The Fast Flash Consensus Algorithm

v1.2 last update: 2/8/2023

HOME Investors White Paper Tech Spec Github

The Fast Flash Consensus Algorithm is a speedy consensus mechanism for blockchain transactions. Its innovative design allows every node to generate and validate blocks at the same time, leading to a high transaction per second(TPS) rate, thereby ensuring quick and secure transactions.

Validators and Wallets in Flash

There are two types of nodes in Flash, Validators and Wallets. Validators create and validate blocks, driving the network’s high TPS, while Wallets create new transactions.

# Unlimited Block Size in Flash

In Flash the block size is unlimited, allowing for a high volume of transactions to be processed in a single block. This is achieved through the combination of two key features:

Snapshots are created by Validators after each block cycle. Snapshots contain a full system record, meaning that information about a Wallet balance can be obtained without downloading the entire blockchain, as only the most recent Snapshot is required.

In traditional consensus algorithms, transactions are typically propagated through all nodes twice - once when they are shared with the network and again when they are included in a block. However, in Flash, Wallets communicate directly with Validators, eliminating the need for multiple propagations. Transactions are received by the Validator only once, either as part of a new block or directly from a Wallet, ensuring the efficient and streamlined processing of transactions.

Combined Voting and Transactions in Flash

In Flash, the voting process and transaction validation are combined in a single block, making blocks the sole communication channel between Validators. This unique approach allows for each Validator to produce only one new block per cycle, streamlining the consensus process and resulting in the creation of a new Snapshot. These Snapshots, which provide a full record of the system, are an integral part of the Flash network and help to maintain its fast and efficient operation.

Three-Step Consensus

Reaching consensus is achieved through a three-step process that Validators run in a loop. Collect Blocks - Validators wait for each Validator in the system to produce a block with the same ID (block N). Create a Snapshot - Validators examine the voting section of blocks N for transactions included in blocks N-1 and create a corresponding Snapshot (Snap N-1). Vote and Create a Block - Validators use Snap N-1 to cast votes on transactions in blocks N and create the next block (block N+1). This simple yet intricate process allows for efficient consensus and the creation of new Snapshots with each block cycle, ensuring a complete system record is always maintained.

Bootstrapping and Recovery

In the event of a system outage or for a new node joining the network for the first time, the Validator must obtain the latest Snapshot and blocks from the last two cycles to start the three steps loop. At least 2/3 of the Validators must be online in order to reach a consensus and create a new Snapshot

Validators List and Performance

The blockchain maintains a list of validators. The performance of each validator is measured and evaluated based on their ability to produce blocks efficiently. The better the performance, the higher the reward.

The network reaches a consensus on the number of validators to include in each block cycle through voting on a specific topic related to adding or removing validators from the list.

Each validator assesses the performance of the other validators by measuring the time it takes to receive a block from them. There are also timeouts in place, beyond which a validator will no longer wait for a block.

The scores that validators receive have an impact on the rewards they receive. In cases where a validator experiences frequent timeouts, they may be removed from the validator list.

Addition of New Validators

New Validators are invited by existing Validators. This process is a manual one and involves communication between human beings. The information about the new Validators is added to a dedicated topic in the block, which is shared with other Validators in the network.

Block Structure

A block in the system consists of several components:

Reward System

The reward system in Flash is based on a combination of stack and performance. The stack serves as a basic evaluation value, while penalties are applied for the following factors: negatively scoring other validators, slow performance, low number of transactions, and low number of blocks produced.

Additionally, the stack is not calculated linearly, but as S^stack where S is a number close to 1 (e.g. 1.0001). This encourages investing in node quality over quantity for better performance.

Holders Incentive

The Flash blockchain empowers holders with a unique reward system. The block reward is split 50/50 between the validators and the holders. To become a holder, coins must be put on hold for a period of 1-120 months. The number of coins held is used as the baseline for calculating rewards.

Holders are incentivized to keep their coins on hold for longer periods by multiplying the rewards by 1.016 for each additional month of holding. The system also encourages holders to keep adding more coins to their holding by implementing a linear penalty that reduces the rewards each day, resulting in zero rewards for holding coins on the last day.


Double-spending Attack

Flash is highly resistant to the double-spend attack as there are no forks in the system, and all blocks become part of the blockchain. Let’s examine some potential cases and see how Flash responds:

Network Attack

Attacks on the network by compromised Validators are not motivated as there is no way for them to double-spend coins. Only an attacker with the goal of sabotaging the system would perform such an attack.

To counter this, Flash encourages Validators to maintain high network performance, and it accounts for all nodes. If any nodes become slow or inaccessible, they are removed from the network.

In the event of a power outage or network attack, Flash waits for a consensus to be established among the Validators. The nodes cannot work on the next block until there is a consensus on the latest snapshot. In such scenarios, the block waiting timeout may be extended indefinitely.

To recover from a network attack, at least 2/3 of the active nodes must be available. In such cases, it is expected that people would pick up the phone and help to troubleshoot the problem manually.