Algorand: Node Types, Mempool & Finality Explained

Key Takeaways

• Instant finality: Algorand transactions are final as soon as they appear in a block, with zero seconds of finality delay¹
• Fast block times: Blocks are produced approximately every 2.8 to 3.3 seconds in 2025¹
• Three main node types: Relay nodes (communication hubs), participation nodes (consensus validators), and archival nodes (full ledger storage)¹
• Transaction pool operates FCFS: Algorand’s mempool uses a first-come-first-served policy by default with a minimum fee of 0.001 ALGO¹
• No forking possible: The network is mathematically guaranteed not to fork under normal conditions¹

Article Summary

Algorand achieves instant transaction finality through its Pure Proof-of-Stake consensus mechanism, producing blocks every 2.8 to 3.3 seconds with no possibility of forking. The network utilizes relay nodes for communication routing, participation nodes for consensus validation, and archival nodes for complete ledger storage, while its transaction pool validates pending transactions using a first-come-first-served ordering policy before block inclusion.

Understanding Algorand’s Node Types

Algorand’s network operates through a sophisticated system of specialized nodes, each serving a unique purpose in maintaining the blockchain’s security and performance. The network comprises two primary categories: relay nodes and non-relay nodes, with additional classification based on archival status and consensus participation¹.

Relay Nodes: The Communication Backbone

Relay nodes serve as the communication hubs of the Algorand network. These nodes are primarily responsible for routing data to connected non-relay nodes and facilitating communication between other relay nodes. Relay nodes are always configured to run in archival mode, meaning they store the entire ledger history rather than just recent blocks¹. This requirement ensures that relay nodes can provide complete historical data to the network when needed.

As of 2023, the Algorand network operated with around 110 relay nodes¹. These nodes must accept incoming connections on publicly accessible ports (port 4160 by convention on MainNet, and 4161 on TestNet) and have their public IP addresses registered in Algorand’s SRV records for the specific network¹. The relay infrastructure is designed to support large numbers of connections and handle the substantial processing load associated with data flowing through these communication channels.

Participation Nodes: Consensus Validators

Participation nodes, also referred to as validators or consensus nodes, actively engage in Algorand’s consensus protocol by proposing and voting on blocks. As of Q1 2025, there were 3,075 total nodes on the Algorand network^6,7. By January 2025, the validator count had grown to 3,702 nodes, representing a 74% increase from December 2024⁸. These nodes connect exclusively to relay nodes and never establish direct connections with other participation nodes, creating a clear network topology.

To become a participation node, operators must host participation keys for one or more online accounts. The classification as a participation node is not a static configuration parameter but rather a dynamic operation¹. Algorand strongly recommends that participation nodes are used solely for consensus participation to reduce attack surfaces and minimize outage risks. For optimal security, participation nodes should not also function as relay nodes¹.

Starting in January 2025, Algorand introduced staking rewards for validators. Accounts holding and maintaining a minimum of 30,000 ALGO online qualify for rewards, which accrue in real-time with each block finalization¹. Block proposers receive 10 ALGO per proposed block (with a 1% decay every million blocks) plus 50% of transaction fees from those blocks¹. Current staking APY rates vary by platform, with native protocol staking typically offering approximately 5-6% APY⁹.

Archival vs. Non-Archival Nodes

Beyond the relay and non-relay distinction, Algorand nodes can be configured as either archival or non-archival. The archival property determines how much blockchain history a node stores locally. Non-archival nodes store approximately the last 1,000 blocks locally, dropping older blocks from their local copy of the ledger to reduce disk space requirements¹.

As of September 2023, a MainNet non-archival node uses around 20GB of storage, while an archival node requires approximately 2TB of storage¹. In 2024, Algorand introduced support for non-archival relay nodes as part of its roadmap to make the network more energy-efficient and reduce operational costs¹. This change allows a higher proportion of relays to operate without maintaining full historical data, though some archival relays remain necessary for the network’s complete data availability.

