Fix session count, concurrency, improve summary

crawler.py

@@ -120,6 +120,7 @@ if __name__ == "__main__":
 				"1.3": 0,
 			}
 		}
+		combinations = {}
 		f = open(sys.argv[2], "r")
 		c = json.load(f)
 		for domain in c:
@@ -141,51 +142,69 @@ if __name__ == "__main__":
 					pass
 					#print("Not found:", line, r)
 
+			cert = ""
 			if "cert_sig" not in domain_stats:
-				summary["cert"]["none"] += 1
+				cert = "none"
 			elif domain_stats["cert_sig"] == "ecdsa_secp256r1_sha256":
-				summary["cert"]["secp256r1"] += 1
+				cert = "secp256r1"
 			elif domain_stats["cert_sig"] == "ecdsa_secp384r1_sha384":
-				summary["cert"]["secp384r1"] += 1
+				cert = "secp384r1"
 			elif domain_stats["cert_sig"] == "ecdsa_secp521r1_sha512":
-				summary["cert"]["secp521r1"] += 1
+				cert = "secp521r1"
 			elif "rsa" in domain_stats["cert_sig"]:
-				summary["cert"]["rsa{}".format(domain_stats["cert_pk_size"])] += 1
+				cert = "rsa{}".format(domain_stats["cert_pk_size"])
+			if cert != "":
+				summary["cert"][cert] += 1
 
+			cipher = ""
 			if "cipher" not in domain_stats:
-				summary["cipher"]["none"] += 1
+				cipher = "none"
 			elif "AES_128" in domain_stats["cipher"] or "AES128" in domain_stats["cipher"]:
-				summary["cipher"]["aes128"] += 1
+				cipher = "aes128"
 			elif "AES_256" in domain_stats["cipher"] or "AES256" in domain_stats["cipher"]:
-				summary["cipher"]["aes256"] += 1
+				cipher = "aes256"
 			elif "CHACHA20" in domain_stats["cipher"]:
-				summary["cipher"]["chacha20"] += 1
+				cipher = "chacha20"
+			if cipher != "":
+				summary["cipher"][cipher] += 1
 
+			kx = ""
 			if "kx" not in domain_stats:
-				summary["kx"]["none"] += 1
+				kx = "none"
 			elif domain_stats["kx"] == "X25519MLKEM768":
-				summary["kx"]["x25519mlkem768"] += 1
+				kx = "x25519mlkem768"
 			elif domain_stats["kx"] == "X25519, 253 bits":
-				summary["kx"]["x25519"] += 1
+				kx = "x25519"
 			elif domain_stats["kx"] == "DH, 2048 bits":
-				summary["kx"]["rsa2048"] += 1
+				kx = "rsa2048"
 			elif domain_stats["kx"] == "DH, 3072 bits":
-				summary["kx"]["rsa3072"] += 1
+				kx = "rsa3072"
 			elif domain_stats["kx"] == "DH, 4096 bits":
-				summary["kx"]["rsa4096"] += 1
+				kx = "rsa4096"
 			elif domain_stats["kx"] == "ECDH, prime256v1, 256 bits":
-				summary["kx"]["secp256r1"] += 1
+				kx = "secp256r1"
 			elif domain_stats["kx"] == "ECDH, secp384r1, 384 bits":
-				summary["kx"]["secp384r1"] += 1
+				kx = "secp384r1"
 			elif domain_stats["kx"] == "ECDH, secp521r1, 521 bits":
-				summary["kx"]["secp521r1"] += 1
+				kx = "secp521r1"
+			if kx != "":
+				summary["kx"][kx] += 1
 
+			protocol = ""
 			if "protocol" not in domain_stats:
-				summary["version"]["none"] += 1
+				protocol = "none"
 			elif domain_stats["protocol"] == "TLSv1.3":
-				summary["version"]["1.3"] += 1
+				protocol = "1.3"
 			elif domain_stats["protocol"] == "TLSv1.2":
-				summary["version"]["1.2"] += 1
+				protocol = "1.2"
+			if protocol != "":
+				summary["version"][protocol] += 1
+			
+			combination = (protocol, cert, kx, cipher)
+			if combination in combinations:
+				combinations[combination] += 1
+			else:
+				combinations[combination] = 1
 
 			#if "kx" in domain_stats and domain_stats["kx"] == "ECDH":
 			#	print(c[domain])
@@ -195,10 +214,15 @@ if __name__ == "__main__":
 				print(f"{cat}:")
 				for item in stats[cat]:
 					print("  {}:\t{}".format(item, stats[cat][item]))
-		elif "-s" in sys.argv: # summary
+		if "-s" in sys.argv: # summary
 			for cat in summary:
 				print(f"{cat}:")
 				for item in summary[cat]:
 					print("  {}:\t{}".format(item, summary[cat][item]))
+		if "-c" in sys.argv: # combinations
+			combinations_list = [c for c in combinations]
+			combinations_list.sort(key=lambda c: combinations[c])
+			for c in combinations_list:
+				print("{}\t{}".format(c, combinations[c]))
 		else:
 			print(stats)

