Execution
The modular nature of Dijets multi-chain architecture helps scale its execution layer.
Storage
Verifiable Data Pruning in state management brings unprecedented optimisation to Dijets storage layer
Consensus
Dijets highly scalable consensus uses a probabilistic model of repeated random sub-sampled voting
Horizontal scaling of Dijets execution layer spreads the computation of transactions across chains whilst maintaining high composability.
Performance
Scalability has always been one of the key defining metric for a blockchain performance. High transaction latencies directly impact end-users experience, e.g., the relatively high finalisation times of Bitcoin transactions (10 minutes) make it unsuitable for interactive trade. execution, storage, and consensus. With the ultimate goal of unlocking blockchain technology for all users and use cases we set out to build Dijets.
Dijets is a high-fidelity, high-security blockchain network that offers unmatched scalability, eliminating the consensus bottleneck of other blockchains without sacrificing security or decentralisation.
To quantify Dijets scalability, we deploy a test network using Dijets-CLI tool with varying network sizes composed of machines ranging from enterprise grade hardware with high computational power (datacenter) to commodity hardware with modest computational power including a Raspberry Pi.
Average throughput with moderate workload of 100-200 Tx requests
Dijets
5480/s
Polkadot
1500/s
Ripple
1200/s
Binance
1000/s
Ethereum
20/s
Solana
2460/s
Avalanche
1800/s
Algorand
1180/s
3.7x
faster throughput
Average throughput when stressed with a constant workload of 1,000 Tx requests
Dijets
4660/s
Polkadot
750/s
Ripple
560/s
Binance
280/s
Ethereum
5/s
Solana
890/s
Avalanche
650/s
Algorand
440/s
6.2x
faster throughput
Note the stated performance results are not indicative of Peak Throughput.
The two performance evaluation setups measure a very specific notion of the availability of a blockchain as its ability to commit submitted transactions in a timely manner even when stressed with load peaks.
A blockchain is more available when it handles more intense bursts of transactions with low latency and without dropping any transaction. As a property this should be desirable for a scalable blockchain in order to handle realistic workloads of thousands of transactions per second wihtout increasing the hardware requirements for running a node.
Communication Overhead
Classical consensus protocols typically include an all-to-all voting and has a designated leader who initiates the decision process and a series of all-to-all communication rounds to ensure that all correct/reputedly-good nodes reach the same decision. Such a protocol typically requires quadratic communication overhead with all-to-all communication of O(n²). This means that for each round:
- - A network of 10 nodes, equates to 100 messages
- - A network of 1,000 nodes, equates to 1,000,000 messages
- - A network of 100,000 nodes equates to 10,000,000,000 messages
Dijets in comparison solves the communication overhead problem by keeping the number of messages each node has to handle per decision constantly low even if the network scales up to millions of validators, each one of those validators can participate in the network consensus without ever effecting its performance.