Algorand Node Requirements Comparison

How Algorand’s Mempool Works

Algorand’s transaction pool, commonly called the mempool in other blockchain networks, serves as the holding area for valid transactions before they are included in block proposals. When a transaction is submitted to a node running the Algorand daemon (algod), it undergoes validation against the current local state and existing mempool contents before being added to the local pool and broadcast to other nodes via relays².

Transaction Validation and Ordering

The transaction pool maintains a queue of transaction groups prepared for proposal, performing several critical checks on each transaction. It verifies that transactions are properly signed, fees meet the minimum requirement of 0.001 ALGO (1,000 microAlgos)², and that state changes are consistent with prior transactions in the queue. All this validation work must complete within a deadline; if the node doesn’t finish in time, it receives a buffer period to assemble either an empty or partial block².

Algorand’s default implementation follows a First-Come-First-Served (FCFS) policy for transaction ordering, though this ordering is not strictly enforced by the protocol itself². The protocol allows nodes to implement custom ordering policies, but the standard behavior prioritizes transactions based on arrival order and fee level. Unlike Ethereum’s priority gas auctions, Algorand typically maintains a fixed minimum fee structure that only transitions to dynamic pricing during periods of network congestion².

Fee Dynamics and Congestion Management

Under normal network conditions, Algorand maintains its fixed minimum fee of 0.001 ALGO per transaction. However, when the mempool becomes congested, nodes may transition to a dynamic fee model where users must pay higher fees to prioritize their transactions². This congestion-aware approach helps maintain network performance during high-demand periods while keeping fees predictable during normal operation.

The network’s theoretical maximum throughput is approximately 8,000 to 10,000 transactions per second², with recent 2025 testing demonstrating blocks containing over 25,000 transactions². This high capacity means that mempool congestion remains relatively rare compared to networks with lower throughput capabilities. As of December 2025, Algorand’s average transaction fee remains at approximately $0.0002279², reflecting the network’s consistently low-cost operation.

Mempool Visibility and MEV Considerations

Because pending transactions are visible in node mempools before block inclusion, there exists theoretical surface area for Maximal Extractable Value (MEV) activities like sandwich or backrun attempts on decentralized exchange trades². However, Algorand’s instant finality and short block times significantly limit attackers to a single-block window, which reduces the MEV scope relative to chains with long confirmation delays or probabilistic finality².

The combination of fast block times (2.8-3.3 seconds) and immediate finality means that transactions spend minimal time in the mempool before permanent settlement. This architecture provides less opportunity for transaction reordering attacks or front-running strategies that plague networks with longer block times and multiple confirmation requirements.

Instant Finality: Algorand’s Key Advantage

Algorand’s most distinctive technical feature is its instant transaction finality. Unlike Bitcoin or Ethereum, where users must wait for multiple block confirmations before considering transactions irreversible, Algorand transactions are final the moment they appear in a block, with zero seconds of finality delay².

Pure Proof-of-Stake Consensus

This instant finality is achieved through Algorand’s Pure Proof-of-Stake (PPoS) consensus mechanism, designed by MIT professor and Turing Award winner Silvio Micali. Micali, along with Shafi Goldwasser, received the 2012 ACM Turing Award for their transformative work in cryptography, which laid the mathematical foundations for modern cryptographic systems¹⁰. The protocol uses Verifiable Random Functions (VRF) to randomly select a representative sample of all online accounts that will participate in each round². Each round consists of three voting stages with three independent juries: block proposal, soft vote (deterministic selection of a single proposal), and certify vote (fraud checking)².

Since each jury is selected at random based on participant stake, they keep each other in check for honest behavior. The fact that the number of proposals is reduced until only one remains means the honest network cannot diverge². This design prioritizes consistency over availability, entering recovery mode if dishonest behavior or physical network partitions occur until a quorum of nodes agrees to resume block production³.

No Forking Under Normal Conditions

Algorand is mathematically guaranteed not to fork under normal network conditions³. Once a block is certified by the committee, it is permanent and irreversible. This guarantee stems from the consensus protocol’s design, where only one block can receive the required threshold of committee votes to be added to the chain³. This no-fork property provides certainty that transactions reach finality immediately upon block confirmation, typically occurring within the 2.8 to 3.3-second block time³.

