def test_te_app():
""" Test a single traffic engineering app"""
topo = complete_topology(4)
for link in topo.links():
topo.set_resource(link, BANDWIDTH, 1)
tcs = [TrafficClass(0, u'classname', 0, 2, array([1]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
appconfig = {
'name': u'te',
'constraints': [(Constraint.ROUTE_ALL, (), {})],
'obj': (Objective.MIN_LINK_LOAD, (BANDWIDTH, ), {}),
'resource_cost': {
BANDWIDTH: (LINKS, 1, None)
}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(BANDWIDTH, cap=1)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# THE solution is 1-objective because of the maximization flip
solution = 1 - opt.get_solved_objective(app)[0]
# Use abs(actual - exprected) because floating point errors
assert solution == .333333 or abs(solution - .33333) <= EPSILON
solution = 1 - opt.get_solved_objective()
assert solution == .333333 or abs(solution - .33333) <= EPSILON
def test_maxflow(cap):
""" Check that maxflow works correctly, for a single traffic class """
# Generate a topology:
topo = complete_topology(4)
for link in topo.links():
topo.set_resource(link, BANDWIDTH, 1)
tcs = [TrafficClass(0, u'classname', 0, 2, array([3]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
appconfig = {
'name': u'mf',
'constraints': [],
'obj': (Objective.MAX_FLOW, (), {}),
'resource_cost': {
BANDWIDTH: (LINKS, 1, None)
}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(BANDWIDTH, cap=cap)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# Ensure that both app objective and global objective are the same
# Also, use abs(actual - exprected) because floating point errors
solution = opt.get_solved_objective(app)[0]
assert solution == cap or abs(solution - cap) <= EPSILON
solution = opt.get_solved_objective()
assert solution == cap or abs(solution - cap) <= EPSILON
def test_mbox_load_balancing():
"""Test the middlebox loadbalancing"""
topo = complete_topology(4)
for n in topo.nodes():
topo.set_resource(n, CPU, 1)
topo.set_mbox(n)
tcs = [TrafficClass(0, u'classname', 0, 2, array([1]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo,
tcs,
has_mbox_predicate,
modify_func=use_mbox_modifier,
cutoff=100)
appconfig = {
'name': u'mb_lb',
'constraints': [(Constraint.ROUTE_ALL, (), {})],
'obj': (Objective.MIN_NODE_LOAD, (CPU, ), {}),
'resource_cost': {
CPU: (MBOXES, 1, None)
}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(CPU, cap=1)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# THE solution is 1-objective because of the maximization flip
solution = 1 - opt.get_solved_objective(app)[0]
# Use abs(actual - exprected) because floating point errors
assert solution == .25 or abs(solution - .25) <= EPSILON
solution = 1 - opt.get_solved_objective()
assert solution == .25 or abs(solution - .25) <= EPSILON
def pptc(topo):
# generate a dummy TM and traffic classes
tm = tmgen.exact_tm(topo.num_nodes(), 1)
tc = traffic_classes(tm, {u'all': 1}, as_dict=False)
# generate all possibe paths
res = generate_paths_tc(topo, tc, null_predicate, 10, max_paths=float('inf'))
return res
def test_min_latency_app():
"""Test a single min latency app"""
topo = complete_topology(4)
for link in topo.links():
topo.set_resource(link, BANDWIDTH, 1)
tcs = [TrafficClass(0, u'classname', 0, 2, array([1]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
appconfig = {
'name': u'te',
'constraints': [(Constraint.ROUTE_ALL, (), {})],
'obj': (Objective.MIN_LATENCY, (), {}),
'resource_cost': {
BANDWIDTH: (LINKS, 1, None)
}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(BANDWIDTH, cap=1)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# norm factor for latency is diameter * n^2
