def distribution_cost(
distribution: Distribution,
computation_graph: ComputationGraph,
agentsdef: Iterable[AgentDef],
computation_memory: Callable[[ComputationNode], float],
communication_load: Callable[[ComputationNode, str], float],
) -> float:
route = route_fonc(agentsdef)
msg_load = msg_load_func(computation_graph, communication_load)
hosting_cost = hosting_cost_func(agentsdef)
comm = 0
agt_names = [a.name for a in agentsdef]
for l in computation_graph.links:
# As we support hypergraph, we may have more than 2 ends to a link
for c1, c2 in combinations(l.nodes, 2):
a1 = distribution.agent_for(c1)
a2 = distribution.agent_for(c2)
comm += route(a1, a2) * msg_load(c1, c2)
hosting = 0
for computation in computation_graph.nodes:
agent = distribution.agent_for(computation.name)
hosting += hosting_cost(agent, computation.name)
cost = RATIO_HOST_COMM * comm + (1 - RATIO_HOST_COMM) * hosting
return cost, comm, hosting
def test_dist_2(self):
d = Distribution({'a1': ['v1', 'v2'], 'a2': ['v3']})
self.assertEqual(len(d.computations_hosted('a1')), 2)
self.assertEqual(len(d.computations_hosted('a2')), 1)
self.assertIn('v1', d.computations_hosted('a1'))
self.assertIn('v2', d.computations_hosted('a1'))
self.assertEqual(d.agent_for('v1'), 'a1')
self.assertEqual(d.agent_for('v2'), 'a1')
self.assertEqual(d.agent_for('v3'), 'a2')
def test_host_on_new_agent(self):
d = Distribution({'a1': ['v1', 'v2'], 'a2': ['v3']})
d.host_on_agent('a3', ['v4'])
self.assertEqual(d.agent_for('v4'), 'a3')
self.assertIn('a3', d.agents)
self.assertIn('v4', d.computations_hosted('a3'))
def distribution_cost(
distribution: Distribution, computation_graph: ComputationGraph,
agentsdef: Iterable[AgentDef],
computation_memory: Callable[[ComputationNode], float],
communication_load: Callable[[ComputationNode, str], float]) -> float:
"""
Compute the cost for a distribution.
In this model, the cost only includes the communication costs based on message size.
Parameters
----------
distribution
computation_graph
agentsdef
computation_memory
communication_load
Returns
-------
"""
# No hosting and route cost here, as this distribution only takes message size
# into account:
# route = route_fonc(agentsdef)
# hosting_cost = hosting_cost_func(agentsdef)
comm = 0
agt_names = [a.name for a in agentsdef]
for l in computation_graph.links:
# As we support hypergraph, we may have more than 2 ends to a link
for c1, c2 in combinations(l.nodes, 2):
if distribution.agent_for(c1) != distribution.agent_for(c2):
edge_cost = communication_load(
computation_graph.computation(c1), c2)
logger.debug(f"edge cost between {c1} and {c2} : {edge_cost}")
comm += edge_cost
else:
logger.debug(
f"On same agent, no edge cost between {c1} and {c2}")
# This distribution model only takes communication cost into account.
# cost = RATIO_HOST_COMM * comm + (1-RATIO_HOST_COMM) * hosting
return comm, comm, 0
def distribution_cost(
distribution: Distribution,
computation_graph: ComputationsFactorGraph,
agentsdef: Iterable[AgentDef],
computation_memory: Callable[[ComputationNode], float],
communication_load: Callable[[ComputationNode, str], float],
) -> float:
"""
Compute the cost of the distribution.
Only takes communication costs into account (no hosting nor route costs).
Parameters
----------
distribution
computation_graph
agentsdef
computation_memory
communication_load
Returns
-------
"""
comm = 0
agt_names = [a.name for a in agentsdef]
for l in computation_graph.links:
# As we support hypergraph, we may have more than 2 ends to a link
for c1, c2 in combinations(l.nodes, 2):
if distribution.agent_for(c1) != distribution.agent_for(c2):
edge_cost = communication_load(
computation_graph.computation(c1), c2)
logger.debug(f"edge cost between {c1} and {c2} : {edge_cost}")
comm += edge_cost
else:
logger.debug(
f"On same agent, no edge cost between {c1} and {c2}")
