Forge test for testing throughput on network purely tuned for throughput

config/src/config/consensus_config.rs

@@ -102,8 +102,8 @@ impl Default for ConsensusConfig {
             max_sending_block_bytes: 600 * 1024, // 600 KB
             max_sending_block_bytes_quorum_store_override: 5 * 1024 * 1024, // 5MB
             max_receiving_block_txns: 10000,
-            max_receiving_block_txns_quorum_store_override: 2
-                * MAX_SENDING_BLOCK_TXNS_QUORUM_STORE_OVERRIDE,
+            max_receiving_block_txns_quorum_store_override: 10000.max(
+                2 * MAX_SENDING_BLOCK_TXNS_QUORUM_STORE_OVERRIDE),
             max_receiving_block_bytes: 3 * 1024 * 1024, // 3MB
             max_receiving_block_bytes_quorum_store_override: 6 * 1024 * 1024, // 6MB
             max_pruned_blocks_in_mem: 100,

testsuite/forge-cli/src/main.rs

@@ -3,7 +3,7 @@
 // SPDX-License-Identifier: Apache-2.0
 
 use anyhow::{format_err, Context, Result};
-use aptos_config::config::ConsensusConfig;
+use aptos_config::config::{ConsensusConfig, PipelineBackpressureValues, ChainHealthBackoffValues};
 use aptos_forge::{
     args::TransactionTypeArg,
     success_criteria::{LatencyType, StateProgressThreshold, SuccessCriteria},
@@ -232,9 +232,9 @@ fn main() -> Result<()> {
     let suite_name: &str = args.suite.as_ref();
 
     let suite_name = if suite_name == "land_blocking" {
-        "realistic_env_load_sweep"
+        "realistic_network_tuned_for_throughput"
     } else {
-        suite_name
+        panic!()
     };
 
     let runtime = Runtime::new()?;
@@ -494,11 +494,12 @@ fn single_test_suite(test_name: &str, duration: Duration) -> Result<ForgeConfig>
     let single_test_suite = match test_name {
         // Land-blocking tests to be run on every PR:
         "land_blocking" => land_blocking_test_suite(duration), // to remove land_blocking, superseeded by the below
-        "realistic_env_max_throughput" => realistic_env_max_throughput_test_suite(duration),
+        "realistic_env_max_throughput" => realistic_env_max_throughput_test(duration),
         "compat" => compat(),
         "framework_upgrade" => upgrade(),
         // Rest of the tests:
         "realistic_env_load_sweep" => realistic_env_load_sweep_test(),
+        "realistic_network_tuned_for_throughput" => realistic_network_tuned_for_throughput_test(),
         "epoch_changer_performance" => epoch_changer_performance(),
         "state_sync_perf_fullnodes_apply_outputs" => state_sync_perf_fullnodes_apply_outputs(),
         "state_sync_perf_fullnodes_execute_transactions" => {
@@ -1365,7 +1366,7 @@ fn land_blocking_test_suite(duration: Duration) -> ForgeConfig {
 }
 
 // TODO: Replace land_blocking when performance reaches on par with current land_blocking
-fn realistic_env_max_throughput_test_suite(duration: Duration) -> ForgeConfig {
+fn realistic_env_max_throughput_test(duration: Duration) -> ForgeConfig {
     ForgeConfig::default()
         .with_initial_validator_count(NonZeroUsize::new(20).unwrap())
         .with_initial_fullnode_count(10)
@@ -1418,6 +1419,45 @@ fn realistic_env_max_throughput_test_suite(duration: Duration) -> ForgeConfig {
         )
 }
 
+fn realistic_network_tuned_for_throughput_test() -> ForgeConfig {
+    ForgeConfig::default()
+        .with_initial_validator_count(NonZeroUsize::new(20).unwrap())
+        .with_initial_fullnode_count(10)
+        .add_network_test(MultiRegionNetworkEmulationTest {
+            override_config: None,
+        })
+        .with_node_helm_config_fn(Arc::new(move |helm_values| {
+            helm_values["validator"]["config"]["consensus"]["max_sending_block_txns_quorum_store_override"] =
+                10000.into();
+            helm_values["validator"]["config"]["consensus"]["pipeline_backpressure"] =
+                serde_yaml::to_value(Vec::<PipelineBackpressureValues>::new()).unwrap();
+            helm_values["validator"]["config"]["consensus"]["chain_health_backoff"] =
+                serde_yaml::to_value(Vec::<ChainHealthBackoffValues>::new()).unwrap();
+
+            helm_values["validator"]["config"]["consensus"]["quorum_store_configs"]["back_pressure"]["backlog_txn_limit_count"] =
+                100000.into();
+
+            // Experimental storage optimizations
+            helm_values["validator"]["config"]["storage"]["rocksdb_configs"]["use_state_kv_db"] = true.into();
+            helm_values["validator"]["config"]["storage"]["rocksdb_configs"]["use_sharded_state_merkle_db"] = true.into();
+        }))
+        .with_success_criteria(
+            SuccessCriteria::new(7000)
+            .add_no_restarts()
+            .add_wait_for_catchup_s(60)
+            .add_system_metrics_threshold(SystemMetricsThreshold::new(
+                // Check that we don't use more than 12 CPU cores for 30% of the time.
+                MetricsThreshold::new(12, 30),
+                // Check that we don't use more than 10 GB of memory for 30% of the time.
+                MetricsThreshold::new(10 * 1024 * 1024 * 1024, 30),
+            ))
+            .add_chain_progress(StateProgressThreshold {
+                max_no_progress_secs: 10.0,
+                max_round_gap: 4,
+            }),
+        )
+}
+
 fn pre_release_suite() -> ForgeConfig {
     ForgeConfig::default()
         .with_initial_validator_count(NonZeroUsize::new(30).unwrap())