Files
oopt-gnpy/gnpy/core/request.py
EstherLerouzic 1ba748f2a4 Changing optical power and nb channels as optional input for requests
- normal way is usually to apply the design optical power for all channels.
  This change uses the default power (same power as used for design) but enables to
  force an arbitrary power if needed. TODO : introduce spectral power density
  to apply power depending on baudrate.
- definition of min max frequency and spacing define the nb of channels:
  uses min max frequencies and spacing to determine nb-channels. It is possible
  to force a different spacing for the request. TODO: check that the value is
  consistant with baudrate and min max values.

Signed-off-by: EstherLerouzic <esther.lerouzic@orange.com>
2018-10-25 15:36:53 +01:00

653 lines
28 KiB
Python

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
gnpy.core.request
=================
This module contains path request functionality.
This functionality allows the user to provide a JSON request
file in accordance with a Yang model for requesting path
computations and returns path results in terms of path
and feasibility
See: draft-ietf-teas-yang-path-computation-01.txt
"""
from collections import namedtuple
from logging import getLogger, basicConfig, CRITICAL, DEBUG, INFO
from networkx import (dijkstra_path, NetworkXNoPath, all_simple_paths)
from networkx.utils import pairwise
from numpy import mean
from gnpy.core.service_sheet import convert_service_sheet, Request_element, Element
from gnpy.core.elements import Transceiver, Roadm, Edfa, Fused
from gnpy.core.network import set_roadm_loss
from gnpy.core.utils import db2lin, lin2db
from gnpy.core.info import create_input_spectral_information, SpectralInformation, Channel, Power
from copy import copy, deepcopy
from csv import writer
logger = getLogger(__name__)
RequestParams = namedtuple('RequestParams','request_id source destination trx_type'+
' trx_mode nodes_list loose_list spacing power nb_channel frequency format baud_rate OSNR bit_rate roll_off')
DisjunctionParams = namedtuple('DisjunctionParams','disjunction_id relaxable link_diverse node_diverse disjunctions_req')
class Path_request:
def __init__(self, *args, **params):
params = RequestParams(**params)
self.request_id = params.request_id
self.source = params.source
self.destination = params.destination
self.tsp = params.trx_type
self.tsp_mode = params.trx_mode
self.baud_rate = params.baud_rate
self.nodes_list = params.nodes_list
self.loose_list = params.loose_list
self.spacing = params.spacing
self.power = params.power
self.nb_channel = params.nb_channel
self.frequency = params.frequency
self.format = params.format
self.OSNR = params.OSNR
self.bit_rate = params.bit_rate
self.roll_off = params.roll_off
def __str__(self):
return '\n\t'.join([ f'{type(self).__name__} {self.request_id}',
f'source: {self.source}',
f'destination: {self.destination}'])
def __repr__(self):
return '\n\t'.join([ f'{type(self).__name__} {self.request_id}',
f'source: \t{self.source}',
f'destination:\t{self.destination}',
f'trx type:\t{self.tsp}',
f'trx mode:\t{self.tsp_mode}',
f'baud_rate:\t{self.baud_rate * 1e-9} Gbaud',
f'bit_rate:\t{self.bit_rate * 1e-9} Gb/s',
f'spacing:\t{self.spacing * 1e-9} GHz',
f'power: \t{round(lin2db(self.power)+30,2)} dBm',
f'nb channels: \t{self.nb_channel}'
'\n'])
class Disjunction:
def __init__(self, *args, **params):
params = DisjunctionParams(**params)
self.disjunction_id = params.disjunction_id
self.relaxable = params.relaxable
self.link_diverse = params.link_diverse
self.node_diverse = params.node_diverse
self.disjunctions_req = params.disjunctions_req
def __str__(self):
return '\n\t'.join([f'relaxable: {self.relaxable}',
f'link-diverse: {self.link_diverse}',
f'node-diverse: {self.node_diverse}',
f'request-id-numbers: {self.disjunctions_req}']
)
def __repr__(self):
return '\n\t'.join([f'relaxable: {self.relaxable}',
f'link-diverse: {self.link_diverse}',
f'node-diverse: {self.node_diverse}',
f'request-id-numbers: {self.disjunctions_req}']
)
class Result_element(Element):
def __init__(self,path_request,computed_path):
self.path_id = path_request.request_id
self.path_request = path_request
self.computed_path = computed_path
hop_type = []
for e in computed_path :
if isinstance(e, Transceiver) :
hop_type.append(' - '.join([path_request.tsp,path_request.tsp_mode]))
else:
hop_type.append('not recorded')
self.hop_type = hop_type
uid = property(lambda self: repr(self))
@property
def pathresult(self):
if not self.computed_path:
return {
'path-id': self.path_id,
