load_balancing.py¶
This example is inspired by a blog entry from yhathq. The article is located here.
This example looks at cloud load balancing to keep a service running in the cloud at reasonable cost by reducing the expense of running cloud servers, minimizing risk and human time due to rebalancing, and doing balance sleeping models across servers.
The different KPIs are optimized in a hierarchical manner: first, the number of active servers is minimized,
then the total number of migrations is minimized,
and finally the sleeping workload is balanced.
This optimization is achieved by the lexicographic_solve()
method.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 | # --------------------------------------------------------------------------
# Source file provided under Apache License, Version 2.0, January 2004,
# http://www.apache.org/licenses/
# (c) Copyright IBM Corp. 2015, 2016
# --------------------------------------------------------------------------
# Source: http://blog.yhathq.com/posts/how-yhat-does-cloud-balancing.html
from collections import namedtuple
import json
from docplex.util.environment import get_environment
from docplex.mp.absmodel import AbstractModel
class TUser(namedtuple("TUser", ["id", "running", "sleeping", "current_server"])):
def __str__(self):
return self.id
DEFAULT_MAX_PROCESSES_PER_SERVER = 50
class LoadBalancingModel(AbstractModel):
def __init__(self, **kwargs):
AbstractModel.__init__(self, 'load_balancing', **kwargs)
# raw data
self.max_processes_per_server = DEFAULT_MAX_PROCESSES_PER_SERVER
self.servers = []
self.users = []
# decision objects
self.active_var_by_server = {}
self.assign_user_to_server_vars = {}
self.number_of_active_servers = None
self.number_of_migrations = None
self.max_sleeping_workload = None
def load_data(self, servers, users, max_process_per_server=DEFAULT_MAX_PROCESSES_PER_SERVER):
self.servers = servers
self.users = [TUser(*user_row) for user_row in users]
self.max_processes_per_server = max_process_per_server
def clear(self):
AbstractModel.clear(self)
self.active_var_by_server = {}
self.assign_user_to_server_vars = {}
self.number_of_active_servers = None
def setup_variables(self):
all_servers = self.servers
all_users = self.users
self.active_var_by_server = self.binary_var_dict(all_servers, name='isActive')
def user_server_pair_namer(u_s):
u, s = u_s
return '%s_to_%s' % (u.id, s)
self.assign_user_to_server_vars = self.binary_var_matrix(all_users, all_servers, user_server_pair_namer)
@staticmethod
def _is_migration(user, server):
""" Returns True if server is not the user's current
Used in setup of constraints.
"""
return server != user.current_server
def setup_constraints(self):
mdl = self
all_servers = self.servers
all_users = self.users
max_proc_per_server = self.max_processes_per_server
mdl.add_constraints(
mdl.sum(self.assign_user_to_server_vars[u, s] * u.running for u in all_users) <= max_proc_per_server
for s in all_servers)
# each assignment var <u, s> is <= active_server(s)
for s in all_servers:
for u in all_users:
ct_name = 'ct_assign_to_active_{0!s}_{1!s}'.format(u, s)
mdl.add_constraint(self.assign_user_to_server_vars[u, s] <= self.active_var_by_server[s], ct_name)
# sum of assignment vars for (u, all s in servers) == 1
for u in all_users:
ct_name = 'ct_unique_server_%s' % (u[0])
mdl.add_constraint(mdl.sum((self.assign_user_to_server_vars[u, s] for s in all_servers)) == 1.0, ct_name)
def setup_objective(self):
mdl = self
self.number_of_active_servers = mdl.sum((self.active_var_by_server[svr] for svr in self.servers))
self.add_kpi(self.number_of_active_servers, "Number of active servers")
self.number_of_migrations = mdl.sum(
self.assign_user_to_server_vars[u, s] for u in self.users for s in self.servers if
self._is_migration(u, s))
mdl.add_kpi(self.number_of_migrations, "Total number of migrations")
max_sleeping_workload = mdl.integer_var(name="max_sleeping_processes")
for s in self.servers:
ct_name = 'ct_define_max_sleeping_%s' % s
mdl.add_constraint(
mdl.sum(
self.assign_user_to_server_vars[u, s] * u.sleeping for u in self.users) <= max_sleeping_workload,
ct_name)
mdl.add_kpi(max_sleeping_workload, "Max sleeping workload")
self.max_sleeping_workload = max_sleeping_workload
# Set objective function
mdl.minimize(self.number_of_active_servers)
def run(self, **kwargs):
mdl = self
mdl.ensure_setup()
mdl.print_information()
# build an ordered sequence of goals
ordered_kpi_keywords = ["servers", "migrations", "sleeping"]
ordered_goals = [mdl.kpi_by_name(k) for k in ordered_kpi_keywords]
return mdl.solve_lexicographic(ordered_goals, **kwargs)
def report(self):
mdl = self
active_servers = sorted([s for s in mdl.servers if mdl.active_var_by_server[s].solution_value == 1])
print("Active Servers: {}".format(active_servers))
print("*** User assignment ***")
for (u, s) in sorted(mdl.assign_user_to_server_vars):
if mdl.assign_user_to_server_vars[(u, s)].solution_value == 1:
print("{} uses {}, migration: {}".format(u, s, "yes" if mdl._is_migration(u, s) else "no"))
print("*** Servers sleeping processes ***")
for s in active_servers:
sleeping = sum(self.assign_user_to_server_vars[u, s].solution_value * u.sleeping for u in self.users)
print("Server: {} #sleeping={}".format(s, sleeping))
def save_solution_as_json(self, json_file):
"""Saves the solution for this model as JSON.
