Hedera 백서

Hedera

원문

bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 6쪽 중 1쪽

1 Dr. Leemon Baird, Mance Harmon, and Paul Madsen The trust layer of the internet Hedera: A Public Hashgraph Network & Governing Council WHITEPAPER V .2.1 L AST UPDATED AUGUST 15, 2020 S UBJECT TO FURTHER REVIEW & UPDATE 2WHITEPAPER Vision To build a trusted, secure, and empowered digital future for all. Mission We are dedicated to building a trusted and secure online world that empowers you. Where you can work, play, buy, sell, create, and engage socially. Where you have safety and privacy in your digital communities. Where you are confident when interacting with others. Where this digital future is available to all. Hello future. © 2018-2020 Hedera Hashgraph, LLC. All rights reserved. 3EXECUTIVE SUMMARY Executive Summary Distributed ledger technologies (DLT) have the potential to disrupt and transform existing markets in multiple industries. However, in our opinion there are five fundamental obstacles to overcome before distributed ledgers can be widely accepted and adopted by enterprises. In this paper we will examine these obstacles and discuss why Hedera Hashgraph is well-suited to support a vast array of applications and become the world’s first mass-adopted public distributed ledger. 1. PERFORMANCE - The most compelling use cases for DLT require hundreds of thousands of transactions per second, and many require consensus latency measured in seconds. These performance metrics are orders of magnitude beyond what current public DLT platforms can achieve. 2. SECURITY - If public DLT platforms are to facilitate the transfer of trillions of dollars of value, they will be targeted by hackers, and so will need the strongest possible network security. Having the strongest possible security starts with the consensus algorithm itself, with its security properties formally proven mathematically. Security vulnerabilities and attack vectors shouldn’t be mitigated; they should be eliminated entirely. To improve performance metrics, some public DLT platforms are making compromises with regard to decentralization, and in so doing are potentially compromising security. 3. GOVERNANCE - A general-purpose public ledger should be governed by representatives from a broad range of market and geographic sectors, each with world-class expertise. Those responsible for network governance need technical expertise so they can competently manage the platform’s underlying software. They need business and economics expertise so they can manage business operations of the organization and its cryptocurrency. They need legal expertise to help navigate the evolving regulatory environment. In other words, the network should be governed by a decentralized group of globally recognized industry leaders, representative of every market in the world. 4. STABILITY - Without technical and legal mechanisms to enforce the decisions of the governing body, public DLT platforms are at risk of devolving into chaos. Strong security and mature governance will enable a stable platform — one that engenders the necessary trust and confidence among those that would build commercial or sensitive applications on it. 5. REGULATORY COMPLIANCE - We expect that governments will continue to extend policy objectives to users, enterprises, and developers utilizing public ledgers and associated cryptocurrencies and tokens. We consider that a public distributed ledger must be capable of providing necessary tools for all members of its ecosystem to comply with applicable laws and regulations, such as the European Union’s General Data Privacy Regulations (GDPR), and enable appropriate identity management to conduct sanctions screening and facilitate Know Your Customer (KYC) and Anti Money Laundering (AML) checks. 4EXECUTIVE SUMMARY What is required to move the DLT industry forward and enable it to realize its full potential? A platform that provides a combination of high performance, strong security, effective governance, technical and legal controls to ensure the platform’s stability, and tools to enable regulatory compliance. Only then do we think mainstream markets will trust a DLT platform enough to adopt it. 5HEDERA HASHGRAPH COUNCIL Introducing Hedera  — a p ublic hashgraph network and governing body designed to address the needs of mainstream markets. The Hedera network will be governed by a council of leading global enterprises, across multiple industries and geographies. Its vision is a cyberspace that is trusted and secure, without the need for centralized parties with inordinate influence. Its licensing and governance model protects users by eliminating the risk of forking, protecting the integrity of the codebase, and providing open access to review the underlying software code. Platform governance will be decentralized through the Hedera Governing Council, which will have a term- limited, rotating set of governing members that each have equal voting rights over key decisions relating to the platform.1 The Hedera network is a distributed ledger platform that resolves the factors that constrain adoption of public DLT by the mainstream. 