Add support for measuring tx e2e latency

scripts/measure_e2e.sh

@@ -0,0 +1,53 @@
+#!/bin/sh
+
+# SLP transaction end-to-end latency measurement wrapper.
+#
+# Usage: sh scripts/measure_e2e.sh [SLP_EVAL_OPTION...] [-- MISSION_ARG...]
+#
+# A preset over scripts/slp_eval.sh that sets the options needed to measure
+# tx end-to-end latency.
+
+SLP_EVAL="$(dirname "$0")/slp_eval.sh"
+
+NETWORK_SIZE_LIMIT=1000
+PUBNET_DATA_FILE="public-network-data-2026-06-03-trimmed-located.json"
+LOADGEN_KEYS_FILE="public-network-data-2026-06-03-loadgenkeys.json"
+
+# Sets DATA_ROOT and SEPARATOR_GIVEN (whether the user supplied "--" in the
+# arguments). Shifts its own copy of the arguments, leaving the caller's "$@"
+# intact.
+scan_args() {
+	DATA_ROOT="$(pwd)"
+	SEPARATOR_GIVEN=0
+
+	while [ "$#" -gt 0 ]; do
+		case "$1" in
+		--)
+			SEPARATOR_GIVEN=1
+			return
+			;;
+		--data-root | --supercluster-root)
+			DATA_ROOT="$2"
+			;;
+		--data-root=* | --supercluster-root=*)
+			DATA_ROOT="${1#*=}"
+			;;
+		esac
+
+		shift
+	done
+}
+
+scan_args "$@"
+
+# Add a separator if the caller didn't supply one
+if [ "$SEPARATOR_GIVEN" -eq 0 ]; then
+	set -- "$@" --
+fi
+
+exec sh "$SLP_EVAL" \
+	--network-size-limit "$NETWORK_SIZE_LIMIT" \
+	--pubnet-data "$DATA_ROOT/data/$PUBNET_DATA_FILE" \
+	"$@" \
+	--loadgen-keys "$DATA_ROOT/data/$LOADGEN_KEYS_FILE" \
+	--measure-e2e-latency

scripts/slp_eval.sh

@@ -3,9 +3,10 @@
 # SLP mixed-load evaluation wrapper.
 #
 # This script runs the MinBlockTimeMixed mission against a stellar-core image
-# using the 2025-06-24 pubnet topology data and the fixed benchmark parameters
-# below. It is intended to answer: "does this image sustain the selected mixed
-# classic/Soroban load at the normal 5s ledger close target?"
+# using the 2025-06-24 pubnet topology data (by default; see --pubnet-data) and
+# the fixed benchmark parameters below. It is intended to answer: "does this
+# image sustain the selected mixed classic/Soroban load at the normal 5s ledger
+# close target?"
 #
 # Benchmark setup:
 # - One mission run is started for each selected Soroban load flag:
@@ -15,14 +16,18 @@
 #   match current network conditions. The flag value supplies only the Soroban
 #   TPS for that run, so total TPS is CLASSIC_TX_RATE + selected Soroban TPS.
 # - The mission uses MinBlockTimeMixed's MIXED_PREGEN_* overlay-only loadgen
-#   mode, simulated pubnet network delay, with NETWORK_SIZE_LIMIT nodes.
+#   mode, simulated pubnet network delay, with a configurable number of nodes
+#   (--network-size-limit, default 277).
 # - The block-time search range is intentionally narrow:
 #   [MIN_BLOCK_TIME_MS, MAX_BLOCK_TIME_MS] = [4900, 5100]. With the mission's
 #   100ms binary-search threshold, this effectively evaluates the 5s target
 #   to match the current network close time.
 # - simulate-apply-duration is derived from SIMULATE_APPLY_BUDGET_MS and the
 #   total TPS so the synthetic apply sleep budget remains roughly constant as
 #   the requested Soroban rate changes.
+# - Arguments after a trailing "--" are appended verbatim to every mission
+#   command line, so one-off flags can be added without editing this script.
+#   scripts/measure_e2e.sh is a preset built on top of that mechanism.
 #
 # Result interpretation:
 # - A zero exit and a "Minimum sustainable block time: ..." log line means the
@@ -41,6 +46,7 @@ STELLAR_CORE_IMAGE=
 SAC_TX_RATE=
 OZ_TX_RATE=
 SOROSWAP_TX_RATE=
+PUBNET_DATA=
 
