2025-02-06 17:50:10 +09:00

296 lines
11 KiB
Python

# !/usr/bin/env python
import argparse
import csv
import json
import os
import subprocess
from collections import OrderedDict
import latency
import mixlog
def bandwidth_result(log_path: str, step_duration_ms: int) -> dict[str, float]:
max_step_id = 0
for topic, json_msg in mixlog.get_input_stream(log_path):
if topic == "MessageFullyUnwrapped":
max_step_id = max(max_step_id, json_msg["message"]["step_id"])
with open(log_path, "r") as file:
for line in file:
if "total_outbound_bandwidth" in line:
line = line[line.find("{") :]
line = line.replace("{ ", '{"')
line = line.replace(": ", '": ')
line = line.replace(", ", ', "')
record = json.loads(line)
elapsed = (max_step_id * step_duration_ms) / 1000.0
return {
"min": float(record["min_node_total_bandwidth"]) / elapsed,
"avg": float(record["avg_node_total_bandwidth"]) / elapsed,
"max": float(record["max_node_total_bandwidth"]) / elapsed,
}
raise Exception("No bandwidth data found in log file")
def topology_result(log_path: str) -> dict[str, int]:
for topic, json_msg in mixlog.get_input_stream(log_path):
if topic == "Topology":
return json_msg
raise Exception("No topology found in log file")
def build_argument_parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(description="Log analysis for nomos-simulations.")
parser.add_argument(
"--step-duration",
type=int,
help="Duration (in ms) of each step in the simulation.",
)
parser.add_argument(
"--params-file",
type=str,
help="A CSV file that contains all parameter sets",
)
parser.add_argument(
"--orig-config-file",
type=str,
help="An original blendnet config JSON file that will be modified as specified in `params_file`",
)
parser.add_argument(
"--outdir",
type=str,
help="A output directory to be created",
)
parser.add_argument(
"--skip-run",
action="store_true",
help="Skip running the simulations and only analyze the logs",
)
return parser
if __name__ == "__main__":
argument_parser = build_argument_parser()
args = argument_parser.parse_args()
# Read the params CSV file
param_sets = []
param_set_header = []
with open(args.params_file, mode="r") as csvfile:
param_set_header = csvfile.readline().strip().split(",")
csvfile.seek(0) # Reset file pointer to the beginning after reading the header
reader = csv.DictReader(csvfile, delimiter=",")
param_sets = list(reader)
# Read the original blendnet json config file
with open(args.orig_config_file, "r") as jsonfile:
original_json = json.load(jsonfile)
# Directory to save modified JSON files
modified_configs_dir = os.path.join(args.outdir, "modified_configs")
os.makedirs(modified_configs_dir, exist_ok=True)
# Modify and save JSON files for each row in CSV
config_paths = []
for idx, param_set in enumerate(param_sets):
output_path = os.path.join(modified_configs_dir, f"{idx}.json")
config_paths.append(output_path)
if args.skip_run:
continue
modified_json = OrderedDict(original_json) # Preserve original field order
# Apply modifications
modified_json["network_settings"]["regions"]["north america west"] = 0.06
modified_json["network_settings"]["regions"]["north america east"] = 0.15
modified_json["network_settings"]["regions"]["north america central"] = 0.02
modified_json["network_settings"]["regions"]["europe"] = 0.47
modified_json["network_settings"]["regions"]["northern europe"] = 0.10
modified_json["network_settings"]["regions"]["east asia"] = 0.10
modified_json["network_settings"]["regions"]["southeast asia"] = 0.07
modified_json["network_settings"]["regions"]["australia"] = 0.03
modified_json["step_time"] = f"{args.step_duration}ms"
modified_json["node_count"] = int(param_set["network_size"])
modified_json["wards"][0]["sum"] = 1000
modified_json["connected_peers_count"] = int(param_set["peering_degree"])
modified_json["data_message_lottery_interval"] = "20s"
modified_json["stake_proportion"] = 0.0
modified_json["persistent_transmission"]["max_emission_frequency"] = 1.0
modified_json["persistent_transmission"]["drop_message_probability"] = 0.0
modified_json["epoch_duration"] = (
f"{int(param_set['cover_slots_per_epoch']) * int(param_set['cover_slot_duration'])}s"
)
modified_json["slots_per_epoch"] = int(param_set["cover_slots_per_epoch"])
modified_json["slot_duration"] = f"{param_set['cover_slot_duration']}s"
modified_json["max_delay_seconds"] = int(param_set["max_temporal_delay"])
modified_json["number_of_hops"] = int(param_set["blend_hops"])
modified_json["number_of_blend_layers"] = int(param_set["blend_hops"])
# Save modified JSON
with open(output_path, "w") as outfile:
json.dump(modified_json, outfile, indent=2)
print("Saved modified JSON to:", output_path)
# Directory to save logs
log_dir = os.path.join(args.outdir, "logs")
os.makedirs(log_dir, exist_ok=True)
log_paths = []
for idx, config_path in enumerate(config_paths):
log_path = f"{log_dir}/{idx}.log"
log_paths.append(log_path)
if args.skip_run:
continue
with open(log_path, "w") as log_file:
print(
f"Running simulation-{idx}: {log_file.name} with config: {config_path}"
)
subprocess.run(
["../../target/release/blendnet-sims", "--input-settings", config_path],
stdout=log_file,
)
print(f"Simulation-{idx} completed: {log_file.name}")
print("Analyzing logs...")