1. 2. PERFO RMANCE - T he platform is built on the hashgraph distributed consensus algorithm, invented by Dr. Leemon Baird. The hashgraph consensus algorithm provides nea r-perfect efficiency i n bandwidth usage and consequently can process hundreds of thousands of transactions per second in a single shard (a fully connected, peer-to-peer mesh of nodes in a network). Initially, we anticipate that the Hedera network will be able to process 10,000 cryptocurrency t ransactions per second. Consensus latency is measured in seconds, not minutes, hours, or days. S ECURIT Y - The hashgraph consensus algorithm achieves the gold standard for security in the field of distributed consensus: asynchronous Byzantine Fault Toler ance (aBFT). Other platforms that use coordinators, leaders, or communication timeouts to improve performance tend to be vulnerable to Distributed Denial of Service (DDoS) attacks. Hashgraph is resilient to these types of attacks against the consensus algorithm because there is no such leader. Achieving this level of security at scale is a fundamental advance in the field of distributed systems. Many applications require that the consensus order of transactions match the actual order in which the transactions are received by the network. It should not be possible for a single party to prevent the flow of transactions into the network, nor influence the order of transactions in the eventual network consensus. A fair c onsensus algorithm ensures that if a user can submit a transaction to the network at all, then the transaction will be received by the network and the order in which it was received will be a fair ordering. Hashgraph uniquely ensures that the actual order transactions are received by the network will be reflected in the consensus order. In other words, hashgraph ensures both Fair Access and Fair Ordering. Formal proofs of the aBFT and fairness properties for the hashgraph consensus algorithm have been available for public review since June 2016. Furthermore, the hashgraph algorithm was validated as aBFT by a math proof checked by computer using the Coq system in October 2018.2 6HEDERA HASHGRAPH COUNCIL 3. GOVERNANCE - The Hedera network will be governed by a council of up to 39 leading global enterprises. Hedera Council members will bring needed experience in process and business expertise that has been absent in previous public ledger platforms. Council membership is designed (i) to reflect a range of industries and geographies, (ii) to have highly respected brands and trusted market positions, and (iii) to encompass competing perspectives. The terms of governance ensure that no single Council member will have control, and no small group of members will have undue influence over the body as a whole. 4. STABILITY - Hedera relies on both technical and legal controls to ensure the stability of the platform. Hedera technical controls enable two capabilities. i) First, the hashgraph technology ensures that software clients validate the pedigree of the Hedera hashgraph ledger prior to use through a shared state mechanism. It isn’t possible for a network node to fork the official version of the Hedera hashgraph platform, make changes, and then have those changes accepted as valid. If the original hashgraph platform and the copy are changed independently, software clients using the Hedera platform will know which is the valid version and which is not. ii) Second, the hashgraph technology makes it possible for the Hedera Council to specify the software changes to be made to network nodes, precisely when those changes are to be adopted , and to confirm that they have been adopted. When the Hedera Council releases a software update, network nodes will have their software automatically updated at exactly the same moment. Any node with invalid software (i.e., one that didn’t install the software update) will no longer be able to modify the ledger or have the world accept their version of the ledger as legitimate. Hedera legal controls ensure the platform will not fork into a competing platform and cryptocurrency. iii) The Hedera codebase will be governed by the Hedera Governing Council, and will be released for public review with Version 1.0, expected to be released in 2020. It will not be open source, but anyone will be able to read the source code, recompile it, and verify that it is correct. No license will be required to use the Hedera platform. No license will be required to write software that uses the services of the Hedera platform. No license will be required to build smart contracts on top of the Hedera platform. Applications built upon the Hedera platform can be open source or proprietary. They do not require any license or any approval from Hedera. Hedera will make the code publicly available for anyone to review it (at Version 1.0), while ensuring stability by using hashgraph software patents defensively to prevent forks. In this way, Hedera will provide a transparent codebase that will provide the stability that markets demand 7HEDERA HASHGRAPH COUNCIL for mainstream adoption of a public ledger. The combination of technical and legal controls provide the governing body with the mechanisms needed to enable meaningful governance, and to bring the stability that we think is required for broad-based adoption. 