norm = topo.diameter() * 16
# the objective is 1-normalized latency, and latency is 1.
# because 1 path with flow fraction of 1.
solution = opt.get_solved_objective(app)[0]
assert solution == 1 - 1 / norm or abs(solution -
(1 - 1 / norm)) <= EPSILON
solution = opt.get_solved_objective()
assert solution == 1 - 1 / norm or abs(solution -
(1 - 1 / norm)) <= EPSILON
def test_mbox_load_balancing_all_tcs():
"""Test the middlebox loadbalancing"""
topo = complete_topology(4)
for n in topo.nodes():
topo.set_resource(n, CPU, 1)
topo.set_mbox(n)
tcs = [TrafficClass(0, u'classname', s, t, array([1])) for (s, t) in product(topo.nodes(), repeat=2)]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, has_mbox_predicate, modify_func=use_mbox_modifier, cutoff=100)
appconfig = {
'name': u'mb_lb',
'constraints': [(Constraint.ROUTE_ALL, (), {})],
'obj': (Objective.MIN_NODE_LOAD, (CPU,), {}),
'resource_cost': {CPU: (MBOXES, 1, None)}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(CPU, cap=1)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# THE solution is 1-objective because of the maximization flip
solution = 1 - opt.get_solved_objective(app)[0]
# Use abs(actual - exprected) because floating point errors
assert solution == 1 or abs(solution - 1) <= EPSILON
solution = 1 - opt.get_solved_objective()
assert solution == 1 or abs(solution - 1) <= EPSILON
def test_maxflow_inapp_caps(cap):
"""Text maxflow, but use the CAP constraint instead of global network caps"""
# Generate a topology:
topo = complete_topology(4)
for link in topo.links():
topo.set_resource(link, BANDWIDTH, 1)
tcs = [TrafficClass(0, u'classname', 0, 2, array([3]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
caps = {link: cap for link in topo.links()}
appconfig = {
'name': u'mf',
'constraints': [(Constraint.CAP_LINKS, (BANDWIDTH, caps), {})],
'obj': (Objective.MAX_FLOW, (), {}),
'resource_cost': {BANDWIDTH: (LINKS, 1, None)}
}
app = App(pptc, **appconfig)
opt = from_app(topo, app, NetworkConfig())
opt.solve()
assert opt.is_solved()
# Ensure that both app objective and global objective are the same
# Also, use abs(actual - exprected) because floating point errors
solution = opt.get_solved_objective(app)[0]
assert solution == cap or abs(solution - cap) <= EPSILON
solution = opt.get_solved_objective()
assert solution == cap or abs(solution - cap) <= EPSILON
def test_te_app():
""" Test a single traffic engineering app"""
topo = complete_topology(4)
for link in topo.links():
topo.set_resource(link, BANDWIDTH, 1)
tcs = [TrafficClass(0, u'classname', 0, 2, array([1]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
appconfig = {
'name': u'te',
'constraints': [(Constraint.ROUTE_ALL, (), {})],
'obj': (Objective.MIN_LINK_LOAD, (BANDWIDTH,), {}),
'resource_cost': {BANDWIDTH: (LINKS, 1, None)}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(BANDWIDTH, cap=1)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# THE solution is 1-objective because of the maximization flip
solution = 1 - opt.get_solved_objective(app)[0]
# Use abs(actual - exprected) because floating point errors
assert solution == .333333 or abs(solution - .33333) <= EPSILON
solution = 1 - opt.get_solved_objective()
assert solution == .333333 or abs(solution - .33333) <= EPSILON
def test_min_latency_app():
"""Test a single min latency app"""
topo = complete_topology(4)
for link in topo.links():
topo.set_resource(link, BANDWIDTH, 1)
tcs = [TrafficClass(0, u'classname', 0, 2, array([1]))]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
appconfig = {
'name': u'te',
'constraints': [(Constraint.ROUTE_ALL, (), {})],
'obj': (Objective.MIN_LATENCY, (), {}),
'resource_cost': {BANDWIDTH: (LINKS, 1, None)}
}
app = App(pptc, **appconfig)
caps = NetworkCaps(topo)
caps.add_cap(BANDWIDTH, cap=1)
opt = from_app(topo, app, NetworkConfig(caps))
opt.solve()
assert opt.is_solved()
