What is Harmony (One)?
Fully Scalable:
Harmony shards not only the network communication and transaction validation like Zilliqa, but also shards the blockchain state.
This makes Harmony a fully scalable blockchain.
Secure sharding:
Harmony’s sharding process is provably secure thanks to the distributed randomness generation (DRG) process which is unpredictable, unbiaseable, verifiable and scalable.
Harmony also reshards the network in a non-interruptive manner to prevent against slowly adaptive byzantine adversaries.
Efficient and Fast Consensus:
Unlike other sharding-based blockchains which require PoW to select validators, Harmony is based on PoS and thus energy efficient.
Consensus is reached with a linearly scalable BFT algorithm that’s 100 times faster than PBFT.
Adaptive-Thresholded PoS:
The threshold of stakes required for a node to join the network is adjusted based on the volume of total staking in a way that malicious stakers cannot concentrate their power in a single shard.
Moreover, the threshold is low enough so that small stakers can still participate in the network and earn rewards.
Scalable Networking Infrastructure:
With RaptorQ fountain code, Harmony can propagate blocks quickly within shards or across network by using the Adaptive Information Dispersal Algorithm.
Harmony also adopts Kademlia routing to achieve cross-shard transactions that scale logarithmically with the number of shards.
Consistent Cross-Shard Transactions:
Harmony supports cross-shard transactions with shards directly communicating with each other. An atomic locking mechanism is used to ensure the consistency of cross-shard transactions.
Consensus Mechanism:
Example: Network communication during a single round of consensus.
Distributed Randomness Generation:
Unpredictable:
No one should be able to predict the random number before it is generated.
Unbiaseable:
The process of generating the random number should not be biasable by any participant.
Verifiable:
The validity of the generated random number should be verifiable by any observer.
Scalable:
The algorithm of randomness generation should scale to a large number of participants.
Scalable Randomness Generation with VRF and VDF:
Example: The VDF (Verifiable Delay Function) delays the revelation of the final randomness.
Staking-based Sharding & Sharding by Voting Shares:
Example: The stakers obtain voting shares proportional to their staked tokens. Voting shares are then randomly assigned to shards. Stakers become validators for the shard(s) where their voting shares are assigned.
Resharding:
Static Round-Adaptive:
Where attackers can only corrupt a subset of nodes at a predetermined stage.
Attackers can only corrupt nodes at the beginning of each epoch.
Slowly Adaptive:
Where attackers can corrupt a subset of nodes over time during the epoch.
Fully Adaptive:
Where attackers can corrupt a subset of nodes instantaneously and at any time.
Fast State Synchronization:
Example: The first block of an epoch contains a hash link to the first block of last epoch and this allows fast state synchronization of new nodes where they can rely only on the blocks in grey to quickly verify the current state.
Shard Chain and Beacon Chain:
Cross-shard Communication:
Main-chain-driven: Projects like Zilliqa rely on the main chain to achieve transactions across shards.
Client-driven: Omniledger proposed a client-driven cross-shard transaction mechanism where the messages between shards are collected and sent to shards by clients.
This adds an extra burden to the client that is not desirable for an adhoc light client.
Shard-driven:
RapidChain proposed that the messages between shards are directly sent by the nodes in the shard without external help.
Hash Link from Shard Chain:
Example: Hash link from beacon chain block to shard chain block.
The hash of its previous block, which must have already been committed in the beacon chain.
The signers of the block’s multi-signature, which must be the correct validators for that shard.
Increases the difficulty of attacking a single shard.
Attackers have to corrupt both the shard chain and beacon chain in order to convince others that an alternative block in the shard chain is valid.
Reduce the network cost of broadcasting the block headers among shards.
There will be a O(N ) network communication if we let each shard broadcast its headers separately.
With the beacon chain as a central relay, the complexity is reduced to O(N ).
Blockchain State Sharding:
Networking:
Kademlia-based Routing.
Efficient Broadcasting with Erasure Code:
Example: Comparison between normal gossip broadcast with gossip broadcast with erasure code.
A newly proposed block needs to be broadcasted by the leader to all validators.
A newly generated master chain block needs to be broadcasted to the whole network.
The cross shard communication requires the broadcast of a message between shards.
FEC-based Unicast, support for home Nodes & support for locator mobility.
Incentive Model:
Consensus rewards, stake slashing & stake withdrawal.
Future Research:
Fraud Proofs & stateless validators.
Ecosystem and Partners:
Warning : Every investment and trading move involves risk. You should conduct your own research before making any decision.
Thanks to all.
The best regards and be save out there.
Native Crypto.
