Bitcoin with only about 8 transaction per second (TPS) and also big latency times (for final confirmations) is not very suitable for most blockchain applications. Similarly with most early generation distributed ledgers. Important things for performance include also the efficiency of virtual machine employed (trend is that the newest projects go toward versions of WebAssembly VM).
Faster and even scalable solutions are now in works. We first describe improvements on Bitcoin and Ethernet networks in the form of lightning and plasma networks on top of those, and then various projects which claim to be scalable, “3rd generation” etc. Essential improvements are various algorithms trying to go beyond linearity of classical blockchain projects, especially those relying on DAGs (directed acyclic graphs) and on so called sharding. As far as consensus algorithms there are scalability improvements with many new algorithms, while PoW is in a way most secure (for public blockchains).
While mining is computationally intensive it is considered that the optimizations with use of GPU and ASIC hardware are bad forces of centralization (not to mention unhealthy and energy consuming chase for success with single purpose hardware). It is considered better to use computational resources for more useful work like validation of smart contracts, security of computation, cryptographic tools for privacy protection and so on. With FPGA the situation is a little better, and some FPGA are claimed to be useful for smart contract processing and alike, see here there are also Chaincode virtual machine-specific ASICs by accelor.io.
Bitcoin solves some of the problems with so called lightning network on top of it, which makes temporary transactions with lots of intermediate transfers only off chain and only small footprint on the main chain, ensuring small price of transactions.
Ethernet recognizes the same problem as Bitcoin, but has significant plans of substantial upgrades of its Mainnet, namely the new Casper protocol, plans for deployment of WebAssembly virtual machine and sharding which, according to Buterin, could together bring million TPS when all employed as planned.
In addition, there is an analogue algorithm to lightning network targeted to Ethereum, with substantial new ideas in the algorithm, called plasma network.
There are several realizations of plasma, mostly used being the Loom network.
There are several major algorithms. Several projects are based on “hashgraph” paper
including Hedera which claims 50000 to 500000 TPS tested,
then so called Consensus project etc.
Somewhat similar but with a new innovative algorithm is Tolar Hashnet, see whitepaper pdf with around 150000 TPS achieved on testnet and low power consumption.
Another kind of scalable algorithms using DAGs is the tangle blockchain (of the iota project)
Telegram (see FAQ and wikipedia) is developing Telegram Open Network (TON) centered around a new generation blockchain project. Besides the blockchain and user interfaces, including on Telegram, it will contain analogues of other parts of the internet structure including TON proxy, TON DNS and TON storage. The blockchain uses delegated proof of stake with a number of innovations including infinite sharding paradigm and instant hypercube rooting. The whitepaper predicts that the platform will be able to support millions of TPS (private access to testnet allegedly given in April 2019). For more details see Telegram Open Network in this wiki and original
Australian academic RedBelly blockchain (also referred as CSIRO) has confirmed 40000 TPS with Amazon platform worldwide testing. Since mid 2017 it had local tests which could achieve up to 400000 but with all nodes in the same location; see the paper about various performance experiments (the largest performance experiment, with measuring various parameters, with up to 1000 nodes in 10 countries!),
and selection of news. The academic group behind the projects has also a number of interesting research papers.
Algorand project, founded by a group at MIT lead by S. Micali, claims 1000 tps and latency of only 5 seconds in a setup with 10000 participants and 500 nodes on testnet world wide. It got 62 M dollars equity funding in 2018 (news) and is issuing Algos coins in auction in June 2019, news. Undergone security audits from Trail of Bits and NCC. While the first prototype in 2016 has been coded in C++, the full implementation is in Golang, see github. The underlying algorithms are described in
There is a theoretical proposal to scale up by 3 orders of magnitude by the BlockReduce algorithm “which only segments consistency”. The paper shortly reflects on other approaches
Last revised on June 14, 2019 at 13:09:09. See the history of this page for a list of all contributions to it.