# norm factor for latency is diameter * n^2
norm = topo.diameter() * 16
# the objective is 1-normalized latency, and latency is 1.
# because 1 path with flow fraction of 1.
solution = opt.get_solved_objective(app)[0]
assert solution == 1 - 1 / norm or abs(solution - (1 - 1 / norm)) <= EPSILON
solution = opt.get_solved_objective()
assert solution == 1 - 1 / norm or abs(solution - (1 - 1 / norm)) <= EPSILON
def pptc(topo):
# generate a dummy TM and traffic classes
tm = tmgen.exact_tm(topo.num_nodes(), 1)
tc = traffic_classes(tm, {u'all': 1}, as_dict=False)
# generate all possibe paths
res = generate_paths_tc(topo,
tc,
null_predicate,
10,
max_paths=float('inf'))
return res
def pptc():
"""
An example paths per traffic class
"""
# get a complete topology
topo = complete_topology(5)
# generate a dummy TM and traffic classes
tm = tmgen.uniform_tm(5, 20, 50, 1)
tc = traffic_classes(tm, {u'all': 1}, as_dict=False)
# generate all possibe paths
res = generate_paths_tc(topo, tc, null_predicate, 10, numpy.inf)
return res
def test_shortest_path():
""" Check that we can correctly implement shortest path routing """
# Generate a topology:
topo = complete_topology(5)
# Generate a single traffic class:
# TrafficClass (id, name, source node, destination node)
tcs = [TrafficClass(0, u'classname', 0, 2)]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
# Application configuration
appconfig = {
'name': u'minLatencyApp',
'constraints': [(Constraint.ROUTE_ALL, (pptc.tcs(), ), {})],
'obj': (Objective.MIN_LATENCY, (), {}),
'resource_cost': {}
}
# Create an application based on our config
app = App(pptc, **appconfig)
# Create and solve an optimization based on the app
# No link capacities will just result in a single shortest path
opt = from_app(topo, app, NetworkConfig(None))
opt.solve()
assert opt.is_solved()
paths = opt.get_paths()
for pi, p in enumerate(paths.paths(tcs[0])):
if list(p.nodes()) == [0, 2]:
assert p.flow_fraction() == 1
else:
assert p.flow_fraction() == 0
# norm factor for latency is diameter * n^2
norm = topo.diameter() * 25
# the objective is 1-normalized latency, and latency is 1.
# because 1 path with flow fraction of 1.
solution = opt.get_solved_objective(app)[0]
assert solution == 1 - 1 / norm or abs(solution - 1 - 1 / norm) <= EPSILON
solution = opt.get_solved_objective()
assert solution == 1 - 1 / norm or abs(solution - 1 - 1 / norm) <= EPSILON
def test_shortest_path():
""" Check that we can correctly implement shortest path routing """
# Generate a topology:
topo = complete_topology(5)
# Generate a single traffic class:
# TrafficClass (id, name, source node, destination node)
tcs = [TrafficClass(0, u'classname', 0, 2)]
# Generate all paths for this traffic class
pptc = generate_paths_tc(topo, tcs, null_predicate, cutoff=100)
# Application configuration
appconfig = {
'name': u'minLatencyApp',
'constraints': [(Constraint.ROUTE_ALL, (pptc.tcs(),), {})],
'obj': (Objective.MIN_LATENCY, (), {}),
'resource_cost': {}
}
# Create an application based on our config
app = App(pptc, **appconfig)
# Create and solve an optimization based on the app
# No link capacities will just result in a single shortest path
opt = from_app(topo, app, NetworkConfig(None))
opt.solve()
assert opt.is_solved()
paths = opt.get_paths()
for pi, p in enumerate(paths.paths(tcs[0])):
if list(p.nodes()) == [0, 2]:
assert p.flow_fraction() == 1
else:
assert p.flow_fraction() == 0
# norm factor for latency is diameter * n^2
norm = topo.diameter() * 25
# the objective is 1-normalized latency, and latency is 1.
# because 1 path with flow fraction of 1.
solution = opt.get_solved_objective(app)[0]
assert solution == 1 - 1 / norm or abs(solution - 1 - 1 / norm) <= EPSILON
solution = opt.get_solved_objective()
assert solution == 1 - 1 / norm or abs(solution - 1 - 1 / norm) <= EPSILON