exp.py

@@ -87,7 +87,7 @@ CONFIGS = {
 			True,
 		],
 		"records": [
-			#{ "filename": "wikipedia", "repeat": 10000, "time": 90 },
+			{ "filename": "wp2", "repeat": 10000, "time": 600 },
 			{ "filename": "yt2-ads", "repeat": 10000, "time": 600 },
 			{ "filename": "yt2-ublock", "repeat": 10000, "time": 600 },
 		],
@@ -107,6 +107,7 @@ CONFIGS = {
 		"idle": "idle - - - - - - - - - - - 600.000081539154 0.0 896 4792 0.5399999999999991 0 -",
 		"notify_listen": ("0.0.0.0", 8090),
 		"notify_addr": "192.168.3.1:8090",
+		"concurrency": "4",
 	},
 	# i7-4790 -> core2
 	"core2": {
@@ -189,11 +190,13 @@ CONFIGS = {
 			"server",
 		],
 		"tls": [
-			False,
+			#False,
 			True,
 		],
 		"records": [
-			{ "filename": "wikipedia", "repeat": 1000 },
+			{ "filename": "wp2", "repeat": 10000, "time": 600 },
+			{ "filename": "yt2-ads", "repeat": 10000, "time": 600 },
+			{ "filename": "yt2-ublock", "repeat": 10000, "time": 600 },
 		],
 		"repo_dir": "/home/tuxmain/reps/tlsbench",
 		"exp_dir": "/dev/shm/exp",
@@ -281,7 +284,7 @@ CONFIGS = {
 		"rapl": False,
 		"sa": False,
 		"listen_port": 8080,
-		"idle": "idle - - - - - - - - - 1772205368.593937 1772206568.6941307 1200.1001937389374 0 298843 2217803 5.53333333333333 0 -",
+		"idle": "idle - - - - - - - - - 1774429500.8909554 1774430100.9141996 600.0232441425323 0 222346 1105446 8.666666666666666 0 -",
 		"notify_listen": ("0.0.0.0", 8090),
 		"notify_addr": "TODO:8090",
 		"ld_preload": {
@@ -343,12 +346,12 @@ CERT_SIGN_ALGS = [
 	"rsa2048", "rsa3072", "rsa4096", # widely used
 ]
 IMPLS = [
-	"aws-lc", # Amazon's crypto widely used in Rust stuff
+	#"aws-lc", # Amazon's crypto widely used in Rust stuff
-	"boring", # Google's fork of OpenSSL used in Chrome and Android
+	#"boring", # Google's fork of OpenSSL used in Chrome and Android
 	#"graviola", # New crypto in Rust
 	"openssl", # widely used
 	#"openssl-static",
-	"ring", # used in most Rust stuff
+	#"ring", # used in most Rust stuff
 	#"symcrypt", # Microsoft's crypto
 	#"wolfcrypt" # used in embedded (won't build with rpxy for now)
 ]
@@ -637,7 +640,10 @@ def run_netreplay_server(ssh, exp_dir, repo_dir, record, listen_addr, listen_por
 	#print(cmdline)
 	return ssh_run_bg(ssh, cmdline, env)
 
-def run_netreplay_client(ssh, exp_dir, repo_dir, record, remote_addr, remote_port, tls, impl, certs_dir, only_record=None, ciphers=None, kexes=None, earlydata="0", debug=False, notify_addr=None, ld_preload=None, skip_verif=False):
+def run_netreplay_client(
+	ssh, exp_dir, repo_dir, record, remote_addr, remote_port, tls, impl, certs_dir,
+	only_record=None, ciphers=None, kexes=None, earlydata="0", debug=False, notify_addr=None, ld_preload=None, skip_verif=False, concurrency=None
+):
 	if exp_dir[-1] != "/":
 		exp_dir += "/"
 	repo_dir = repo_dir.removesuffix("/")
@@ -661,6 +667,8 @@ def run_netreplay_client(ssh, exp_dir, repo_dir, record, remote_addr, remote_por
 		cmd += ["-n", notify_addr]
 	if skip_verif:
 		cmd.append("-s")
+	if concurrency != None:
+		cmd += ["--concurrency", concurrency]
 	cmdline = " ".join(cmd)
 	#print(cmdline)
 	return ssh_run_bg(ssh, cmdline, env)
@@ -876,6 +884,7 @@ def run_exp(config, only_record=None, idle=False, shutdown=False, debug=False):
 												debug=debug,
 												notify_addr=config["notify_addr"],
 												ld_preload=config["ld_preload"] if "ld_preload" in config else None,
+												concurrency=config.get("concurrency"),
 											)
 										else:
 											netreplay = run_netreplay_server(
@@ -912,6 +921,7 @@ def run_exp(config, only_record=None, idle=False, shutdown=False, debug=False):
 												notify_addr=config["notify_addr"],
 												ld_preload=config["ld_preload"] if "ld_preload" in config else None,
 												skip_verif=True,
+												concurrency=config.get("concurrency"),
 											)
 
 										prof_filename = "-"
@@ -942,6 +952,8 @@ def run_exp(config, only_record=None, idle=False, shutdown=False, debug=False):
 												if rec_count == 0xffffffff:
 													break
 												if rec_count > session_count:
+													if rec_count - session_count > 1:
+														print("count", session_count, "->", rec_count)
 													session_count = rec_count
 										except Timeout:
 											print("TIMEOUT: stop")
@@ -962,6 +974,14 @@ def run_exp(config, only_record=None, idle=False, shutdown=False, debug=False):
 											ssh_run(ssh, "killall netreplay-"+impl)
 										except invoke.exceptions.UnexpectedExit as e:
 											pass
+
+										# flush the UDP queue
+										signal.alarm(2)
+										try:
+											while True:
+												notify_socket.recv(4)
+										except Timeout:
+											pass
 										