 IMAGE_REPOSITORY="746476062914.dkr.ecr.us-east-1.amazonaws.com/dev"
 
@@ -64,11 +70,15 @@ NUM_PREGENERATED_TXS=1000000
 GENESIS_TEST_ACCOUNT_COUNT=1000000
 SIMULATE_APPLY_WEIGHT=100
 SIMULATE_APPLY_BUDGET_MS=600
-NETWORK_SIZE_LIMIT=277
+DEFAULT_NETWORK_SIZE_LIMIT=277
+NETWORK_SIZE_LIMIT="$DEFAULT_NETWORK_SIZE_LIMIT"
+DEFAULT_PUBNET_DATA_FILE="public-network-data-2025-06-24.json"
 
 usage() {
 	cat <<EOF
-Usage: $0 --stellar-core-image IMAGE [--data-root PATH] [--sac RATE] [--oz RATE] [--soroswap RATE]
+Usage: $0 --stellar-core-image IMAGE [--data-root PATH] [--network-size-limit N]
+       [--pubnet-data PATH] [--sac RATE] [--oz RATE] [--soroswap RATE]
+       [-- MISSION_ARG...]
 
 Runs one MinBlockTimeMixed mission per selected load flag against IMAGE.
 Each run always generates ${CLASSIC_TX_RATE} classic payment TPS plus the
@@ -78,19 +88,24 @@ TPS + 50 SAC Soroban TPS; "--sac 50 --oz 25" runs two separate benchmarks.
 Options:
   --stellar-core-image, --image IMAGE   Stellar Core image to evaluate. Required.
   --data-root, --supercluster-root PATH Root containing data/. Defaults to pwd.
+  --network-size-limit N                Number of pubnet nodes to run.
+                                        Defaults to ${DEFAULT_NETWORK_SIZE_LIMIT}.
+  --pubnet-data PATH                    Pubnet connectivity graph to run against.
+                                        Defaults to <data-root>/data/${DEFAULT_PUBNET_DATA_FILE}.
   --sac RATE                            Run SAC load with the given Soroban tx rate.
                                         Can be supplied with other load flags to run benchmarks sequentially.
   --oz RATE                             Run OZ load with the given Soroban tx rate.
                                         Can be supplied with other load flags to run benchmarks sequentially.
   --soroswap RATE                       Run Soroswap load with the given Soroban tx rate.
                                         Can be supplied with other load flags to run benchmarks sequentially.
   -h, --help                            Show this help.
+  -- MISSION_ARG...                     Everything after "--" is appended verbatim to each
+                                        mission command line. Use it for one-off flags this
+                                        wrapper does not expose.
 
 Benchmark constants:
   Classic TPS:        ${CLASSIC_TX_RATE}
   Target close time:  ${BLOCK_TIME_MS}ms, evaluated via [${MIN_BLOCK_TIME_MS}, ${MAX_BLOCK_TIME_MS}]
-  Network size limit: ${NETWORK_SIZE_LIMIT}
-  Data set:           data/public-network-data-2025-06-24.json
 
 Results:
   PASS: command exits 0 and logs "Minimum sustainable block time: ...".
@@ -112,7 +127,13 @@ is_nonnegative_integer() {
 	esac
 }
 