# This distribution model only takes communication cost into account.
# cost = RATIO_HOST_COMM * comm + (1-RATIO_HOST_COMM) * hosting
return comm, comm, 0
def fg_secp_ilp(
cg: ComputationsFactorGraph,
agents: List[AgentDef],
already_assigned: Distribution,
computation_memory: Callable[[ComputationNode], float],
communication_load: Callable[[ComputationNode, str], float],
) -> Distribution:
variables = [n for n in cg.nodes if n.type == "VariableComputation"]
factors = [n for n in cg.nodes if n.type == "FactorComputation"]
agents = list(agents)
agents_names = [a.name for a in agents]
# Only keep computations for which we actually need to find an agent.
vars_to_host = [
v.name for v in variables
if not already_assigned.has_computation(v.name)
]
facs_to_host = [
f.name for f in factors if not already_assigned.has_computation(f.name)
]
# x_i^k : binary variable indicating if var x_i is hosted on agent a_k.
xs = _build_xs_binvar(vars_to_host, agents_names)
# f_j^k : binary variable indicating if factor f_j is hosted on agent a_k.
fs = _build_fs_binvar(facs_to_host, agents_names)
# alpha_ijk : binary variable indicating if x_i and f_j are both on a_k.
alphas = _build_alphaijk_binvars(cg, agents_names)
logger.debug(f"alpha_ijk {alphas}")
# LP problem with objective function (total communication cost).
pb = LpProblem("distribution", LpMinimize)
pb += (
secp_dist_objective_function(cg, communication_load, alphas,
agents_names),
"Communication costs",
)
# Constraints.
# All variable computations must be hosted:
for i in vars_to_host:
pb += (
lpSum([xs[(i, k)] for k in agents_names]) == 1,
"var {} is hosted".format(i),
)
# All factor computations must be hosted:
for j in facs_to_host:
pb += (
lpSum([fs[(j, k)] for k in agents_names]) == 1,
"factor {} is hosted".format(j),
)
# Each agent must host at least one computation:
# We only need this constraints for agents that do not already host a
# computation:
empty_agents = [
a for a in agents_names if not already_assigned.computations_hosted(a)
]
for k in empty_agents:
pb += (
lpSum([xs[(i, k)] for i in vars_to_host]) +
lpSum([fs[(j, k)] for j in facs_to_host]) >= 1,
"atleastone {}".format(k),
)
# Memory capacity constraint for agents
for a in agents:
# Decrease capacity for already hosted computations
capacity = a.capacity - sum([
secp_computation_memory_in_cg(c, cg, computation_memory)
for c in already_assigned.computations_hosted(a.name)
])
pb += (
lpSum([
secp_computation_memory_in_cg(i, cg, computation_memory) * xs[
(i, a.name)] for i in vars_to_host
]) + lpSum([
secp_computation_memory_in_cg(j, cg, computation_memory) * fs[
(j, a.name)] for j in facs_to_host
]) <= capacity,
"memory {}".format(a.name),
)
# Linearization constraints for alpha_ijk.
for link in cg.links:
i, j = link.variable_node, link.factor_node
for k in agents_names:
if i in vars_to_host and j in facs_to_host:
pb += alphas[((i, j), k)] <= xs[(i, k)], "lin1 {}{}{}".format(
i, j, k)
pb += alphas[((i, j), k)] <= fs[(j, k)], "lin2 {}{}{}".format(
i, j, k)
pb += (
alphas[((i, j), k)] >= xs[(i, k)] + fs[(j, k)] - 1,
"lin3 {}{}{}".format(i, j, k),
)
elif i in vars_to_host and j not in facs_to_host:
# Var is free, factor is already hosted
if already_assigned.agent_for(j) == k:
pb += alphas[((i, j), k)] == xs[(i, k)]
else:
pb += alphas[((i, j), k)] == 0
elif i not in vars_to_host and j in facs_to_host:
# if i is not in vars_vars_to_host, it means that it's a
# computation that is already hosted (from hints)
if already_assigned.agent_for(i) == k:
pb += alphas[((i, j), k)] == fs[(j, k)]
else:
pb += alphas[((i, j), k)] == 0
else:
# i and j are both alredy hosted
if (already_assigned.agent_for(i) == k
and already_assigned.agent_for(j) == k):
pb += alphas[((i, j), k)] == 1
else:
pb += alphas[((i, j), k)] == 0
# Now solve our LP
# status = pb.solve(GLPK_CMD())
# status = pb.solve(GLPK_CMD(mip=1))
# status = pb.solve(GLPK_CMD(mip=0, keepFiles=1,
# options=['--simplex', '--interior']))
status = pb.solve(GLPK_CMD(keepFiles=0, msg=False, options=["--pcost"]))
if status != LpStatusOptimal:
raise ImpossibleDistributionException("No possible optimal"
" distribution ")
else:
logger.debug("GLPK cost : %s", pulp.value(pb.objective))
comp_dist = already_assigned
for k in agents_names:
agt_vars = [
i for i, ka in xs if ka == k and pulp.value(xs[(i, ka)]) == 1
]
comp_dist.host_on_agent(k, agt_vars)
agt_rels = [
j for j, ka in fs if ka == k and pulp.value(fs[(j, ka)]) == 1
]
comp_dist.host_on_agent(k, agt_rels)
return comp_dist
def cg_secp_ilp(
cg: ComputationConstraintsHyperGraph,
agents: List[AgentDef],
already_assigned: Distribution,
computation_memory: Callable[[ComputationNode], float],
communication_load: Callable[[ComputationNode, str], float],
timeout=600, # Max 10 min
) -> Distribution:
start_t = time.time()
agents = list(agents)
agents_names = [a.name for a in agents]