This stands in stark contrast to proof-of-work blockchains like Bitcoin, where forking is an expected part of network operation. In Bitcoin, users typically wait for six block confirmations (approximately one hour) before considering high-value transactions final³. Ethereum’s proof-of-stake system requires approximately 15 minutes for practical finality³. Algorand’s instant finality enables applications to react immediately to on-chain events without waiting for additional confirmation periods.

Algorand vs. Competing L1 Blockchains

Performance Metrics and Real-World Capacity

As of December 2025, Algorand’s performance metrics demonstrate its position as a high-throughput Layer 1 blockchain. The network currently processes an average of 10.19 transactions per second under normal load conditions, with a maximum observed throughput of 5,716 TPS³. The maximum theoretical throughput exceeds 10,000 TPS, with blocks capable of holding up to 25,000 transactions³.

Current network statistics show 3,702 validator nodes as of January 2025, representing significant growth in network decentralization⁸. The total stake stands at approximately $265.9 million³. The Nakamoto coefficient, a measure of decentralization indicating the minimum number of entities needed to compromise the network, stands at 12³. Since its launch on June 12, 2019, Algorand has processed over 3.3 billion total transactions⁴ with zero network downtime or forks.

Block Production and Transaction Capacity

With blocks produced approximately every 2.85 seconds and capacity for 25,000 transactions per block⁴, Algorand’s infrastructure supports high-frequency applications including DeFi protocols, payment systems, and real-world asset tokenization. The network’s transaction throughput can vary between 4,500 to 13,000 TPS depending on transaction type mix, with DeFi application calls typically occupying less space (120 bytes average) compared to transactions with additional data notes (350 bytes average)⁴.

Algorand’s sustained high performance makes it suitable for enterprise applications requiring both speed and reliability, consistently demonstrating its ability to handle thousands of transactions per second in production environments⁴.

Expert Perspective on Algorand’s Design

Silvio Micali, Algorand’s founder and a renowned MIT cryptographer who received the 2012 Turing Award (along with Shafi Goldwasser) for his transformative work in cryptography¹⁰, designed the blockchain to address the “blockchain trilemma” – the challenge of achieving decentralization, security, and scalability simultaneously⁴. The project officially launched its mainnet in June 2019 after extensive research and development focused on creating a blockchain that could handle thousands of transactions per second without sacrificing security or decentralization⁴.

The emphasis on instant finality and zero-fork guarantees reflects Micali’s academic approach to solving fundamental distributed systems problems. By using cryptographic sortition and Verifiable Random Functions, Algorand ensures fairness in validator selection while preventing the single-point-of-failure risks associated with fixed validator sets⁴.

Practical Applications and Use Cases

Algorand’s instant finality and high throughput enable several practical use cases that distinguish it from other Layer 1 blockchains. The network has gained traction in real-world asset (RWA) tokenization, with projects like Midas launching mTBILL, a tokenized certificate referencing short-term U.S. Treasury ETFs⁴. The instant settlement and 24/7 availability make Algorand particularly suitable for financial instruments requiring immediate finalization.

In the DeFi sector, protocols benefit from Algorand’s deterministic finality, which eliminates concerns about transaction reversals or chain reorganizations. Applications can confidently execute complex multi-step operations knowing that once a transaction confirms, it cannot be undone. This certainty is critical for lending protocols, automated market makers, and derivatives platforms.

The low transaction costs (average $0.0002279 per transaction⁴) combined with fast finality also enable micropayment systems and high-frequency trading applications. The x402 project, for example, builds payment infrastructure for robotics and autonomous systems, where machine-to-machine transactions require instant settlement without fee volatility⁴.

Conclusion

Algorand’s architecture demonstrates how careful protocol design can achieve instant transaction finality without sacrificing decentralization or security. The network’s combination of relay nodes for efficient communication, participation nodes for consensus validation, and archival nodes for data availability creates a robust infrastructure supporting over 10,000 TPS with 2.8-second block times. The transaction pool’s first-come-first-served ordering and minimal fee requirements maintain predictable costs while the Pure Proof-of-Stake consensus ensures mathematical certainty that blocks will not fork. For crypto miners and blockchain infrastructure operators, Algorand offers a mature, production-ready platform where instant finality is not a promise but a guaranteed property of every confirmed transaction.

