Skip to main content

Consensus & Throughput

RP1 achieves high throughput through optimized CometBFT consensus parameters and parallel transaction execution.

Consensus Parameters​

Default Configuration (Low Latency Network)​

// 200ms total block time
TimeoutPropose:    100ms  // Wait for block proposal
TimeoutPrevote:    50ms   // Prevote voting round
TimeoutPrecommit:  50ms   // Precommit voting round
SkipTimeoutCommit: true   // No delay between blocks

Total Block Time: ~200ms

Production Configuration (Geo-Distributed)​

// 600ms total block time for global validators
TimeoutPropose:    200ms
TimeoutPrevote:    100ms
TimeoutPrecommit:  100ms
TimeoutCommit:     200ms
SkipTimeoutCommit: false

Total Block Time: ~600ms

Throughput Analysis​

Theoretical Maximum​

ParameterValue
Block Time200ms
Blocks/Second5
Tx/Block (Target)20,000
Max TPS100,000

Realistic Throughput​

ScenarioTPSNotes
Simple transfers50,000+Bank module only
Privacy transactions10,000ZK proof verification overhead
DEX swaps30,000AMM calculations
Mixed workload25,000Typical production

Instant Finality​

Unlike probabilistic finality chains (Bitcoin, Ethereum PoW), RP1 provides instant finality:

Block N proposed → Prevotes (2/3+) → Precommits (2/3+) → Block N FINAL
                   └────────────────────────────────────┘
                            ~200-600ms total

No reorgs. No confirmations needed. Final is final.

Consensus Flow​

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  Proposer   │────▶│   Prevote   │────▶│  Precommit  │
│   (100ms)   │     │   (50ms)    │     │   (50ms)    │
└─────────────┘     └─────────────┘     └─────────────┘
       │                   │                   │
       ▼                   ▼                   ▼
   Block built      2/3+ prevotes       2/3+ precommits
   from mempool     received            received
                                              │
                                              ▼
                                    ┌─────────────────┐
                                    │  Block Committed│
                                    │    (FINAL)      │
                                    └─────────────────┘

Parallel Execution​

RP1 executes independent transactions in parallel:

// Transaction dependency analysis
tx1: A → B (1000 tokens)    // Group 1
tx2: C → D (500 tokens)     // Group 2 (independent, parallel)
tx3: B → E (200 tokens)     // Group 1 (depends on tx1)

// Execution order:
// [tx1, tx2] execute in parallel
// [tx3] executes after tx1 completes

Mempool Lanes​

Transactions are prioritized by type:

LanePriorityUse Case
OracleHighestPrice updates (protocol critical)
MEV-ProtectedHighEncrypted transactions
StandardMediumRegular transfers
Low PriorityLowBatch operations
// Lane configuration
type LaneConfig struct {
    MaxBlockSpace: 0.3,  // Max 30% of block per lane
    MaxTxs:        1000, // Max transactions per lane
}

Validator Requirements​

For optimal consensus performance:

ResourceMinimumRecommended
CPU8 cores16+ cores
RAM32 GB64 GB
StorageNVMe SSDNVMe RAID
Network1 Gbps10 Gbps
Latency to peers<100ms<50ms

Benchmarks​

Tested on c5.4xlarge (16 vCPU, 32GB RAM):

BenchmarkSendTransfer-16         50000    23000 ns/op
BenchmarkPrivacyShield-16        10000   150000 ns/op
BenchmarkDEXSwap-16              30000    35000 ns/op
BenchmarkOracleUpdate-16        100000     8000 ns/op

Configuration​

app.toml​

[mempool]
max_txs = 10000
max_tx_bytes = 1048576
cache_size = 100000

[state-sync]
snapshot_interval = 1000
snapshot_keep_recent = 2

config.toml​

[consensus]
timeout_propose = "100ms"
timeout_propose_delta = "10ms"
timeout_prevote = "50ms"
timeout_prevote_delta = "10ms"
timeout_precommit = "50ms"
timeout_precommit_delta = "10ms"
timeout_commit = "100ms"
skip_timeout_commit = true

Next Steps​