# Only keep computations for which we actually need to find an agent.
comps_to_host = [
c for c in cg.node_names() if not already_assigned.has_computation(c)
]
# x_i^k : binary variable indicating if var x_i is hosted on agent a_k.
xs = _build_cs_binvar(comps_to_host, agents_names)
# alpha_ijk : binary variable indicating if x_i and f_j are both on a_k.
alphas = _build_alphaijk_binvars(cg, agents_names)
logger.debug(f"alpha_ijk {alphas}")
# LP problem with objective function (total communication cost).
pb = LpProblem("distribution", LpMinimize)
pb += (
_objective_function(cg, communication_load, alphas, agents_names),
"Communication costs",
)
# Constraints.
# All variable computations must be hosted:
for i in comps_to_host:
pb += (
lpSum([xs[(i, k)] for k in agents_names]) == 1,
"var {} is hosted".format(i),
)
# Each agent must host at least one computation:
# We only need this constraints for agents that do not already host a
# computation:
empty_agents = [
a for a in agents_names if not already_assigned.computations_hosted(a)
]
for k in empty_agents:
pb += (
lpSum([xs[(i, k)] for i in comps_to_host]) >= 1,
"atleastone {}".format(k),
)
# Memory capacity constraint for agents
for a in agents:
# Decrease capacity for already hosted computations
capacity = a.capacity - sum([
secp_computation_memory_in_cg(c, cg, computation_memory)
for c in already_assigned.computations_hosted(a.name)
])
pb += (
lpSum([
secp_computation_memory_in_cg(i, cg, computation_memory) *
xs[(i, a.name)] for i in comps_to_host
]) <= capacity,
"memory {}".format(a.name),
)
# Linearization constraints for alpha_ijk.
for (i, j), k in alphas:
if i in comps_to_host and j in comps_to_host:
pb += alphas[((i, j), k)] <= xs[(i, k)], "lin1 {}{}{}".format(
i, j, k)
pb += alphas[((i, j), k)] <= xs[(j, k)], "lin2 {}{}{}".format(
i, j, k)
pb += (
alphas[((i, j), k)] >= xs[(i, k)] + xs[(j, k)] - 1,
"lin3 {}{}{}".format(i, j, k),
)
elif i in comps_to_host and j not in comps_to_host:
# Var is free, factor is already hosted
if already_assigned.agent_for(j) == k:
pb += alphas[((i, j), k)] == xs[(i, k)]
else:
pb += alphas[((i, j), k)] == 0
elif i not in comps_to_host and j in comps_to_host:
# if i is not in vars_vars_to_host, it means that it's a
# computation that is already hosted (from hints)
if already_assigned.agent_for(i) == k:
pb += alphas[((i, j), k)] == xs[(j, k)]
else:
pb += alphas[((i, j), k)] == 0
else:
# i and j are both alredy hosted
if (already_assigned.agent_for(i) == k
and already_assigned.agent_for(j) == k):
pb += alphas[((i, j), k)] == 1
else:
pb += alphas[((i, j), k)] == 0
# the timeout for the solver must be monierd by the time spent to build the pb:
remaining_time = round(timeout - (time.time() - start_t)) - 2
# Now solve our LP
status = pb.solve(
GLPK_CMD(keepFiles=0,
msg=False,
options=["--pcost", "--tmlim",
str(remaining_time)]))
if status != LpStatusOptimal:
raise ImpossibleDistributionException("No possible optimal"
" distribution ")
else:
logger.debug("GLPK cost : %s", pulp.value(pb.objective))
comp_dist = already_assigned
for k in agents_names:
agt_vars = [
i for i, ka in xs if ka == k and pulp.value(xs[(i, ka)]) == 1
]
comp_dist.host_on_agent(k, agt_vars)
return comp_dist
def factor_graph_lp_model(cg: ComputationsFactorGraph,
agents: List[AgentDef],
must_host: Dict[str, List],
computation_memory=None,
communication_load=None):
"""
To distribute we need:
* com : the communication cost of an edge between a var and a fact
* mem_var : the memory footprint of a variable computation
* mem_fac : the memory footprint of a factor computation
These function depends on the algorithm.
Here
* mem_var and mem_fac are given by the computation_memory method.
