Polaris (PLR)

What is Polaris

We aim to combat fundamental weaknesses correlated with blockchain architecture. Pre-existing developments have exhausting problems such as slow block times and high transaction costs. Blockchain commercialization and utilization becomes more costly and taxing for companies this way. Our set of development tools will remedy this, gathering advanced blockchain technology to experienced developers and amateur developers alike. We present the first candidate for a secure sharding protocol for permission-less blockchains. At its core, it will scale up the agreement throughput near linearly with the computational power of the network and tolerates byzantine adversaries which controls up to one-fourth computation capacity. It offers promising scalability and usability in next-generation cryptocurrencies.

Abstract

Currently the issue with cryptocurrencies is that in real world applications they are limited by the transaction speed and scalability which makes them un-realistic for mass adoption. We take a new approach to overcome these problems by breaking down processing transactions into smaller groups of nodes called shards.

Large networks require increased computational power which leads to slower transactions. By breaking down these transactions into smaller shards, it maximizes performance significantly allowing the system to scale much more efficiently.

Unfortunately, standard sharding based blockchain protocols only benefit partially from sharding due to the fact that they still require a limited amount of communication from participants per transaction.

Due to this issue, it causes the throughput and latency protocols to slow down significantly which decreases the security and limits their potential to become an everyday, mainstream payment solution.

In this paper we propose Polaris the first sharding based public blockchain protocol that can achieve full communication, computation, and storage overhead of all processing transactions without relying on any trusted setup. It is completely safe from Byzantine faults from up to 2/3 fraction of its participants.

To ensure all blocks are secure, Polaris uses a flawless intra committee consensus algorithm, that can create extremely high throughput via block pipelining, an innovative hearsay protocol for large blocks.

Our protocol uses a dynamic cross-shard transaction verification procedure to avoid hearsay across the entire network.

With all this said Polaris can process and authenticate a speed of 1,000,000 tps with confirmation latency of 4 seconds in a network of 5,000 nodes with no failure.

Results.

Without loss of generality, we assume that the net- work contains n processors which have equivalent computational power. Polaris exhibits almost linear scalability with computation capacity and does not require quadratic number of messages as the network grows. Polaris tolerates up to f < n/4 adaptive byzantine adversaries, where f and n are bounds on the adversarial and total computational power respectively. 1 The protocol can support the same blockchain data structure format (a hash-chain) as Bitcoin; but, for further scalability, we propose a modification that permits better efficiency parameters.

From an efficiency perspective, our protocol shards the network into an almost linear number of committees that scales with computation capacity. Within each committee of size c (a few hundred) identities, we run a secure consensus protocol which has message complexity of O(c2) (best case) to O(c3) (worst case). Overall, this yields a message complexity of at most O(nc3), where messages are of constant size.

We implement Polaris based on the most popular client for Bitcoin [23]. Our implementation adds roughly 5, 000 C++ LoCs on top of Bitcoin. The throughput of our prototype scales near linearly with respect to available computation i.e., O(n/ loglog(n)), when runs on our network simulation. With the same network implementation as in Bitcoin, the scale up (blocks per epoch) for 100, 200, 400, 800 and 1, 600 nodes with equal computational power 2 are as theoretical expectation, namely 1, 1.89, 3.61, 6.98 and 13.5 times respectively.

Finally, Polaris’s clean-slate design decouples the consensus from block-data broadcasts, hence the bandwidth spent by each node remains almost constant, regardless of the size of the network. Our simulations are necessarily on a smaller scale than Bitcoin; however, if we project our results to a full deployment to a network of Bitcoin’s scale, we can expect a scale up of 10, 000 in the number of agreed values per epoch. This agreement throughput is 4 orders of magnitude larger than Bitcoin’s.

Contributions. We claim the following contributions.