 										# Measure CPU after (as it may update only after the process is killed)
 										new_cpu = 0

makecerts.py

@@ -15,7 +15,8 @@ ALGS = [
 	"rsa4096"
 ]
 DOMAINS = [
-	#"wikipedia.org",
+	"wikipedia.org",
+	"wikimedia.org",
 	"youtube.com",
 	"rr2---sn-gxo5uxg-jqbl.googlevideo.com",# main domain googlevideo.com returns certificate for google.com only
 	"googleusercontent.com",

plots.py

@@ -72,6 +72,10 @@ def impl_title(impl):
 # Where gnuplot files, data files and images are output
 PLOTS_DIR = "/dev/shm/plots"
 
+# Energy per byte on the network, in W/s
+# Estimated at 0.06 kWh/GB by Aslan 2018
+WS_PER_BYTE = 0.06 * 1000 * 3600 / 1024**3
+
 def gnuplot_histogram(**kwargs):
 	if "machine" in kwargs and kwargs["machine"] != None:
 		kwargs["machine"] = ", " + kwargs["machine"]
@@ -525,6 +529,7 @@ def make_summary(logs):
 			n = float(log.get("n", "1000"))
 			plain_results[plain_result_key(log)] = {
 				"cpu": (float(log["cpu"]) - idle_val["cpu"] * float(log["time"])) / n,
+				"total_energy": float(log["Wh"]) * 3600 / n,
 				"energy": (float(log["Wh"]) * 3600 - idle_val["energy"] * float(log["time"])) / n,
 				"in": (float(log["bytes_in"]) - idle_val["in"] * float(log["time"])) / n,
 				"out": (float(log["bytes_out"]) - idle_val["out"] * float(log["time"])) / n,
@@ -533,23 +538,38 @@ def make_summary(logs):
 			n = float(log.get("n", "1000"))
 			results[result_key(log)] = {
 				"cpu": (float(log["cpu"]) - idle_val["cpu"] * float(log["time"])) / n,
+				"total_energy": float(log["Wh"]) * 3600 / n,
 				"energy": (float(log["Wh"]) * 3600 - idle_val["energy"] * float(log["time"])) / n,
 				"in": (float(log["bytes_in"]) - idle_val["in"] * float(log["time"])) / n,
 				"out": (float(log["bytes_out"]) - idle_val["out"] * float(log["time"])) / n,
 			}
 	
-	lines = [["key", "idle (W)", "no_tls (Ws/S)", "tls (Ws/S)", "tls_only (Ws/S)", "tls_in (1)", "tls_out (1)"]]
+	lines = [[
+		"key",
+		"idle (W)",
+		"no_tls (Ws/S)",
+		"tls (Ws/S)",
+		"tls_min (Ws/S)",
+		#"tls_max (Ws/S)",
+		"tls_in (MB/S)",
+		"tls_out (MB/S)",
+		"tls_io (MB/S)",
+		"tls_net (Ws/S)",
+	]]
 	for k in results:
 		r = results[k]
 		p = plain_results[k[:3]]
 		lines.append([
 			"/".join([str(i) for i in k]),
-			str(idle_val["energy"]),
+			str(round(idle_val["energy"], 1)),
-			str(p["energy"]),
+			str(round(p["energy"], 1)),
-			str(r["energy"]),
+			str(round(r["energy"], 1)),
-			str(r["energy"] - p["energy"]),
+			str(round(r["energy"] - p["energy"], 1)) + " (" + str(round((r["energy"] - p["energy"])/r["energy"]*100, 1)) + "%)",
-			str((r["in"] - p["in"]) / r["in"]),
+			#str(round(r["total_energy"] - p["total_energy"], 1)),
-			str((r["out"] - p["out"]) / r["out"]),
+			str(round((r["in"] - p["in"])/1024**2, 2)) + " (" + str(round((r["in"] - p["in"])/r["in"]*100, 1)) + "%)",
+			str(round((r["out"] - p["out"])/1024**2, 2)) + " (" + str(round((r["out"] - p["out"])/r["out"]*100, 1)) + "%)",
+			str(round((r["in"] + r["out"] - p["in"] - p["out"])/1024**2, 2)) + " (" + str(round((r["in"] + r["out"] - p["in"] - p["out"])/(r["in"] + r["out"])*100, 1)) + "%)",
+			str(round(r["energy"] - p["energy"] + (r["out"] + r["in"] - p["out"] - p["in"]) * WS_PER_BYTE, 2)),
 		])
 	print(tabulate(lines))