Disclaimer: Speculative Rationality only look closely at and write about projects we believe in, have invested in or plan on investing in. Information below aims to be impartial; however it is not investment advice. Do your due diligence. Good luck!
Please attribute sourcing of content and images from this article to www.spec-rationality.com
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Recommended Reading to assist in better informing this post:
- In Depth Reading: Whitepaper
- Additional reading: TreeReduce and TreeAggregate Demystified
- Video Interview: Constellation Execs talk with BlockchainBrad about their Horizontally Scalable Distributed OS
Scalability is one of the greatest challenges faced by distributed ledger technologies (DLT) as they evolve towards maturity. A great deal of energy has been committed to finding appropriate solutions, with proposals taking many forms- ranging from second layer solutions, alternate consensus methodologies and even to alternate data structures. The question of how to scale a truly decentralised public ledger remains without a definitive answer.
Blockchains have been central to the narrative so far with the likes of Bitcoin and Ethereum playing vital roles in bringing attention to the space. As the demands on Blockchain networks have grown we have seen lively and at times divisive debate erupt about how best to scale those networks. One class of solution is the implementation of “Level 2 solutions” like Lightning and Plasma which take the burden of increased transactional volume or computational complexity “off-chain”. Another class of solution is to push towards Consensus mechanisms which lend themselves more easily to scaling solutions than Proof of Work chains; one of the most popular alternate consensus methodologies has been Proof of Stake (PoS), with titans like Ethereum moving towards a PoS model in future upgrades with a view to implement sharding.
Perhaps one of the most interesting proposals is the move away from Blockchain itself, with novel and new data structures that do away with traditional Blockchain architecture in varying degrees. Directed Acyclic Graphs (DAGs) are one example of alternate DLT architecture; DAG’s themselves are not homogenous, with various iterations manifesting themselves and diverse approaches to consensus, security and adaptability emerging.
Constellation is a new project which introduces to the DAG space — Hylochain, A lightweight consensus architecture. Hylochain couples a reputation based consensus model with the Directed Acyclic Graph (DAG) data structure. The table below illustrates where Constellation fits into the current landscape.
View Table: Projects Comparison
Constellation, in the context of DAG architecture is particularly noteworthy for both its introduction of an interesting new consensus methodology coined ‘Proof of Meme’ and for its implementation of smart contracts as microservices. This article will help to illustrate the Hylochain Consensus Architecture and its unique properties and a follow up article will cover the implementation of smart contracts as microservices.
Achieving Scalability: The Hylochain Consensus Architecture
Notable features of the Constellation Consensus Architecture:
- HyloChain — Hylochain takes its name from the function of Hylomorphism that is utilised in the underlying consensus architecture. A technical description — hylomorphism is a recursive function, corresponding to the composition of an anamorphism (which first builds a set of results; also known as ‘unfolding’) followed by a catamorphism (which then folds these results into a final return value). Fusion of these two recursive computations into a single recursive pattern then avoids building the intermediate data structure. Hylochain also leverages dynamic partitioning of nodes (second protocol on top of the consensus protocol) within the network to ensure network is always working optimally.
- ExtendedTrustChain — Architecture that implements multiple node types, each with their own responsibility and role in the network to asynchronously govern different aspects of the protocol. Standard nodes, Checkpoint nodes and Validators nodes. In ExtendedTrustChain, each node operates as an account, maintains the history of its own chain, and relies upon transaction order for validity (similar to HashGraph). (Whitepaper)
- Proof of Meme — Reputation based consensus mechanism
- MapReduce — process of breaking computation into simple operations that can be fed into an asynchronous DAG of computation, thereby increasing the efficiency of an already concurrent program.
- Gossip Protocol — allows for large networks to communicate total network state at a scale that is orders of magnitude higher than existing blockchain technology.
What does consensus in Constellation look like in practice?
“ A round of consensus takes the resulting hash block of a previous round and adds it as a regular transaction to the transaction pool. The filling of the transaction pool is isomorphic to an unfold operation. Once this checkpoint block is filled with the previous round plus new transactions, it is hashed. This is isomorphic to a fold operation.” — Whitepaper
To visualise this process let’s take a look at the necessary steps taken throughout the consensus process.