* com is given by computation_memory
:return:
"""
variables = [n for n in cg.nodes if n.type == 'VariableComputation']
factors = [n for n in cg.nodes if n.type == 'FactorComputation']
agents = list(agents)
agents_names = [a.name for a in agents]
fixed_dist = Distribution(must_host)
# Only keep computations for which we actually need to find an agent.
vars_to_host = [
v.name for v in variables if not fixed_dist.has_computation(v.name)
]
facs_to_host = [
f.name for f in factors if not fixed_dist.has_computation(f.name)
]
# x_i^k : binary variable indicating if var x_i is hosted on agent a_k.
xs = _build_xs_binvar(vars_to_host, agents_names)
# f_j^k : binary variable indicating if factor f_j is hosted on agent a_k.
fs = _build_fs_binvar(facs_to_host, agents_names)
# alpha_ijk : binary variable indicating if x_i and f_j are both on a_k.
alphas = _build_alphaijk_binvars(cg, agents_names)
# LP problem with objective function (total communication cost).
pb = LpProblem('distribution', LpMinimize)
pb += _objective_function(cg, communication_load, alphas,
agents_names), 'Communication costs'
# Constraints.
# All variable computations must be hosted:
for i in vars_to_host:
pb += lpSum([xs[(i, k)] for k in agents_names]) == 1, \
'var {} is hosted'.format(i)
# All factor computations must be hosted:
for j in facs_to_host:
pb += lpSum([fs[(j, k)] for k in agents_names]) == 1, \
'factor {} is hosted'.format(j)
# Each agent must host at least one computation:
# We only need this constraints for agents that do not already host a
# computation:
empty_agents = [a for a in agents_names if not must_host[a]]
for k in empty_agents:
pb += lpSum([xs[(i, k)] for i in vars_to_host]) + \
lpSum([fs[(j, k)] for j in facs_to_host]) >= 1, \
'atleastone {}'.format(k)
# Memory capacity constraint for agents
for a in agents:
# Decrease capacity for already hosted computations
capacity = a.capacity - \
sum([_computation_memory_in_cg(c, cg, computation_memory)
for c in must_host[a.name]])
pb += lpSum([_computation_memory_in_cg(i, cg, computation_memory) *
xs[(i, a.name)] for i in vars_to_host]) \
+ lpSum([_computation_memory_in_cg(j, cg, computation_memory) *
fs[(j, a.name)] for j in facs_to_host]) <= capacity, \
'memory {}'.format(a.name)
# Linearization constraints for alpha_ijk.
for link in cg.links:
i, j = link.variable_node, link.factor_node
for k in agents_names:
if i in vars_to_host and j in facs_to_host:
pb += alphas[((i, j), k)] <= xs[(i, k)], \
'lin1 {}{}{}'.format(i, j, k)
pb += alphas[((i, j), k)] <= fs[(j, k)], \
'lin2 {}{}{}'.format(i, j, k)
pb += alphas[((i, j), k)] >= xs[(i, k)] + fs[(j, k)] - 1, \
'lin3 {}{}{}'.format(i, j, k)
elif i in vars_to_host and j not in facs_to_host:
# Var is free, factor is already hosted
if fixed_dist.agent_for(j) == k:
pb += alphas[((i, j), k)] == xs[(i, k)]
else:
pb += alphas[((i, j), k)] == 0
elif i not in vars_to_host and j in facs_to_host:
# if i is not in vars_vars_to_host, it means that it's a
# computation that is already hosted (from hints)
if fixed_dist.agent_for(i) == k:
pb += alphas[((i, j), k)] == fs[(j, k)]
else:
pb += alphas[((i, j), k)] == 0
else:
# i and j are both alredy hosted
if fixed_dist.agent_for(i) == k and fixed_dist.agent_for(j) \
== k:
pb += alphas[((i, j), k)] == 1
else:
pb += alphas[((i, j), k)] == 0
# Now solve our LP
# status = pb.solve(GLPK_CMD())
# status = pb.solve(GLPK_CMD(mip=1))
# status = pb.solve(GLPK_CMD(mip=0, keepFiles=1,
# options=['--simplex', '--interior']))
status = pb.solve(GLPK_CMD(keepFiles=0, msg=False, options=['--pcost']))
if status != LpStatusOptimal:
raise ImpossibleDistributionException("No possible optimal"
" distribution ")
else:
logger.debug('GLPK cost : %s', value(pb.objective))
comp_dist = fixed_dist
for k in agents_names:
agt_vars = [
i for i, ka in xs if ka == k and value(xs[(i, ka)]) == 1
]
comp_dist.host_on_agent(k, agt_vars)
agt_rels = [
j for j, ka in fs if ka == k and value(fs[(j, ka)]) == 1
]
comp_dist.host_on_agent(k, agt_rels)
return comp_dist