def test_experiment_full(task_scale=1, cluster_scale=1):
dim = DIM4
experiments_iterator(
'full',
[{}],
[30 * task_scale],
[
(TaskGeneratorEqual,
dict(task_definitions=TASKS_3TYPES, replicas=10 * task_scale, dimensions=dim)),
(TaskGeneratorClasses, dict(task_definitions=TASKS_3TYPES,
counts=dict(mbw=3 * task_scale, cpu=12 * task_scale,
mem=15 * task_scale), dimensions=dim)),
(TaskGeneratorClasses, dict(task_definitions=TASKS_3TYPES,
counts=dict(mbw=20 * task_scale, cpu=5 * task_scale,
mem=5 * task_scale), dimensions=dim)),
(TaskGeneratorClasses, dict(task_definitions=TASKS_3TYPES,
counts=dict(mbw=30 * task_scale, cpu=0 * task_scale,
mem=0 * task_scale), dimensions=dim)),
(TaskGeneratorClasses, dict(task_definitions=TASKS_3TYPES,
counts=dict(mbw=0 * task_scale, cpu=0 * task_scale,
mem=30 * task_scale), dimensions=dim)),
],
[
prepare_nodes(NODES_DEFINITIONS_2TYPES, dict(aep=0, dram=6 * cluster_scale),
dim),
prepare_nodes(NODES_DEFINITIONS_2TYPES,
dict(aep=2 * cluster_scale, dram=6 * cluster_scale), dim),
prepare_nodes(NODES_DEFINITIONS_2TYPES, dict(aep=0, dram=8 * cluster_scale),
dim),
prepare_nodes(NODES_DEFINITIONS_2TYPES, dict(aep=cluster_scale, dram=cluster_scale),
dim, ),
],
[
(NOPAlgorithm, {}),
(Fit, dict(dimensions=DIM2)),
(Fit, dict(dimensions=DIM4)),
(LeastUsed, dict(dimensions=DIM2)),
(LeastUsed, dict(dimensions=DIM4)),
(BAR, dict(dimensions=DIM2)),
(HierBAR, dict(dimensions=DIM2)),
(LeastUsedBAR,
dict(alias='kubernetes_baseline', dimensions={CPU, MEM})),
(LeastUsedBAR, dict(alias='BAR__LU_OFF', dimensions=DIM4, least_used_weight=0)),
(LeastUsedBAR,
dict(alias='BAR__LU_ON__WEIGHTS_EQUAL', dimensions=DIM4, least_used_weight=1)),
(LeastUsedBAR,
dict(alias='BAR__LU_ON__WEIGHTS_UNEQUAL', dimensions=DIM4, least_used_weight=1,
bar_weights={MEM: 0.5})),
],
)
def test_experiment_bar():
task_scale = 1
cluster_scale = 1
nodes_dimensions = DIM4
results = experiments_iterator(
'bar_weights',
[{}],
[30 * task_scale, ],
[
(TaskGeneratorClasses, dict(task_definitions=taskset_dimensions(nodes_dimensions,
TASKS_6TYPES),
counts=dict(mbw=20 * task_scale, cpu=5 * task_scale,
mem=5 * task_scale))),
],
[
prepare_nodes(NODES_DEFINITIONS_2TYPES,
dict(aep=2 * cluster_scale, dram=6 * cluster_scale), nodes_dimensions),
],
[
(LeastUsedBAR,
dict(alias='BAR__LU_ON__WEIGHTS_UNEQUAL', dimensions=DIM4, least_used_weight=1,
bar_weights={MEM: 0.5})),
(LeastUsedBAR,
dict(alias='BAR__LU_ON__WEIGHTS_EQUAL', dimensions=DIM4, least_used_weight=1)),
(LeastUsedBAR,
dict(alias='BAR__LU_ON__WEIGHTS_100x', dimensions=DIM4, least_used_weight=100)),
(LeastUsedBAR, dict(alias='BAR__LU_OFF', dimensions=DIM4, least_used_weight=0)),
],
)
assert len(results) == 4
def test_experiment_debug():
task_scale = 1
cluster_scale = 1
dim = DIM4
length = 90
experiments_iterator(
'debug',
[dict(retry_scheduling=True)],
[length * task_scale],
[
(TaskGeneratorEqual,
dict(task_definitions=TASKS_3TYPES, replicas=10 * task_scale, duration=30,
alias='long', dimensions=dim)),
(TaskGeneratorEqual,
dict(task_definitions=TASKS_3TYPES, replicas=10 * task_scale, duration=5,
alias='short', dimensions=dim)),
(TaskGeneratorEqual,
dict(task_definitions=TASKS_3TYPES, replicas=10 * task_scale, duration=None,
dimensions=dim)),
],
[
prepare_nodes(NODES_DEFINITIONS_2TYPES,
dict(aep=2 * cluster_scale, dram=6 * cluster_scale), dim),
],
[
(LeastUsedBAR, dict(alias='BAR__LU_OFF', dimensions=DIM4, least_used_weight=0)),
],
)
def test_experiment_fit():
dim = DIM2
results = experiments_iterator(
'fit',
[dict(retry_scheduling=True)], # Simulator configuration
[1],
[
(TaskGeneratorEqual,
dict(task_definitions=TASKS_2TYPES, dimensions=dim, replicas=3,
duration=None, node_name='dram_0')),
],