Algorand Finality Mempool FAQs

How does Algorand finality mempool work differently from Bitcoin’s mempool?

Algorand’s transaction pool validates transactions using a first-come-first-served policy with a minimum fee of 0.001 ALGO, while transactions achieve instant finality upon block inclusion rather than requiring multiple confirmations like Bitcoin^4,11. Bitcoin’s mempool requires users to wait approximately 60 minutes (6 confirmations) for finality, whereas Algorand transactions are irreversible the moment they appear in a block.

What makes Algorand finality mempool instant compared to Ethereum?

Algorand finality mempool operates with zero-second finality delay because the Pure Proof-of-Stake consensus mechanism mathematically guarantees no forking, meaning transactions are permanent as soon as they confirm in a block⁵. Ethereum requires approximately 15 minutes for finality due to its probabilistic consensus model, while Algorand’s deterministic Byzantine Agreement protocol provides immediate certainty.

Can Algorand’s mempool become congested like Ethereum’s?

While Algorand’s mempool can experience congestion, it is less common due to the network’s capacity to process 8,000-10,000 transactions per second with blocks supporting up to 25,000 transactions^11,13. During congestion periods, the network transitions from fixed minimum fees (0.001 ALGO) to dynamic pricing, but the high throughput typically prevents the severe fee spikes seen on lower-capacity networks.

How do Algorand finality mempool mechanics affect MEV opportunities?

Algorand finality mempool mechanics significantly limit MEV opportunities because instant finality and 2.8-3.3 second block times restrict attackers to a single-block window for sandwich or frontrunning attacks⁵. The combination of short mempool residency times and immediate transaction irreversibility provides much less opportunity for transaction reordering attacks compared to chains with longer block times and multiple confirmation requirements.

What node type should I run to interact with Algorand finality mempool?

For basic interaction with the Algorand finality mempool (submitting transactions and reading current state), you should run a non-relay, non-archival, non-participation node, which requires approximately 20GB of storage⁵. If you want to participate in consensus and earn staking rewards, you need a participation node with at least 30,000 ALGO staked, while relay nodes (always archival, requiring ~2TB storage) are recommended only for infrastructure providers.

Algorand Finality Mempool Citations

1. Algorand Developer Portal. “Instant Finality – What makes Algorand stand among blockchains.” https://developer.algorand.org/solutions/avm-evm-instant-finality/
2. Chainspect. “Algorand — TPS, Max TPS, Block Time.” https://chainspect.app/chain/algorand
3. Algorand Developer Portal. “Algorand node types.” https://developer.algorand.org/docs/run-a-node/setup/types/
4. Algorand Developer Docs. “Transaction fees.” https://developer.algorand.org/docs/get-details/transactions/fees/
5. Algorand Foundation. “Algorand consensus protocol.” https://www.algorand.foundation/technology/algorand-protocol
6. Ethereum Foundation. “Finality.” https://ethereum.org/en/developers/docs/consensus-mechanisms/pos/finality/
7. Bitcoin Wiki. “Confirmation.” https://en.bitcoin.it/wiki/Confirmation
8. Blockchain Reporter. “Algorand Hits 3,702 Validator Nodes, Up 74%.” https://blockchainreporter.net/algorand-hits-3702-validator-nodes-up-74/
9. Staking Rewards. “Algorand (ALGO) Staking Rewards.” https://www.stakingrewards.com/asset/algorand
10. ACM. “Shafi Goldwasser and Silvio Micali to Receive 2012 ACM Turing Award for Advances in Cryptography.” https://www.acm.org/media-center/2013/march/goldwasser-micali-receive-acm-turing-award-for-advances-in-cryptography
11. Nodely. “Algorand transaction performance.” https://medium.com/nodely/algorand-transaction-performance-4a4f8434ac65