'path-properties':{
'path-metric': [
{
'metric-type': 'SNR@bandwidth',
'accumulative-value': 'None'
},
{
'metric-type': 'SNR@0.1nm',
'accumulative-value': 'None'
},
{
'metric-type': 'OSNR@bandwidth',
'accumulative-value': 'None'
},
{
'metric-type': 'OSNR@0.1nm',
'accumulative-value': 'None'
},
{
'metric-type': 'reference_power',
'accumulative-value': self.path_request.power
}
],
'path-srlgs': {
'usage': 'not used yet',
'values': 'not used yet'
},
'path-route-objects': [
{
'path-route-object': {
'index': 0,
'unnumbered-hop': {
'node-id': self.path_request.source,
'link-tp-id': self.path_request.source,
'hop-type': ' - '.join([self.path_request.tsp, self.path_request.tsp_mode]),
'direction': 'not used'
},
'label-hop': {
'te-label': {
'generic': 'not used yet',
'direction': 'not used yet'
}
}
}
},
{
'path-route-object': {
'index': 1,
'unnumbered-hop': {
'node-id': self.path_request.destination,
'link-tp-id': self.path_request.destination,
'hop-type': ' - '.join([self.path_request.tsp, self.path_request.tsp_mode]),
'direction': 'not used'
},
'label-hop': {
'te-label': {
'generic': 'not used yet',
'direction': 'not used yet'
}
}
}
}
]
}
}
else:
return {
'path-id': self.path_id,
'path-properties':{
'path-metric': [
{
'metric-type': 'SNR@bandwidth',
'accumulative-value': round(mean(self.computed_path[-1].snr),2)
},
{
'metric-type': 'SNR@0.1nm',
'accumulative-value': round(mean(self.computed_path[-1].snr+lin2db(self.path_request.baud_rate/12.5e9)),2)
},
{
'metric-type': 'OSNR@bandwidth',
'accumulative-value': round(mean(self.computed_path[-1].osnr_ase),2)
},
{
'metric-type': 'OSNR@0.1nm',
'accumulative-value': round(mean(self.computed_path[-1].osnr_ase_01nm),2)
},
{
'metric-type': 'reference_power',
'accumulative-value': self.path_request.power
}
],
'path-srlgs': {
'usage': 'not used yet',
'values': 'not used yet'
},
'path-route-objects': [
{
'path-route-object': {
'index': self.computed_path.index(n),
'unnumbered-hop': {
'node-id': n.uid,
'link-tp-id': n.uid,
'hop-type': self.hop_type[self.computed_path.index(n)],
'direction': 'not used'
},
'label-hop': {
'te-label': {
'generic': 'not used yet',
'direction': 'not used yet'
}
}
}
} for n in self.computed_path
]
}
}
@property
def json(self):
return self.pathresult
def compute_constrained_path(network, req):
trx = [n for n in network.nodes() if isinstance(n, Transceiver)]
roadm = [n for n in network.nodes() if isinstance(n, Roadm)]
edfa = [n for n in network.nodes() if isinstance(n, Edfa)]
source = next(el for el in trx if el.uid == req.source)
# start the path with its source
# TODO : avoid loops due to constraints , guess name based on string,
# avoid crashing if on req is not correct
total_path = [source]
for n in req.nodes_list:
try :
node = next(el for el in trx if el.uid == n)
except StopIteration:
try:
node = next(el for el in roadm if el.uid == n)
except StopIteration:
try:
# TODO this test is not giving good results: full name of the
# amp is required to avoid ambiguity on the direction
node = next(el for el in edfa
if el.uid.find(f'{n}'))
except StopIteration:
msg = f'could not find node : {n} in network topology: \
not a trx, roadm, edfa or fused element'
logger.critical(msg)
raise ValueError(msg)
# extend path list without repeating source -> skip first element in the list
try:
total_path.extend(dijkstra_path(network, source, node)[1:])
source = node
except NetworkXNoPath:
if req.loose_list[req.nodes_list.index(n)] == 'loose':
print(f'could not find a path from {source.uid} to loose node : {n} in network topology')
print(f'node {n} is skipped')
else:
msg = f'could not find a path from {source.uid} to node : {n} in network topology'
logger.critical(msg)
print(msg)
total_path = []
return total_path
def propagate(path, req, equipment, show=False):
#update roadm loss in case of power sweep (power mode only)
set_roadm_loss(path, equipment, lin2db(req.power*1e3))
si = create_input_spectral_information(
req.frequency['min'], req.roll_off,
req.baud_rate, req.power, req.spacing, req.nb_channel)
for el in path:
si = el(si)
if show :
print(el)
return path
def jsontocsv(json_data,equipment,fileout):
# read json path result file in accordance with:
# Yang model for requesting Path Computation
# draft-ietf-teas-yang-path-computation-01.txt.
# and write results in an CSV file
mywriter = writer(fileout)
mywriter.writerow(('path-id','source','destination','transponder-type',\
'transponder-mode','baud rate (Gbaud)', 'input power (dBm)','path',\
'OSNR@bandwidth','OSNR@0.1nm','SNR@bandwidth','SNR@0.1nm','Pass?'))