+# Parses the wrapper's own options and stops at a "--" separator. Sets
+# PARSED_ARG_COUNT to the number of arguments consumed (including the
+# separator) so the caller can shift them off and keep the mission arguments
+# that followed in "$@".
 parse_args() {
+	total_arg_count="$#"
+
 	while [ "$#" -gt 0 ]; do
 		case "$1" in
 		--stellar-core-image | --image)
@@ -180,6 +201,43 @@ parse_args() {
 			SOROSWAP_TX_RATE="${1#*=}"
 			shift
 			;;
+		--network-size-limit)
+			if [ "$#" -lt 2 ] || [ -z "$2" ]; then
+				printf '%s\n' "--network-size-limit requires a node count." >&2
+				usage >&2
+				exit 1
+			fi
+			NETWORK_SIZE_LIMIT="$2"
+			shift 2
+			;;
+		--network-size-limit=*)
+			NETWORK_SIZE_LIMIT="${1#*=}"
+			shift
+			;;
+		--pubnet-data)
+			if [ "$#" -lt 2 ] || [ -z "$2" ]; then
+				printf '%s\n' "--pubnet-data requires a path." >&2
+				usage >&2
+				exit 1
+			fi
+			PUBNET_DATA="$2"
+			shift 2
+			;;
+		--pubnet-data=*)
+			PUBNET_DATA="${1#*=}"
+			# An empty value here would silently fall back to the
+			# default data set, so reject it like the two-token form.
+			if [ -z "$PUBNET_DATA" ]; then
+				printf '%s\n' "--pubnet-data requires a path." >&2
+				usage >&2
+				exit 1
+			fi
+			shift
+			;;
+		--)
+			shift
+			break
+			;;
 		-h | --help)
 			usage
 			exit 0
@@ -191,6 +249,8 @@ parse_args() {
 			;;
 		esac
 	done
+
+	PARSED_ARG_COUNT=$((total_arg_count - $#))
 }
 
 validate_tx_rate() {
@@ -227,6 +287,13 @@ validate_args() {
 	if [ -n "$SOROSWAP_TX_RATE" ]; then
 		validate_tx_rate "--soroswap" "$SOROSWAP_TX_RATE"
 	fi
+
+	# Non-numeric values make "[" exit 2 rather than 1, so test for the
+	# accepted range and negate: anything unparseable is rejected too.
+	if ! [ "$NETWORK_SIZE_LIMIT" -ge 1 ] 2>/dev/null; then
+		printf '%s\n' "--network-size-limit must be a positive integer." >&2
+		exit 1
+	fi
 }
 
 calculate_simulate_apply_duration() {
@@ -266,8 +333,9 @@ resolve_min_block_time_mixed_mode() {
 }
 
 run_min_block_time_mixed() {
-	mode_alias="${1:?usage: run_min_block_time_mixed MODE SOROBAN_TX_RATE}"
-	soroban_tx_rate="${2:?usage: run_min_block_time_mixed MODE SOROBAN_TX_RATE}"
+	mode_alias="${1:?usage: run_min_block_time_mixed MODE SOROBAN_TX_RATE [MISSION_ARG...]}"
+	soroban_tx_rate="${2:?usage: run_min_block_time_mixed MODE SOROBAN_TX_RATE [MISSION_ARG...]}"
+	shift 2
 	min_block_time_mixed_mode="$(resolve_min_block_time_mixed_mode "$mode_alias")"
 	simulate_apply_duration="$(calculate_simulate_apply_duration "$CLASSIC_TX_RATE" "$soroban_tx_rate")"
 
@@ -296,23 +364,28 @@ run_min_block_time_mixed() {
 		--tier1-keys "$TIER1_KEYS" \
 		--network-size-limit "$NETWORK_SIZE_LIMIT" \
 		--require-node-labels=purpose:largetests \
-		--tolerate-node-taints=largetests
+		--tolerate-node-taints=largetests \
+		"$@"
 }
 
 parse_args "$@"
+shift "$PARSED_ARG_COUNT"
 validate_args
 
-PUBNET_DATA="$DATA_ROOT/data/public-network-data-2025-06-24.json"
+if [ -z "$PUBNET_DATA" ]; then
+	PUBNET_DATA="$DATA_ROOT/data/$DEFAULT_PUBNET_DATA_FILE"
+fi
+
 TIER1_KEYS="$DATA_ROOT/data/tier1keys.json"
 
 if [ -n "$SAC_TX_RATE" ]; then
-	run_min_block_time_mixed sac "$SAC_TX_RATE"
+	run_min_block_time_mixed sac "$SAC_TX_RATE" "$@"
 fi
 
 if [ -n "$OZ_TX_RATE" ]; then
-	run_min_block_time_mixed oz "$OZ_TX_RATE"
+	run_min_block_time_mixed oz "$OZ_TX_RATE" "$@"
 fi
 
 if [ -n "$SOROSWAP_TX_RATE" ]; then
-	run_min_block_time_mixed soroswap "$SOROSWAP_TX_RATE"
+	run_min_block_time_mixed soroswap "$SOROSWAP_TX_RATE" "$@"
 fi

src/App/Program.fs

@@ -73,8 +73,10 @@ type MissionOptions
         apiRateLimit: int,
         httpProxyReplicas: int,
         pubnetData: string option,
+        measureE2eLatency: bool,
         flatQuorum: bool option,
         tier1Keys: string option,
+        loadgenKeys: string option,
         maxConnections: int option,
         fullyConnectTier1: bool,
         byteCountValues: seq<int>,
@@ -318,12 +320,21 @@ type MissionOptions
     [<Option("pubnet-data", HelpText = "JSON file containing pubnet connectivity graph data", Required = false)>]
     member self.PubnetData = pubnetData
 
