Add new mode to apply load for tx set validation

docs/apply-load-benchmark-sac.cfg

@@ -7,6 +7,11 @@
 APPLY_LOAD_MODE="benchmark"
 APPLY_LOAD_MODEL_TX="sac"
 
+# Which timing path to use: "apply" preserves the historical apply-only
+# benchmark, while "txset-validation-and-apply" simulates a non-leader receiving
+# and validating a tx set before applying it. Tx-set creation is not measured.
+APPLY_LOAD_TIMING_PHASES = "apply"
+
 # Whether to time the write part of the apply stage. This can be
 # disabled to get less noisy results for non-write related changes,
 # but should be enabled to get more comprehensive e2e numbers.
@@ -62,4 +67,4 @@ NODE_SEED="SDQVDISRYN2JXBS7ICL7QJAEKB3HWBJFP2QECXG7GZICAHBK4UNJCWK2 self"
 
 [QUORUM_SET]
 THRESHOLD_PERCENT=100
-VALIDATORS=["$self"]
+VALIDATORS=["$self"]

docs/metrics.md

@@ -62,6 +62,7 @@ crypto.verify.miss                        | meter     | number of signature cach
 crypto.verify.total                       | meter     | sum of both hits and misses
 crypto.verify.tx-valid-hit                | meter     | signature cache hits that occurred while validating transactions (outside of background signature validation)
 crypto.verify.tx-valid-total              | meter     | sum of both hits and misses during transaction validation (outside of background signature validation)
+herder.txset.validate                     | timer     | time spent turning a received tx set into an applicable tx set and validating it on a validity-cache miss; Tracy labels the enclosing call as a cache hit or miss
 herder.pending[-soroban]-txs.age0         | counter   | number of gen0 pending transactions
 herder.pending[-soroban]-txs.age1         | counter   | number of gen1 pending transactions
 herder.pending[-soroban]-txs.age2         | counter   | number of gen2 pending transactions

docs/software/commands.md

@@ -20,9 +20,11 @@ Common options can be placed at any place in the command line.
 Command options can only by placed after command.
 
 * **apply-load**: Benchmarks Soroban transaction application time using
-    synthetic transactions. The benchmark is isolated to mostly just executing
-    the transactions and thus it omits a lot of the supporting mechanisms
-    (such as overlay, SCP, mempool etc). This command will generate enough
+    synthetic transactions. The benchmark omits the overlay and mempool, but
+    each iteration reconstructs the tx set from serialized bytes and runs real
+    consensus with the node as its own single-validator quorum. It does not
+    simulate network transport, peer fetching, or multi-node timing.
+    This command will generate enough
     transactions to fill up a synthetic transaction queue (it's just a list of
     transactions with the same limits as the real queue), and then create a
     transaction set off of that to apply. This can also be used to record the
@@ -36,6 +38,20 @@ Command options can only by placed after command.
       consisting only of fast SAC transfer.
     - `APPLY_LOAD_MODE="benchmark"`: benchmarks a fixed-size ledger of model
       transactions. Use `APPLY_LOAD_MODEL_TX` to select the model transaction.
+  * `APPLY_LOAD_TIMING_PHASES` selects one of two timing paths:
+    - `"apply"`: the default historical apply-only benchmark. Its close helper
+      still calls `checkValid`, but that happens before the recorded ledger-close
+      timer and leaves the caches warm, as consensus validation would on a live
+      node. Output remains in the historical format.
+    - `"txset-validation-and-apply"`: simulates a non-leader receiving the tx
+      set over the wire, validating it through local consensus, and then applying
+      it. It reports validation, ledger close, and end-to-end time. Leader-side
+      tx-set creation and signing happen before the measured span. The signature
+      verification cache is cleared before validation, then retained so apply
+      sees the warm cache produced by validation.
+    `"txset-validation-and-apply"` is not supported with
+    `APPLY_LOAD_MODE="max-sac-tps"`; that search retains its historical
+    apply-only timing objective.
   * Load generation is configured in the Core config file. The relevant settings
     all begin with `APPLY_LOAD_`. See full example configurations with
     per-setting documentation in the `docs` directory