Note that this is not a CPLEX Solution file, as this is the result of post-processing a CPLEX solution
"""
mdl = self
solution_dict = {}
# active server
active_servers = sorted([s for s in mdl.servers if mdl.active_var_by_server[s].solution_value == 1])
solution_dict["active servers"] = active_servers
# sleeping processes by server
sleeping_processes = {}
for s in active_servers:
sleeping = sum(self.assign_user_to_server_vars[u, s].solution_value * u.sleeping for u in self.users)
sleeping_processes[s] = sleeping
solution_dict["sleeping processes by server"] = sleeping_processes
# user assignment
user_assignment = []
for (u, s) in sorted(mdl.assign_user_to_server_vars):
if mdl.assign_user_to_server_vars[(u, s)].solution_value == 1:
n = {
'user': u.id,
'server': s,
'migration': "yes" if mdl._is_migration(u, s) else "no"
}
user_assignment.append(n)
solution_dict['user assignment'] = user_assignment
json_file.write(json.dumps(solution_dict, indent=3).encode('utf-8'))
SERVERS = ["server002", "server003", "server001", "server006", "server007", "server004", "server005"]
USERS = [("user013", 2, 1, "server002"),
("user014", 0, 2, "server002"),
("user015", 0, 4, "server002"),
("user016", 1, 4, "server002"),
("user017", 0, 3, "server002"),
("user018", 0, 2, "server002"),
("user019", 0, 2, "server002"),
("user020", 0, 1, "server002"),
("user021", 4, 4, "server002"),
("user022", 0, 1, "server002"),
("user023", 0, 3, "server002"),
("user024", 1, 2, "server002"),
("user025", 0, 1, "server003"),
("user026", 0, 1, "server003"),
("user027", 1, 1, "server003"),
("user028", 0, 1, "server003"),
("user029", 2, 1, "server003"),
("user030", 0, 5, "server003"),
("user031", 0, 2, "server003"),
("user032", 0, 3, "server003"),
("user033", 1, 1, "server003"),
("user034", 0, 1, "server003"),
("user035", 0, 1, "server003"),
("user036", 4, 1, "server003"),
("user037", 7, 1, "server003"),
("user038", 2, 1, "server003"),
("user039", 0, 3, "server003"),
("user040", 1, 2, "server003"),
("user001", 0, 2, "server001"),
("user002", 0, 3, "server001"),
("user003", 5, 4, "server001"),
("user004", 0, 1, "server001"),
("user005", 0, 1, "server001"),
("user006", 0, 2, "server001"),
("user007", 0, 4, "server001"),
("user008", 0, 1, "server001"),
("user009", 5, 1, "server001"),
("user010", 7, 1, "server001"),
("user011", 4, 5, "server001"),
("user012", 0, 4, "server001"),
("user062", 0, 1, "server006"),
("user063", 3, 5, "server006"),
("user064", 0, 1, "server006"),
("user065", 0, 3, "server006"),
("user066", 3, 1, "server006"),
("user067", 0, 1, "server006"),
("user068", 0, 1, "server006"),
("user069", 0, 2, "server006"),
("user070", 3, 2, "server006"),
("user071", 0, 1, "server006"),
("user072", 5, 3, "server006"),
("user073", 0, 1, "server006"),
("user074", 0, 1, "server006"),
("user075", 0, 2, "server007"),
("user076", 1, 1, "server007"),
("user077", 1, 1, "server007"),
("user078", 0, 1, "server007"),
("user079", 0, 3, "server007"),
("user080", 0, 1, "server007"),
("user081", 4, 1, "server007"),
("user082", 1, 1, "server007"),
("user041", 0, 1, "server004"),
("user042", 2, 1, "server004"),
("user043", 5, 2, "server004"),
("user044", 5, 2, "server004"),
("user045", 0, 2, "server004"),
("user046", 1, 5, "server004"),
("user047", 0, 1, "server004"),
("user048", 0, 3, "server004"),
("user049", 5, 1, "server004"),
("user050", 0, 2, "server004"),
("user051", 0, 3, "server004"),
("user052", 0, 3, "server004"),
("user053", 0, 1, "server004"),
("user054", 0, 2, "server004"),
("user055", 0, 3, "server005"),
("user056", 3, 1, "server005"),
("user057", 0, 3, "server005"),
("user058", 0, 2, "server005"),
("user059", 0, 1, "server005"),
("user060", 0, 5, "server005"),
("user061", 0, 2, "server005")
]
class DefaultLoadBalancingModel(LoadBalancingModel):
def __init__(self, context=None, **kwargs):
LoadBalancingModel.__init__(self, context=context, **kwargs)
self.load_data(SERVERS, USERS)
if __name__ == '__main__':
"""DOcplexcloud credentials can be specified with url and api_key in the code block below.
Alternatively, Context.make_default_context() searches the PYTHONPATH for
the following files:
* cplex_config.py
* cplex_config_<hostname>.py
* docloud_config.py (must only contain context.solver.docloud configuration)
These files contain the credentials and other properties. For example,
something similar to::
context.solver.docloud.url = "https://docloud.service.com/job_manager/rest/v1"
context.solver.docloud.key = "example api_key"
"""
url = None # put your url here
key = None # put your api key here
lbm = DefaultLoadBalancingModel()
# Run the model. If a key has been specified above, the model will run on
# IBM Decision Optimization on cloud.
if lbm.run(url=url, key=key):
lbm.report()
with get_environment().get_output_stream("solution.json") as fp:
lbm.save_solution_as_json(fp)
|