+    [<Option("measure-e2e-latency",
+             HelpText = "Set to enable the loadgen e2e metrics",
+             Required = false,
+             Default = false)>]
+    member self.MeasureE2eLatency = measureE2eLatency
+
     [<Option("flat-quorum", HelpText = "Use flat Tier1 quorum", Required = false)>]
     member self.FlatQuorum = flatQuorum
 
     [<Option("tier1-keys", HelpText = "JSON file containing list of 'tier-1' pubkeys from pubnet", Required = false)>]
     member self.Tier1Keys = tier1Keys
 
+    [<Option("loadgen-keys", HelpText = "JSON file containing list of pubkeys to generate load", Required = false)>]
+    member self.LoadgenKeys = loadgenKeys
+
     [<Option("max-connections",
              HelpText = "Maximum number of connections to allow any node in pubnet data to have. When enabled, this option will prune connections for any node with more than this number of connections. (default: no limit)",
              Required = false)>]
@@ -743,6 +754,12 @@ let main argv =
             0
 
         | :? MissionOptions as mission ->
+            if mission.LoadgenKeys.IsSome && mission.PubnetData.IsNone then
+                failwith "Error: --loadgen-keys requires --pubnet-data to be set"
+
+            if mission.MeasureE2eLatency && mission.LoadgenKeys.IsNone then
+                failwith "Error: --measure-e2e-latency requires --loadgen-keys"
+
             let _ = logToConsoleAndFile (sprintf "%s/stellar-supercluster.log" mission.Destination)
 
             let ll =
@@ -848,8 +865,10 @@ let main argv =
                                apiRateLimit = mission.ApiRateLimit
                                httpProxyReplicas = mission.HttpProxyReplicas
                                pubnetData = mission.PubnetData
+                               measureE2eLatency = mission.MeasureE2eLatency
                                flatQuorum = mission.FlatQuorum
                                tier1Keys = mission.Tier1Keys
+                               loadgenKeys = mission.LoadgenKeys
                                maxConnections = mission.MaxConnections
                                fullyConnectTier1 = mission.FullyConnectTier1
                                byteCountDistribution =

src/FSLibrary.Tests/Tests.fs

@@ -71,8 +71,10 @@ let ctx : MissionContext =
       apiRateLimit = 10
       httpProxyReplicas = 2
       pubnetData = None
+      measureE2eLatency = false
       flatQuorum = None
       tier1Keys = None
+      loadgenKeys = None
       maxConnections = None
       fullyConnectTier1 = false
       peerReadingCapacity = None

src/FSLibrary/MaxTPSTest.fs

@@ -165,9 +165,15 @@ let maxTPSTest (context: MissionContext) (baseLoadGen: LoadGen) (setupCfg: LoadG
         List.find (fun (cs: CoreSet) -> cs.name.StringName = "stellar" || cs.name.StringName = "sdf") allNodes
 
     let tier1 = List.filter (fun (cs: CoreSet) -> cs.options.tier1 = Some true) allNodes
+    let loadGenNodes = List.filter (fun (cs: CoreSet) -> cs.options.generatesLoad) allNodes
+
+    let loadGenNodes =
+        if List.isEmpty loadGenNodes then
+            // On smaller networks, run loadgen on all nodes to better balance the overhead of load generation
+            if List.length allNodes > smallNetworkSize then tier1 else allNodes
+        else
+            loadGenNodes
 
-    // On smaller networks, run loadgen on all nodes to better balance the overhead of load generation
-    let loadGenNodes = if List.length allNodes > smallNetworkSize then tier1 else allNodes
     let isLoadGenNode cs = List.exists (fun (cs': CoreSet) -> cs' = cs) loadGenNodes
 
     // Assign pre-generated transaction information to each load generator node.