tspjsondata = equipment['Transceiver']
#print(tspjsondata)
for p in json_data['path']:
path_id = p['path-id']
source = p['path-properties']['path-route-objects'][0]\
['path-route-object']['unnumbered-hop']['node-id']
destination = p['path-properties']['path-route-objects'][-1]\
['path-route-object']['unnumbered-hop']['node-id']
pth = ' | '.join([ e['path-route-object']['unnumbered-hop']['node-id']
for e in p['path-properties']['path-route-objects']])
[tsp,mode] = p['path-properties']['path-route-objects'][0]\
['path-route-object']['unnumbered-hop']['hop-type'].split(' - ')
# find the min acceptable OSNR, baud rate from the eqpt library based on tsp (tupe) and mode (format)
# loading equipment already tests the existence of tsp type and mode:
[minosnr, baud_rate] = next([m['OSNR'] , m['baud_rate']]
for m in equipment['Transceiver'][tsp].mode if m['format']==mode)
output_snr = next(e['accumulative-value']
for e in p['path-properties']['path-metric'] if e['metric-type'] == 'SNR@0.1nm')
output_snrbandwidth = next(e['accumulative-value']
for e in p['path-properties']['path-metric'] if e['metric-type'] == 'SNR@bandwidth')
output_osnr = next(e['accumulative-value']
for e in p['path-properties']['path-metric'] if e['metric-type'] == 'OSNR@0.1nm')
output_osnrbandwidth = next(e['accumulative-value']
for e in p['path-properties']['path-metric'] if e['metric-type'] == 'OSNR@bandwidth')
power = next(e['accumulative-value']
for e in p['path-properties']['path-metric'] if e['metric-type'] == 'reference_power')
if isinstance(output_snr, str):
isok = ''
else:
isok = output_snr >= minosnr
mywriter.writerow((path_id,
source,
destination,
tsp,
mode,
baud_rate*1e-9,
round(lin2db(power)+30,2),
pth,
output_osnrbandwidth,
output_osnr,
output_snrbandwidth,
output_snr,
isok
))
def compute_path_dsjctn(network, equipment, pathreqlist, disjunctions_list):