print("=================")
with open(os.path.join(args.outdir, "output.csv"), "w", newline="") as file:
print(f"Writing results to: {file.name}")
csv_writer = csv.writer(file)
csv_writer.writerow(
param_set_header
+ [
"network_diameter",
"msg_count",
"min_latency_sec",
"avg_latency_sec",
"median_latency_sec",
"max_latency_sec",
"min_latency_msg_id",
"min_latency_msg_persistent_latency_sec",
"min_latency_msg_persistent_index",
"min_latency_msg_temporal_latency_sec",
"min_latency_msg_temporal_index",
"max_latency_msg_id",
"max_latency_msg_persistent_latency_sec",
"max_latency_msg_persistent_index",
"max_latency_msg_temporal_latency_sec",
"max_latency_msg_temporal_index",
"min_conn_latency_sec",
"avg_conn_latency_sec",
"med_conn_latency_sec",
"max_conn_latency_sec",
"min_bandwidth_kbps",
"avg_bandwidth_kbps",
"max_bandwidth_kbps",
]
)
for idx, log_path in enumerate(log_paths):
csv_row = []
csv_row.extend([param_sets[idx][key] for key in param_set_header])
csv_row.append(topology_result(log_path)["diameter"])
latency_analysis = latency.LatencyAnalysis.build(
mixlog.get_input_stream(log_path)
)
csv_row.append(latency_analysis.total_messages)
csv_row.append(float(latency_analysis.min_latency_ms) / 1000.0)
csv_row.append(float(latency_analysis.avg_latency_ms) / 1000.0)
csv_row.append(float(latency_analysis.median_latency_ms) / 1000.0)
csv_row.append(float(latency_analysis.max_latency_ms) / 1000.0)
csv_row.append(latency_analysis.min_latency_analysis.message_id)
csv_row.append(
",".join(
map(
str,
[
ms / 1000.0
for ms in latency_analysis.min_latency_analysis.persistent_latencies_ms
],
)
)
)
csv_row.append(
",".join(
map(str, latency_analysis.min_latency_analysis.persistent_indices)
)
)
csv_row.append(
",".join(
map(
str,
[
ms / 1000.0
for ms in latency_analysis.min_latency_analysis.temporal_latencies_ms
],
)
)
)
csv_row.append(
",".join(
map(str, latency_analysis.min_latency_analysis.temporal_indices)
)
)
csv_row.append(latency_analysis.max_latency_analysis.message_id)
csv_row.append(
",".join(
map(
str,
[
ms / 1000.0
for ms in latency_analysis.max_latency_analysis.persistent_latencies_ms
],
)
)
)
csv_row.append(
",".join(
map(str, latency_analysis.max_latency_analysis.persistent_indices)
)
)
csv_row.append(
",".join(
map(
str,
[
ms / 1000.0
for ms in latency_analysis.max_latency_analysis.temporal_latencies_ms
],
)
)
)
csv_row.append(
",".join(
map(str, latency_analysis.max_latency_analysis.temporal_indices)
)
)
csv_row.append(
float(latency_analysis.conn_latency_analysis.min_ms) / 1000.0
)
csv_row.append(
float(latency_analysis.conn_latency_analysis.avg_ms) / 1000.0
)
csv_row.append(
float(latency_analysis.conn_latency_analysis.med_ms) / 1000.0
)
csv_row.append(
float(latency_analysis.conn_latency_analysis.max_ms) / 1000.0
)
bandwidth_res = bandwidth_result(log_path, args.step_duration)
csv_row.append(bandwidth_res["min"] * 8 / 1000.0)
csv_row.append(bandwidth_res["avg"] * 8 / 1000.0)
csv_row.append(bandwidth_res["max"] * 8 / 1000.0)
csv_writer.writerow(csv_row)
print(f"The outputs have been successfully written to {file.name}")