The diagram below illustrates the following steps:
- Bob is a Star Node and his transaction is recorded on his node’s local ledger, as are all interactions with the network.
- Transactions are signed by the initiator (Bob) and counterparty then broadcasted to the network (via gossip)
- The message gets pooled with other gossiped message into a local transaction pool (Anamorphism — unfold)
- A Star Cluster formed of Stars who elected to take part in consensus form a locality sensitive hash block. It forms the block by agreeing on all transactions and the order in which they occured within the transaction pool and hashing it into a locality sensitive (belonging to a locale) hash block. The locality sensitive hash block is the catamorphic (fold) function of the transaction pool which also includes the previous block’s catamorphic value.
- As the locality sensitive hash blocks are gossiped upwards they are pooled into transaction pools on higher tiers, galaxy tier pools (Anamorphism of locality sensitive hash blocks). A Galaxy Cluster then forms consensus by agreeing on all transactions and their ordering to form the Black Hole block (Catamorphism). They also add the hash of the previous black hole to this catamorphic function. “Black Holes are blocks of hashed locality sensitive hash blocks. It is equivalent to refer to them as blocks in the blockchain.” — Whitepaper
So what is the unique property of this process compared to other architectures?
Constellation utilises dynamic partitioning/topology that continuously updates the network consensus structure to be at its most efficient. That is to say Constellation implements a secondary protocol on top of ordinary consensus which balances the trie structure (Further reading). This in effect allows the “tiering” of nodes based on reputation and is intrinsically tied to validator rewards. For example as nodes enter the network, to ensure that network load is balanced and throughput is not compromised — the network adds intermediary tiers, ordering the network to the most efficient structure. The upward tiering allows for a theoretically infinite scalability as the network doesn’t stagnate when a certain number of nodes are reached.In the simplest terms — the architecture developed by Constellation dynamically adjusts its structure between rounds of consensus to ensure efficient throughput and effective load-balancing of network resources.
So how does this differ to current or proposed implementation of typical sharding?
Sharding utilises a linear topology — linear sub-shards connected to a linear mainchain (an incremental advance from a single main chain). As new nodes enter the network they are assigned to one of the shards wherein each shard has an upper limit of nodes that marks the optimal point for function and operations. As the network grows — it grows (linearly) with each additional shard, however in the absence of an ability to dynamically re-order/re-structure, sharding as a solution to scaling taps out at some point. For example when a given shard becomes congested, the network is not set to reorder itself to alleviate the issue automatically, nodes or participants would have to move to another less congested shard or wait for congestion to naturally unwind for optimal throughput to be restored.Constellation tries to ensure that as nodes enter the network, dynamic partitioning prevents the network from “tapping out” by iteratively re-organising to the most efficient structure. This in effect means that when a resource-intensive/high throughput event takes place it can do so without congesting certain parts of the network.
If we are to take the example above and extrapolate with additional tiers we can visualise the following steps:
- Bob is a Star Node and his transaction is recorded on his node’s local ledger, as are all interactions with the network.
- Transactions are signed by the initiator (Bob) and counterparty then broadcasted to the network (via gossip)
- The message gets pooled with other gossiped message into a local transaction pool (Anamorphism — unfold)
- A Star Cluster formed of Stars who elected to take part in consensus form a locality sensitive hash block. It forms the block by agreeing on all transactions and the order in which they occured within the transaction pool and hashing it into a locality sensitive (belonging to a locale) hash block. The locality sensitive hash block is the catamorphic (fold) function of the transaction pool which also includes the previous block’s catamorphic value.
- As the locality sensitive hash blocks are gossiped upwards they are pooled into transaction pools on higher tiers, Galaxy Tier 1 pool (Anamorphism of locality sensitive hash blocks). A Galaxy Cluster then forms consensus by agreeing on all transactions and their ordering to form the Tier 1 Block (Catamorphism). They also add the hash of the previous Tier 1 Block to this catamorphic function.
