2/ Decentralisation is a key component to what sets blockchain apart from traditional solutions, it increases security, censorship resistance, enables trust and is more inclusive. Centralisation makes them a target for law enforcement to enforce any changes.
3/ Classical consensus protocols are based on all-to-all voting and typically have a designated leader who initiates the decision process and a series of rounds of all-to-all communication to ensure that all correct nodes reach the same decision.
4/ They typically require quadratic communication overhead with all-to-all communication of O(n²). This means for each round if there is: A network of 10 nodes = 100 messages A network of 1,000 nodes = 1,000,000 messages A network of 100,000 nodes = 10,000,000,000 messages
5/ So an easy way that many blockchains cut down on that overhead and achieve better performance is simply to sacrifice decentralisation and only run a small number of nodes, or by only selecting a small committee size of nodes to participate in consensus. Examples:
6/ Binance Smart Chain uses a centralised Proof of Authority consensus with just 21 nodes, which the majority of nodes are controlled by Binance. The validator set for BSC is determined by Binance Chain which has just 11 validators (a permissioned network controlled by Binance)
7/ Classical consensus requires 66%+1 of the vote in order to finalise blocks, so if more than 33% of the network goes offline, the network will halt. Once a vote has reached 66%+1 vote then it doesn’t matter what the other validators vote as quorum has been reached.
8/ There are platforms which not only have a small number of total nodes but an even smaller number of those holding the majority of stake. Polygon’s $MATIC side chain has 2 nodes which control 45% of the network. If those 2 nodes go offline, the entire network will halt.
9/ Only 5 Nodes control 66%+1 of the stake, making the other 95 validators irrelevant. If those 5 nodes collude then security of the network is compromised. In classical consensus security can actually be compromised with as little as 33%+1 of the vote with a network partition.
10/ Not only does Polygon have a small total validator set of which just a handful of nodes control the network, the entire security of Polygon and all the apps built on it is controlled by a 5-out-of-8 multisig (4 of which are controlled by the Polygon team)…
11/ Others such as Fantom $FTM again have a very small number of validators with an even smaller number that control the network. Recently 2 nodes went offline which halted the entire network as they controlled over 33% of the stake.
12/ Dfinity $ICP is a permissioned network consisting of many subnets, all of which are controlled by the Network Nervous System. The subnets are their own blockchain and each secured by as little as 7 validators in total (without any rotation of validator sets).
13/ Dfinity requires highly customised spec (each node has 60 CPU cores, 2 TB of memory and 3.5 TB of NVRAM with enormous amounts of network throughput) and have to run them from approved datacenters, complete full KYC and apply / be approved to be even able to run a node
14/ Dfinity alone control the Network Nervous system. Whilst the DFINITY Foundation and the Internet Computer Association own 40% of voting power from their own neurons, the community are delegating their vote to them because they don’t have the knowledge / information and time
15/ to be frequently asked to vote on complex tech / decisions. They will just take the easy route, delegate their vote to the team and earn the rewards but this leads to centralisation, not just of the NNS, but for the entire network / all the subnets as NNS controls them all.
16/ Solana $SOL is another blockchain which requires very high resources, minimum 12 Cores / 24 threads with at least 2.8 GHz, 128 GB Memory, Multiple TB NVMe Drives, High end GPUs and 1 GB Network. It is already seeing high skip slot rates as some validators struggle to keep up.
17/ Solana is the only blockchain where consensus votes are done on-chain. Up to around 80% of the current TPS is used just by voting on consensus. As it is done on-chain validators also have to pay transaction fees (which is around 1 SOL per day or $13k a year). As the number of
18/ tps increases, so too will the transaction fees each validator will have to pay and as the validators with the largest stake will receive more of the rewards it leads to further centralisation. The more validators there are the more of the available tps is used just by voting
For more info on the disadvantages / inflated metrics of Solana see this twitter thread
19/ Others select a small sample of the overall validator set to actively participate in consensus to increase performance at the expense of decentralisation. Polkadot $DOT can have as little as 14 validators validating each parachain
20/ Decentralisation can also be reduced through the distribution method. Rather than a distributed peer to peer model where nodes connect to each other, Algorand $ALGO uses a central hub and spoke where all transactions have to pass through a permissioned set of relay nodes.
21/ Which performs signature verification / validity checks before forwarding them to nodes to vote. Leaving the potential for censorship of transactions flowing through the network. Whilst malicious nodes can’t compromise security of the network, they can cause it to halt.
22/ $ETH Layer 2 Solutions such as Optimism’s optimistic rollup will be centralised as initially everyone that is using the rollup sends their transactions to a single centralised sequencer that puts them into a batch and calculates the new state root to be posted on chain.
For more info about rollups see this thread
23/ The sequencer can choose to ignore transactions for a batch / include their own — to benefit from MEV / front running etc for a period of time. This also raises potential for censorship resistance through regulatory pressures to prevent certain DeFi activity.
24/ There is the potential that you could have just a single entity (maybe even nobody) checking the validity of a transaction by performing the computation off-chain with optimistic rollups, offering weaker security.
For more info about how rollups are currently centralised and have private keys which can steal funds see this thread
25/36 It’s clear the common theme is that decentralisation is being sacrificed to achieve higher performance. Thankfully there is a new revolutionary consensus called Avalanche which offers high throughput, sub second finality and can scale to millions of validators.
26/ Avalanche Consensus is the biggest breakthrough since Nakamoto. It uses repeated random sub-sampling of the entire network to quickly achieve consensus with minimal overhead per node for incredible performance. See this article for more details
27/ The number of messages each node has to handle per decision remains constant as the validator set increases, enabling it to be an inclusive consensus that can scale to millions of validators all participating in consensus to enable unparalleled levels of decentralisation.
28/ Hardware requirements are minimal compared to other blockchains with just 2 cores, 4 GB Memory and 32 GB Hard Drive enabling anyone to run one. The minimum stake amount is due to be reduced significantly once governance is implemented shortly.
29/ Being a leaderless blockchain, not limited to a small no. of validators, combined with sub second finality makes MEV pretty much a non-issue compared to other platforms. So not only do you benefit from cheap fees and a decentralised secure network but also don’t get front run
30/ But what about other blockchains that are restricted to small number of nodes? Avalanche is a platform of platforms where blockchains can build on top, swap out their legacy consensus with Avalanche consensus enabling higher performance and scale to millions of validators.
31/ Avalanche supports any custom VM so the Blockchains mentioned above could integrate their VM into Avalanche, migrate the state and benefit from the decentralisation and performance Avalanche provides whilst still being able to use their own native token.
32/ All blockchains built on Avalanche will be interoperable. As all validators also validate the primary network as they are not restricted to small amounts of validators like with Cosmos, then faster and more efficient interoperability will be possible compared to IBC
34/ Avalanche offers an incredibly customisable platform that not only caters for, but is the best platform for all requirements from DeFi, Payments, NFTs to Enterprises and the 1000’s of trillions of dollars in the traditional finance world.
35/ Avalanche is the only consensus which can scale to enormous amounts of validators all actively participating in consensus at the same time. Decentralisation is essential for the future of blockchain, and the Avalanche platform can help all other blockchains achieve that.