# pathreqlist is a list of Path_request objects
# disjunctions_list a list of Disjunction objects
# given a network, a list of requests with the set of disjunction features between
# request, the function computes the set of path satisfying : first the disjunction
# constraint and second the routing constraint if the request include an explicit
# set of elements to pass through.
# the algorithm used allows to specify disjunction for demands not sharing source or
# destination.
# a request might be declared as disjoint from several requests
# it is a iterative process:
# first computes a list of all shortest path (this may add computation time)
# second elaborate the set of path solution for each synchronization vector
# third select only the candidates that satisfy all synchronization vectors they belong to
# fourth apply route constraints : remove candidate path that do not satisfy the constraint
# fifth select the first candidate among the set of candidates.
# the example network used in comments has been added to the set of data tests files
# define the list to be returned
path_res_list = []
# all disjctn must be computed at once together to avoid blocking
# 1 1
# eg a----b-----c
# |1 |0.5 |1
# e----f--h--g
# 1 0.5 0.5
# if I have to compute a to g and a to h
# I must not compute a-b-f-h-g, otherwise there is no disjoint path remaining for a to h
# instead I should list all most disjoint path and select the one that have the less
# number of commonalities
# \ path abfh aefh abcgh
# \___cost 2 2.5 3.5
# path| cost
# abfhg| 2.5 x x x
# abcg | 3 x x
# aefhg| 3 x x x
# from this table abcg and aefh have no common links and should be preferred
# even they are not the shortest paths
# build the list of pathreqlist elements not concerned by disjunction
global_disjunctions_list = [e for d in disjunctions_list for e in d.disjunctions_req ]
pathreqlist_simple = [e for e in pathreqlist if e.request_id not in global_disjunctions_list]
pathreqlist_disjt = [e for e in pathreqlist if e.request_id in global_disjunctions_list]
# use a mirror class to record path and the corresponding requests
class Pth:
def __init__(self, req, pth, simplepth):
self.req = req
self.pth = pth
self.simplepth = simplepth
# step 1
# for each remaining request compute a set of simple path
allpaths = {}
rqs = {}
simple_rqs = {}
simple_rqs_reversed = {}
for pathreq in pathreqlist_disjt :
all_simp_pths = list(all_simple_paths(network,\
source=next(el for el in network.nodes() if el.uid == pathreq.source),\
target=next(el for el in network.nodes() if el.uid == pathreq.destination)))
# sort them
all_simp_pths = sorted(all_simp_pths, key=lambda path: len(path))
# reversed direction paths required to check disjunction on both direction
all_simp_pths_reversed = []
for pth in all_simp_pths:
all_simp_pths_reversed.append(find_reversed_path(pth,network))
rqs[pathreq.request_id] = all_simp_pths
temp =[]
for p in all_simp_pths :
# build a short list representing each roadm+direction with the first item
# start enumeration at 1 to avoid Trx in the list
s = [e.uid for i,e in enumerate(p[1:-1]) \
if (isinstance(e,Roadm) | (isinstance(p[i],Roadm) ))]
temp.append(s)
# id(s) is unique even if path is the same: two objects with same
# path have two different ids
allpaths[id(s)] = Pth(pathreq,p,s)
simple_rqs[pathreq.request_id] = temp
temp =[]
for p in all_simp_pths_reversed :
# build a short list representing each roadm+direction with the first item
# start enumeration at 1 to avoid Trx in the list
temp.append([e.uid for i,e in enumerate(p[1:-1]) \
if (isinstance(e,Roadm) | (isinstance(p[i],Roadm) ))] )
simple_rqs_reversed[pathreq.request_id] = temp
# step 2
# for each set of requests that need to be disjoint
# select the disjoint path combination
candidates = {}
for d in disjunctions_list :
dlist = d.disjunctions_req.copy()
# each line of dpath is one combination of path that satisfies disjunction
dpath = []
for i,p in enumerate(simple_rqs[dlist[0]]):
dpath.append([p])
# allpaths[id(p)].d_id = d.disjunction_id
# in each loop, dpath is updated with a path for rq that satisfies
# disjunction with each path in dpath
# for example, assume set of requests in the vector (disjunction_list) is {rq1,rq2, rq3}
# rq1 p1: abfhg
# p2: aefhg
# p3: abcg
# rq2 p8: bf
# rq3 p4: abcgh
# p6: aefh
# p7: abfh
# initiate with rq1
# dpath = [[p1]
# [p2]
# [p3]]
# after first loop:
# dpath = [[p1 p8]
# [p3 p8]]
# since p2 and p8 are not disjoint
# after second loop:
# dpath = [ p3 p8 p6 ]
# since p1 and p4 are not disjoint
# p1 and p7 are not disjoint
# p3 and p4 are not disjoint
# p3 and p7 are not disjoint
for e1 in dlist[1:] :
temp = []
for j,p1 in enumerate(simple_rqs[e1]):
# allpaths[id(p1)].d_id = d.disjunction_id
# can use index j in simple_rqs_reversed because index
# of direct and reversed paths have been kept identical
p1_reversed = simple_rqs_reversed[e1][j]
# print(p1_reversed)
# print('\n\n')
for k,c in enumerate(dpath) :
# print(f' c: \t{c}')
temp2 = c.copy()
all_disjoint = 0
for p in c :
all_disjoint += isdisjoint(p1,p)+ isdisjoint(p1_reversed,p)
if all_disjoint ==0:
temp2.append(p1)