src/herder/HerderSCPDriver.cpp

@@ -76,6 +76,8 @@ HerderSCPDriver::SCPMetrics::SCPMetrics(Application& app)
           {"scp", "timing", "self-to-others-externalize-lag"}))
     , mBallotBlockedOnTxSet(app.getMetrics().NewTimer(
           {"scp", "timing", "ballot-blocked-on-txset"}))
+    , mTxSetValidation(
+          app.getMetrics().NewTimer({"herder", "txset", "validate"}))
     , mEmptyTxSetExternalized(
           app.getMetrics().NewCounter({"scp", "empty-tx-set", "externalized"}))
     , mEmptyTxSetValueReplaced(app.getMetrics().NewCounter(
@@ -1859,12 +1861,18 @@ HerderSCPDriver::checkAndCacheTxSetValid(TxSetXDRFrame const& txSet,
                                          LedgerHeaderHistoryEntry const& lcl,
                                          uint64_t closeTimeOffset) const
 {
+    ZoneScoped;
+
     auto key = TxSetValidityKey{lcl.hash, txSet.getContentsHash(),
                                 closeTimeOffset, closeTimeOffset};
 
     bool* pRes = mTxSetValidCache.maybeGet(key);
     if (pRes == nullptr)
     {
+        std::string zoneTxt("miss");
+        ZoneText(zoneTxt.c_str(), zoneTxt.size());
+        auto validationTime = mSCPMetrics.mTxSetValidation.TimeScope();
+
         // The invariant here is that we only validate tx sets nominated
         // to be applied to the current ledger state. However, in case
         // if we receive a bad SCP value for the current state, we still
@@ -1895,6 +1903,8 @@ HerderSCPDriver::checkAndCacheTxSetValid(TxSetXDRFrame const& txSet,
     }
     else
     {
+        std::string zoneTxt("hit");
+        ZoneText(zoneTxt.c_str(), zoneTxt.size());
         return *pRes;
     }
 }

src/herder/HerderSCPDriver.h

@@ -268,6 +268,9 @@ class HerderSCPDriver : public SCPDriver
         // download
         medida::Timer& mBallotBlockedOnTxSet;
 
+        // Timer tracking time to check and cache a tx set
+        medida::Timer& mTxSetValidation;
+
         // Tracks how many ledgers we externalized an empty-tx-set value.
         medida::Counter& mEmptyTxSetExternalized;
 

src/main/Config.cpp

@@ -464,6 +464,23 @@ parseApplyLoadModelTx(ConfigItem const& item)
         "invalid 'APPLY_LOAD_MODEL_TX', expected one of: sac, custom_token, "
         "soroswap");
 }
+
+ApplyLoadTimingPhases
+parseApplyLoadTimingPhases(ConfigItem const& item)
+{
+    auto phases = readString(item);
+    if (phases == "apply")
+    {
+        return ApplyLoadTimingPhases::APPLY_ONLY;
+    }
+    if (phases == "txset-validation-and-apply")
+    {
+        return ApplyLoadTimingPhases::TX_SET_VALIDATION_AND_APPLY;
+    }
+    throw std::invalid_argument(
+        "invalid 'APPLY_LOAD_TIMING_PHASES', expected one of: apply, "
+        "txset-validation-and-apply");
+}
 #endif
 
 template <typename T>
@@ -1866,6 +1883,11 @@ Config::processConfig(std::shared_ptr<cpptoml::table> t)
                  }},
                 {"APPLY_LOAD_TIME_WRITES",
                  [&]() { APPLY_LOAD_TIME_WRITES = readBool(item); }},
+                {"APPLY_LOAD_TIMING_PHASES",
+                 [&]() {
+                     APPLY_LOAD_TIMING_PHASES =
+                         parseApplyLoadTimingPhases(item);
+                 }},
 #endif // BUILD_TESTS
                 {"GENESIS_TEST_ACCOUNT_COUNT",
                  [&]() {

src/main/Config.h

@@ -83,6 +83,13 @@ enum class ApplyLoadModelTx
     CUSTOM_TOKEN,
     SOROSWAP
 };
+
+// Which apply-load timing path to use.
+enum class ApplyLoadTimingPhases
+{
+    APPLY_ONLY,
+    TX_SET_VALIDATION_AND_APPLY
+};
 #endif
 
 class Config : public std::enable_shared_from_this<Config>
@@ -425,6 +432,10 @@ class Config : public std::enable_shared_from_this<Config>
     // If set to true, database writes will count towards TPS calculation.
     // Otherwise, BucketList writes will not be counted.
     bool APPLY_LOAD_TIME_WRITES = true;
+
+    // Which apply-load timing path to use.
+    ApplyLoadTimingPhases APPLY_LOAD_TIMING_PHASES =
+        ApplyLoadTimingPhases::APPLY_ONLY;
 #endif // BUILD_TESTS
 
     // Waits for merges to complete before applying transactions during catchup