def experiment_changing_input_size(repetitions=100):
# Create a graph
fn = syn.generate_network_with(num_nodes=100000,
num_nodes_per_table=10,
num_schema_sim=90000,
num_content_sim=90000,
num_pkfk=90000)
api = API(fn)
perf_results = dict()
# input size from 1 to 100
for i in range(50):
i = i + 1
nodes = fn.fields_degree(i)
nids = [x for x, y in nodes]
info = fn.get_info_for(nids)
hits = fn.get_hits_from_info(info)
in_drs = api.drs_from_hits(hits)
q2, q3, q4 = run_all_queries(repetitions,
api_obj=api,
in_drs_obj=in_drs)
percentile_results = get_percentiles([q2, q3, q4])
perf_results[i] = percentile_results
return perf_results
def experiment_changing_max_hops_tc_queries(repetitions=100):
perf_results = dict()
for i in range(10):
i = i + 1
fn = syn.generate_network_with(num_nodes=100000,
num_nodes_per_table=10,
num_schema_sim=100000,
num_content_sim=100000,
num_pkfk=100000)
api = API(fn)
nodes = fn.fields_degree(1)
nids = [x for x, y in nodes]
info = fn.get_info_for(nids)
hits = fn.get_hits_from_info(info)
in_drs = api.drs_from_hits(hits)
query_times = []
for repet in range(repetitions):
s = time.time()
res = api.traverse(in_drs, Relation.SCHEMA_SIM, max_hops=i)
e = time.time()
query_times.append((e - s))
percentile_results = get_percentiles([query_times])
perf_results[i] = percentile_results
return perf_results
def test():
# Fixed graph density, differing sizes (nodes)
fn = syn.generate_network_with(num_nodes=100,
num_nodes_per_table=10,
num_schema_sim=200,
num_content_sim=150,
num_pkfk=50)
api = API(fn)
nodes = fn.fields_degree(3)
nids = [x for x, y in nodes]
info = fn.get_info_for(nids)
hits = fn.get_hits_from_info(info)
in_drs = api.drs_from_hits(hits)
q2, q3, q4 = run_all_queries(100, api_obj=api, in_drs_obj=in_drs)
nq2 = np.array(q2)
p5 = np.percentile(nq2, 5)
p50 = np.percentile(nq2, 50)
p95 = np.percentile(nq2, 95)
print("q2: " + str(p5) + " - " + str(p50) + " - " + str(p95))
nq3 = np.array(q3)
p5 = np.percentile(nq3, 5)
p50 = np.percentile(nq3, 50)
p95 = np.percentile(nq3, 95)
print("q3: " + str(p5) + " - " + str(p50) + " - " + str(p95))
nq4 = np.array(q4)
p5 = np.percentile(nq4, 5)
p50 = np.percentile(nq4, 50)
p95 = np.percentile(nq4, 95)
print("q4: " + str(p5) + " - " + str(p50) + " - " + str(p95))
def experiment_changing_graph_density_constant_size(repetitions=10):
size = 100000
densities = [100, 1000, 10000, 100000, 1000000]
perf_results = dict()
for density in densities:
fn = syn.generate_network_with(num_nodes=size,
num_nodes_per_table=10,
num_schema_sim=density,
num_content_sim=density,
num_pkfk=density)
api = API(fn)
nodes = fn.fields_degree(3)
nids = [x for x, y in nodes]
info = fn.get_info_for(nids)
hits = fn.get_hits_from_info(info)
in_drs = api.drs_from_hits(hits)
q2, q3, q4 = run_all_queries(repetitions,
api_obj=api,
in_drs_obj=in_drs)
percentile_results = get_percentiles([q2, q3, q4])
perf_results[density] = percentile_results
return perf_results