[
prepare_nodes(NODES_DEFINITIONS_2TYPES, dict(aep=1, dram=1), dim),
],
[
(Fit, dict(dimensions=dim)),
],
rmtree=False,
charts=False,
metrics=False, # can be as list of metric names
)
assert len(results) == 1
assert results == [{'ALGO': 'Fit(2)',
'NODES': '2(aep=1,dram=1)',
'SIM': 'retry=1',
'TASKS': '6(cputask=3,memtask=3)',
'assigned%': 100,
'assigned_broken%': 0,
'balance': 0.7465277777777778,
'cpu_util%': 8.333333333333332,
'cpu_util(AEP)%': 0.0,
'mem_util%': 16.778523489932887,
'mem_util(AEP)%': 0.0,
'scheduled': 1,
'utilization%': 12}]
def test_experiment_hierbar():
dim = DIM4
iterations = 30
results = experiments_iterator(
'hierbar',
[{}],
[iterations],
[
(TaskGeneratorEqual, dict(task_definitions=TASKS_3TYPES, replicas=10, dimensions=dim)),
(TaskGeneratorClasses,
dict(task_definitions=TASKS_3TYPES, counts=dict(mem=30), dimensions=dim)), # AEP
(TaskGeneratorClasses,
dict(task_definitions=TASKS_3TYPES, counts=dict(mbw=30), dimensions=dim)), # no AEP
(TaskGeneratorClasses,
dict(task_definitions=TASKS_3TYPES, counts=dict(cpu=5, mem=5, mbw=20),
dimensions=dim)),
(TaskGeneratorClasses,
dict(task_definitions=TASKS_3TYPES, counts=dict(cpu=5, mem=20, mbw=5),
dimensions=dim)),
],
[
prepare_nodes(NODES_DEFINITIONS_3TYPES, dict(aep=2, sml=4, big=2), dim)
],
[
(LeastUsedBAR, dict(dimensions=DIM2, alias='native')),
(HierBAR, dict(dimensions=dim)),
(HierBAR, dict(dimensions=dim, merge_threshold=1, alias='extender')),
],
)
assert len(results) == 15
def test_experiment_score():
dim = DIM2
results = experiments_iterator(
'score',
[ # Simulator & data provider configuration
dict(retry_scheduling=True, data_provider_args=dict(normalization_dimension=CPU)),
dict(retry_scheduling=True, data_provider_args=dict(normalization_dimension=MEM)),
],
[20],
[
(TaskGeneratorEqual,
dict(task_definitions=TASKS_3TYPES, dimensions=dim, replicas=3,
duration=None, node_name='dram_0')),
],
[
prepare_nodes(NODES_DEFINITIONS_2TYPES, dict(aep=1, dram=1), dim),
],
[
(Score, dict(dimensions=dim)),
],
rmtree=False,
charts=False,
metrics=False, # can be as list of metric names
)
assert len(results) == 2
assert results == [{'ALGO': 'Score(2)',
'NODES': '2(aep=1,dram=1)',
'SIM': 'retry=1,norm=cpu',
'TASKS': '9(cpu=3,mbw=3,mem=3)',
'assigned%': 100,
'assigned_broken%': 0,
'balance': 0.90892333984375,
'cpu_util%': 30.0,
'cpu_util(AEP)%': 7.5,
'mem_util%': 38.00335570469799,
'mem_util(AEP)%': 30.0,
'scheduled': 9,
'utilization%': 34},
{'ALGO': 'Score(2)',
'NODES': '2(aep=1,dram=1)',
'SIM': 'retry=1,norm=mem',
'TASKS': '9(cpu=3,mbw=3,mem=3)',
'assigned%': 100,
'assigned_broken%': 0,
'balance': 0.90892333984375,
'cpu_util%': 30.0,
'cpu_util(AEP)%': 7.5,
'mem_util%': 38.00335570469799,
'mem_util(AEP)%': 30.0,
'scheduled': 9,
'utilization%': 34}]
def test_experiment_static_assigner():
dim = DIM2
targeted_assigned_apps_counts = \
{'aep_0': {'cputask': 1, 'memtask': 2},
'dram_0': {'cputask': 1, 'memtask': 0}}
results = experiments_iterator(
'static_assigner',
[{}],
[6],
[
(TaskGeneratorClasses,
dict(task_definitions=TASKS_2TYPES,
counts=dict(cputask=2, memtask=2),
dimensions=dim)),
],
[
prepare_nodes(NODES_DEFINITIONS_2TYPES, dict(aep=1, dram=1), dim),
],
[
(StaticAssigner,
dict(alias='StaticAssigner',
targeted_assigned_apps_counts=targeted_assigned_apps_counts,
dimensions=dim)),
],
)
assert len(results) == 1
result = results[0]
assert result == {
'ALGO': 'StaticAssigner',
'NODES': '2(aep=1,dram=1)',
'SIM': 'default',
'TASKS': '4(cputask=2,memtask=2)',
'assigned%': 100,
'assigned_broken%': 0,
'balance': 0.8671381944444445,
'cpu_util%': 50.0,
'cpu_util(AEP)%': 100.0,
'mem_util%': 40.26845637583892,
'mem_util(AEP)%': 44.0,
'scheduled': 4,
'utilization%': 45
}