hop_count_paths) > 1: new_path = random.choice(lowest_hop_count_paths) else: new_path = lowest_hop_count_paths[0] else: new_path = candidate_path_info_w_reservable_bw[0] # Change LSP path into more verbose form and set LSP's path self._add_lsp_path_data(lsp, new_path) for interface in [ interface for interface in lsp.path["interfaces"] if lsp.path != "Unrouted" ]: interface.reserved_bandwidth += lsp.reserved_bandwidth def _make_weighted_network_graph_routed_lsp(self, lsp, needed_bw=0): """ Returns a networkx weighted network directional graph from the input Model object. Considers edges with needed_bw of reservable_bandwidth and also takes into account reserved_bandwidth by the lsp on Interfaces in the existing LSP path. :param include_failed_circuits: failed circuits can be included in the graph as functional edges :param lsp: LSP to be considered :param needed_bw: amount of reservable bandwidth an interface must have to be added to the graph :return: """ # The Interfaces that the lsp is routed over currently lsp_path_interfaces = lsp.path["interfaces"] eligible_interface_generator = ( interface for interface in self.interface_objects if (interface.failed is False and interface.rsvp_enabled is True) ) eligible_interfaces = set() # Find only the interfaces that are not failed and that have # enough reservable_bandwidth for interface in eligible_interface_generator: # Add back the lsp's reserved bandwidth to Interfaces already in its path if interface in lsp_path_interfaces: effective_reservable_bw = ( interface.reservable_bandwidth + lsp.reserved_bandwidth ) else: effective_reservable_bw = interface.reservable_bandwidth if effective_reservable_bw >= needed_bw: eligible_interfaces.add(interface) # Get edge names in eligible_interfaces edge_names = ( ( interface.node_object.name, interface.remote_node_object.name, interface.cost, ) for interface in eligible_interfaces ) # Make a new graph with the eligible interfaces (interfaces # with enough effective_reservable_bw) G = nx.MultiDiGraph() # Add edges to networkx MultiDiGraph G.add_weighted_edges_from(edge_names, weight="cost") # Add all the nodes node_name_iterator = (node.name for node in self.node_objects) G.add_nodes_from(node_name_iterator) return G
[docs] @classmethod def load_model_file(cls, data_file): # TODO - allow commas instead of tabs """ Opens a network_modeling data file, returns a model containing the info in the data file, and runs update_simulation(). The data file must be of the appropriate format to produce a valid model. This cannot be used to open multiple models in a single python instance - there may be unpredictable results in the info in the models. The format for the file must be a tab separated value file. CIRCUIT ID (circuit_id) MUST BE SPECIFIED AS THIS IS WHAT ALLOWS THE CLASS TO DISCERN WHAT MULTIPLE, PARALLEL INTERFACES BETWEEN THE SAME NODES MATCH UP INTO WHICH CIRCUIT. THE circuit_id CAN BE ANY COMMON KEY, SUCH AS IP SUBNET ID OR DESIGNATED CIRCUIT ID FROM PRODUCTION. This docstring you are reading may not display the table info explanations/examples below correctly on https://pyntm.readthedocs.io/en/latest/api.html. Recommend either using help(Model.load_model_file) at the python3 cli or looking at one of the sample model data_files in github: https://github.com/tim-fiola/network_traffic_modeler_py3/blob/master/examples/sample_network_model_file.csv https://github.com/tim-fiola/network_traffic_modeler_py3/blob/master/examples/lsp_model_test_file.csv The following headers must exist, with the following tab-column names beneath:: INTERFACES_TABLE - node_object_name - name of node where interface resides - remote_node_object_name - name of remote node - name - interface name - cost - IGP cost/metric for interface - capacity - capacity - circuit_id - id of the circuit; used to match two Interfaces into Circuits; - each circuit_id can only appear twice in the model - circuit_id can be string or integer - rsvp_enabled (optional) - is interface allowed to carry RSVP LSPs? True|False; default is True - percent_reservable_bandwidth (optional) - percent of capacity allowed to be reserved by RSVP LSPs; this value should be given as a percentage value - ie 80% would be given as 80, NOT .80. Default is 100 Note - The existence of Nodes will be inferred from the INTERFACES_TABLE. So a Node created from an Interface does not have to appear in the NODES_TABLE unless you want to add additional attributes for the Node such as latitude/longitude NODES_TABLE - - name - name of node - lon - longitude (or y-coordinate) - lat - latitude (or x-coordinate) - igp_shortcuts_enabled(default=False) Note - The NODES_TABLE is present for 2 reasons: - to add a Node that has no interfaces - and/or to add additional attributes for a Node inferred from the INTERFACES_TABLE DEMANDS_TABLE - source - source node name - dest - destination node name - traffic - amount of traffic on demand - name - name of demand RSVP_LSP_TABLE (this table is optional) - source - source node name - dest - destination node name - name - name of LSP - configured_setup_bw - if LSP has a fixed, static configured setup bandwidth, place that static value here, if LSP is auto-bandwidth, then leave this blank for the LSP (optional) - manual_metric - manually assigned metric for LSP, if not using default metric from topology shortest path (optional) Functional model files can be found in this directory in https://github.com/tim-fiola/network_traffic_modeler_py3/tree/master/examples Here is an example of a data file. Example:: INTERFACES_TABLE node_object_name remote_node_object_name name cost capacity circuit_id rsvp_enabled percent_reservable_bandwidth # noqa E501 A B A-to-B_1 20 120 1 True 50 B A B-to-A_1 20 120 1 True 50 A B A-to-B_2 20 150 2 B A B-to-A_2 20 150 2 A B A-to-B_3 10 200 3 False B A B-to-A_3 10 200 3 False NODES_TABLE name lon lat igp_shortcuts_enabled(default=False) A 50 0 True B 0 -50 False DEMANDS_TABLE source dest traffic name A B 80 dmd_a_b_1 RSVP_LSP_TABLE source dest name configured_setup_bw manual_metric A B lsp_a_b_1 10 19 A B lsp_a_b_2 6 :param data_file: file with model info :return: Model object """ # TODO - allow user to add user-defined columns in NODES_TABLE and add that as an attribute to the Node interface_set = set() node_set = set() demand_set = set() lsp_set = set() # Open the file with the data, read it, and split it into lines with open(data_file, "r", encoding="utf-8-sig") as f: data = f.read() lines = data.splitlines() # Define the Interfaces from the data and extract the presence of # Nodes from the Interface data int_info_begin_index = lines.index("INTERFACES_TABLE") + 2 int_info_end_index = find_end_index(int_info_begin_index, lines) # Check that each circuit_id appears exactly 2 times circuit_id_list = [] for line in lines[int_info_begin_index:int_info_end_index]: try: circuit_id_item = line.split("\t")[5] circuit_id_list.append(circuit_id_item) except IndexError: pass bad_circuit_ids = [ {"circuit_id": item, "appearances": circuit_id_list.count(item)} for item in set(circuit_id_list) if circuit_id_list.count(item) != 2 ] if len(bad_circuit_ids) != 0: msg = ( "Each circuit_id value must appear exactly twice; the following circuit_id values " "do not meet that criteria: {}".format(bad_circuit_ids) ) raise ModelException(msg) interface_set, node_set = cls._extract_interface_data_and_implied_nodes( int_info_begin_index, int_info_end_index, lines ) # Define the explicit nodes info from the file nodes_info_begin_index = lines.index("NODES_TABLE") + 2 nodes_info_end_index = find_end_index(nodes_info_begin_index, lines) node_lines = lines[nodes_info_begin_index:nodes_info_end_index] for node_line in node_lines: node_set = cls._add_node_from_data( demand_set, interface_set, lsp_set, node_line, node_set ) # Define the demands info demands_info_begin_index = lines.index("DEMANDS_TABLE") + 2 demands_info_end_index = find_end_index(demands_info_begin_index, lines) # There may or may not be LSPs in the model, so if there are not, # set the demands_info_end_index as the last line in the file if not demands_info_end_index: demands_info_end_index = len(lines) demands_lines = lines[demands_info_begin_index:demands_info_end_index] for demand_line in demands_lines: try: cls._add_demand_from_data(demand_line, demand_set, lines, node_set) except ModelException as e: err_msg = e.args[0] raise ModelException(err_msg) # Define the LSP info (if present) try: lsp_info_begin_index = lines.index("RSVP_LSP_TABLE") + 2 cls._add_lsp_from_data(lsp_info_begin_index, lines, lsp_set, node_set) except ValueError: print("RSVP_LSP_TABLE not in file; no LSPs added to model") except ModelException as e: err_msg = e.args[0] raise ModelException(err_msg) return cls(interface_set, node_set, demand_set, lsp_set)
@classmethod def _extract_interface_data_and_implied_nodes( cls, int_info_begin_index, int_info_end_index, lines ): """ Extracts interface data from lines and adds Interface objects to a set. Also extracts the implied Nodes from the Interfaces and adds those Nodes to a set. :param int_info_begin_index: Index position in lines where interface info begins :param int_info_end_index: Index position in lines where interface info ends :param lines: lines of data describing a Model objects :return: set of Interface objects, set of Node objects created from lines """ interface_set = set() node_set = set() interface_lines = lines[int_info_begin_index:int_info_end_index] # Add the Interfaces to a set for interface_line in interface_lines: # Read interface characteristics if len(interface_line.split("\t")) == 6: [ node_name, remote_node_name, name, cost, capacity, circuit_id, ] = interface_line.split("\t") rsvp_enabled_bool = True percent_reservable_bandwidth = 100 elif len(interface_line.split("\t")) == 7: [ node_name, remote_node_name, name, cost, capacity, circuit_id, rsvp_enabled, ] = interface_line.split("\t") if rsvp_enabled in [True, "T", "True", "true"]: rsvp_enabled_bool = True else: rsvp_enabled_bool = False percent_reservable_bandwidth = 100 elif len(interface_line.split("\t")) >= 8: [ node_name, remote_node_name, name, cost, capacity, circuit_id, rsvp_enabled, percent_reservable_bandwidth, ] = interface_line.split("\t") if rsvp_enabled in [True, "T", "True", "true"]: rsvp_enabled_bool = True else: rsvp_enabled_bool = False else: msg = ( "node_name, remote_node_name, name, cost, capacity, circuit_id " "must be defined for line {}, line index {}".format( interface_line, lines.index(interface_line) ) ) raise ModelException(msg) node_names = [node.name for node in node_set] if node_name in node_names: node_object = [node for node in node_set if node.name == node_name][0] else: node_object = Node(node_name) if remote_node_name in node_names: remote_node_object = [ node for node in node_set if node.name == remote_node_name ][0] else: remote_node_object = Node(remote_node_name) new_interface = Interface( name, int(cost), int(capacity), node_object, remote_node_object, circuit_id, rsvp_enabled_bool, float(percent_reservable_bandwidth), ) if new_interface._key not in { interface._key for interface in interface_set }: interface_set.add(new_interface) else: print( "{} already exists in model; disregarding line {}".format( new_interface, lines.index(interface_line) ) ) # Derive Nodes from the Interface data if node_name not in {node.name for node in node_set}: node_set.add(new_interface.node_object) if remote_node_name not in {node.name for node in node_set}: node_set.add(new_interface.remote_node_object) return interface_set, node_set
class Parallel_Link_Model(FlexModel): """ This is the legacy Parallel_Link_Model class, now a subclass of the more aptly named FlexModel class. This has been added to attempt to keep any legacy code, written in pyNTM 1.6 or earlier, from breaking. """ def __init__( self, interface_objects=set(), node_objects=set(), demand_objects=set(), rsvp_lsp_objects=set(), ): self.interface_objects = interface_objects self.node_objects = node_objects self.demand_objects = demand_objects self.circuit_objects = set() self.rsvp_lsp_objects = rsvp_lsp_objects self.srlg_objects = set() self._parallel_lsp_groups = {} super().__init__( interface_objects, node_objects, demand_objects, rsvp_lsp_objects )