- Similar to Step “e” Tier 1 Blocks are gossiped upwards and are pooled into transaction pools on those respective tiers and initiate the same process as “e”.
- This is an ongoing cycle of blocks going “upwards” towards the “black hole” block. Blocks from “lower” galaxies gossip their blocks upwards, and when they hit the top tier of nodes they are merged into the global state of the network.
Constellation’s Hylochain architecture enables dynamic auto scaling based on network resources, allowing for consistent throughput.Constellation’s dynamic consensus takes into account the issue of “sleepy nodes”.
“All nodes in Constellation can be ”sleepy” i.e. they can join and leave the network at any time. As new nodes join the network, their resources are allocated to a sub-net. Once the sub-net reaches the threshold for transaction throughput and cryptographic security (the number of facilitators vs. the number of participants) as outlined [Page 11 of the whitepaper], entering nodes must form a new sub-net. Thus, new sub-nets will be dynamically allocated as members leave and join the network. It is trivial to show that this architecture exhibits the power law intrinsic of scale-free networks which are notorious for their application to fault tolerance in complex systems such as distributed computing.” — Constellation Whitepaper
The table below compares this consensus architecture to typical Blockchain and DAG architectures to better illustrate the key differences:
Proof of Meme — Alleviating pressure towards centralisation
It is important to acknowledge Constellation’s efforts to create a truly decentralised network by alleviating the pressure towards centralisation in their reputation based consensus mechanism, Proof of Meme (PoM). We can extrapolate from current world systems and attempts at decentralised networks such as Bitcoin and Ethereum that there will always be a constant pressure towards centralisation in regards to power of governance over a network or monopolising the incentives of a network. Proof of Meme introduces a low threshold for new nodes to enter and participate (incentives and governance) on equal footing with nodes (even where those other nodes have far more resources and history). The table below breaks down popular consensus mechanisms and their pressures towards centralisation.
View Table: Consensus Mechanism
Adoption: The Primary Mandate
Constellation’s primary mandate is Adoption. The keys to wide-spread/sustainable adoption are difficult to decipher, particularly in an increasingly saturated crypto space. If we’re talking about adoption — Constellation’s approach to technological innovation (scalable DLT — Hylochain Consensus), ease-of participation (Proof of Meme) and developer reach are notable.
In introducing a scalable DLT through the Hylochain Consensus architecture, the Constellation team has sought to solve for one of the key roadblocks faced by the space — scalability. A truly scalable smart-contract protocol can certainly be seen as a pre-cursor to widespread adoption, we will watch with interest to see how the platform develops as we approach testnet.
Throughput and scalability tend to dominate discourse in the space at the moment, however just as important are community involvement/participation. The Constellation team recognising this, has implemented The Proof of Meme consensus mechanism which ensures low threshold — widespread participation. Though we should note that with enough economic incentive there are always those who would engage in monopolistic behavior, Constellation’s efforts with Proof of Meme should act to alleviate this pressure to some degree.
Developer reach is another pivotal requirement for any hopeful smart-contract platform. Constellation is implementing an interface in which JVM can be utilised by all developers on the constellation network, ensuring that any language compatible with JVM can be utilised. They take this a step further by implementing Smart contracts as micro services, providing a modular aspect to the creation of applications and contracts (beyond the scope of this article, will be expanded on in a follow up article).
The current crypto landscape has become increasingly saturated and competition is rife, it is more difficult now than ever before to pick the winners in the space. What stands out to us about Constellation in both our analysis of their architecture and in our contact with the team is the synthesis of strong technology, business acumen and community awareness into a coherent platform and project narrative. We will watch with interest as emerging technologies like Constellation look to locomote the space forward.
_“Competitors do two things very well. They validate what you’re doing because there’s somebody else in the space tackling the same problem with a similar approach which is often g_ood. It also pushes you to develop faster in a more lean way to kind of beat them and it’s not necessarily competitive, it’s spurring on innovation.” — Brendan Playford, CEO of Constellation (Source: Blockchain Brad interview)
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Join SpecR Telegram for more active and focused discussions on tech and the idea of decentralized economy and its implications.
Join Constellation Telegram to engage in community discussions around the project.