temp.append(temp2)
# print(f' coucou {e1}: \t{temp}')
dpath = temp
# print(dpath)
candidates[d.disjunction_id] = dpath
# for i in disjunctions_list :
# print(f'\n{candidates[i.disjunction_id]}')
# step 3
# now for each request, select the path that satisfies all disjunctions
# path must be in candidates[id] for all concerned ids
# for example, assume set of sync vectors (disjunction groups) is
# s1 = {rq1 rq2} s2 = {rq1 rq3}
# candidate[s1] = [[p1 p8]
# [p3 p8]]
# candidate[s2] = [[p3 p6]]
# for rq1 p3 should be preferred
for pathreq in pathreqlist_disjt:
concerned_d_id = [d.disjunction_id for d in disjunctions_list if pathreq.request_id in d.disjunctions_req]
# for each set of solution, verify that the same path is used for the same request
candidate_paths = simple_rqs[pathreq.request_id]
# print('coucou')
# print(pathreq.request_id)
for p in candidate_paths :
iscandidate = 0
for sol in concerned_d_id :
test = 1
# for each solution test if p is part of the solution
# if yes, then p can remain a candidate
for i,m in enumerate(candidates[sol]) :
if p in m:
if allpaths[id(m[m.index(p)])].req.request_id == pathreq.request_id :
test = 0
break
iscandidate += test
if iscandidate != 0:
for l in concerned_d_id :
for m in candidates[l] :
if p in m :
candidates[l].remove(m)
# for i in disjunctions_list :
# print(i.disjunction_id)
# print(f'\n{candidates[i.disjunction_id]}')
# step 4 apply route constraints : remove candidate path that do not satisfy the constraint
# only in the case of disjounction: the simple path is processed in request.compute_constrained_path
# TODO : keep a version without the loose constraint
for d in disjunctions_list :
temp = []
for j,sol in enumerate(candidates[d.disjunction_id]) :
testispartok = True
for i,p in enumerate(sol) :
# print(f'test {allpaths[id(p)].req.request_id}')
# print(f'length of route {len(allpaths[id(p)].req.nodes_list)}')
if allpaths[id(p)].req.nodes_list :
# if p does not containt the ordered list node, remove sol from the candidate
# except if this was the last solution: then check if the constraint is loose or not
if not ispart(allpaths[id(p)].req.nodes_list, p) :
# print(f'nb of solutions {len(temp)}')
if j < len(candidates[d.disjunction_id])-1 :
msg = f'removing {sol}'
logger.info(msg)
testispartok = False
#break
else:
if 'loose' in allpaths[id(p)].req.loose_list:
logger.info(f'Could not apply route constraint'+
f'{allpaths[id(p)].req.nodes_list} on request {allpaths[id(p)].req.request_id}')
else :
logger.info(f'removing last solution from candidate paths\n{sol}')
testispartok = False
if testispartok :
temp.append(sol)
candidates[d.disjunction_id] = temp
# step 5 select the first combination that works
pathreslist_disjoint = {}
for d in disjunctions_list :
test_sol = True
while test_sol:
# print('coucou')
if candidates[d.disjunction_id] :
for p in candidates[d.disjunction_id][0]:
if allpaths[id(p)].req in pathreqlist_disjt:
# print(f'selected path :{p} for req {allpaths[id(p)].req.request_id}')
pathreslist_disjoint[allpaths[id(p)].req] = allpaths[id(p)].pth
pathreqlist_disjt.remove(allpaths[id(p)].req)
candidates = remove_candidate(candidates, allpaths, allpaths[id(p)].req, p)
test_sol = False
else:
msg = f'No disjoint path found with added constraint'
logger.critical(msg)
print(f'{msg}\nComputation stopped.')
# TODO in this case: replay step 5 with the candidate without constraints
exit()
# for i in disjunctions_list :
# print(i.disjunction_id)
# print(f'\n{candidates[i.disjunction_id]}')
# list the results in the same order as initial pathreqlist
for req in pathreqlist :
req.nodes_list.append(req.destination)
# we assume that the destination is a strict constraint
req.loose_list.append('strict')
if req in pathreqlist_simple:
path_res_list.append(compute_constrained_path(network, req))
else:
path_res_list.append(pathreslist_disjoint[req])
return path_res_list
def isdisjoint(p1,p2) :
# returns 0 if disjoint
edge1 = list(pairwise(p1))
edge2 = list(pairwise(p2))
for e in edge1 :
if e in edge2 :
return 1
return 0
def find_reversed_path(p,network) :
# select of intermediate roadms and find the path between them
# note that this function may not give an exact result in case of multiple
# links between two adjacent nodes.
# TODO add some indication on elements to indicate from which other they
# are the reversed direction
reversed_roadm_path = list(reversed([e for e in p if isinstance (e,Roadm)]))
source = p[-1]
destination = p[0]
total_path = [source]
for node in reversed_roadm_path :
total_path.extend(dijkstra_path(network, source, node)[1:])
source = node
total_path.append(destination)
return total_path
def ispart(a,b) :
j = 0
for i, el in enumerate(a):
if el in b :
if b.index(el) >= j :
j = b.index(el)
else:
return False
else:
return False
return True
def remove_candidate(candidates, allpaths, rq, pth) :
# print(f'coucou {rq.request_id}')
for key, candidate in candidates.items() :
temp = candidate.copy()
for i,sol in enumerate(candidate) :
for p in sol :
if allpaths[id(p)].req.request_id == rq.request_id :
if id(p) != id(pth) :
temp.remove(sol)
break
candidates[key] = temp
return candidates