5. REGULATORY COMPLIANCE – The Hedera technical framework includes controlled mutability of the network state and the potential to request or attach additional data to transactions, such as identity certificates. These features enable future functionality such as the erasure of personal data and other uploaded files and opt-in verified identity mechanisms – all optional and within the control of the end users. We intend to work with regulators and encourage development of such tools to allow enterprises to fulfill their consumer protection and regulatory compliance obligations. 1 Swirlds, Inc. is a Delaware corporation that holds the patent titles to the hashgraph consensus algorithm. Swirlds is a permanent member of the Hedera Governing Council. 2 See Appendix 3 for a full definition of the hashgraph algorithm, including proofs of aBFT. Also see https://www.hedera.com/blog/coq- proof-completed-by-carnegie-mellon-professor-confirms-hashgraph-consensus-algorithm-is-asynchronous-byzantine-fault-tolerant for access to the formalized aBFT proof. 8 Part 1 An introduction to Hedera Hashgraph INTRODUCTION 9INTRODUCTION The hashgraph data structure and consensus algorithm provides a new platform for distributed consensus. This introduction gives an overview of how hashgraph works, and some of its properties. The goal of a distributed consensus algorithm is to allow a community of users to come to an agreement on the order in which some of the users generated transactions, when no single member is trusted by everyone. In this way, it is a system for generating trust, when individuals do not already trust each other. Hashgraph achieves this in a fundamentally new way. A blockchain is like a tree that is continuously pruned as it grows – this pruning is necessary to keep the branches of blocks from growing out of control and to ensure the ledger consists of just one chain of blocks. In hashgraph, rather than pruning new growth, such growth is woven back into the body of the ledger. 10INTRODUCTION In both blockchain and hashgraph ledgers, any user can create a transaction, which will eventually be put into a container (the “block”) and will then spread throughout the distributed network. In blockchain, those “blocks” of containers are intended to form a single, long chain. If two blocks are created at the same time, the network nodes will eventually choose one chain to continue and discard the other one, lest the blockchain “fork” into two different chains. It is like a growing tree that is constantly having all but one of its branches chopped off. In hashgraph, every container of transactions is incorporated into the ledger — none are discarded — so it is more efficient than blockchains. All the branches continue to exist forever, and are woven together into a single whole. Furthermore, blockchain fails if the new containers arrive too quickly, because new branches sprout faster than they can be pruned. That is why blockchain needs proof-of-work or some other mechanism to artificially slow down the growth. In hashgraph, though, nothing is thrown away. There is no harm in the hashgraph data structure growing quickly. Every member can create transactions and containers whenever they want. Finally, because the hashgraph does not require pruning of would-be “forks” (as every container of transactions is incorporated into the ledger), hashgraph allows more powerful mathematical guarantees, such as Byzantine agreement and fairness. Distributed databases such as Paxos are Byzantine, but do not guarantee fairness in the ordering of transactions. Blockchain is neither Byzantine nor fair. The hashgraph algorithm accomplishes being fair, fast, Byzantine, ACID compliant, efficient, inexpensive, timestamped, and DoS resistant. 11PERFORMANCE Performance COST A hashgraph distributed ledger is inexpensive to operate compared to blockchain distributed ledgers, as it avoids energy-intensive proof-of-work. Individuals and organizations who want to run hashgraph nodes will not need to purchase expensive custom mining rigs. Instead, they will be able to run hashgraph nodes via readily available hardware that is less expensive than such specialized mining rigs. EFFICIENCY The hashgraph is 100% efficient , as that term is used in the blockchain community. In blockchain, work is sometimes wasted mining a block that later is considered stale and is discarded by the network of nodes. In hashgraph, the equivalent of a “block” of transactions never becomes stale. Hashgraph is also efficient in its use of bandwidth. Whatever the amount of bandwidth required to inform all the nodes of a given transaction (even without achieving consensus on a timestamp for that transaction), hashgraph adds only a very small overhead of additional bandwidth to achieve a consensus timestamp and put the transactions into order. Additionally, the hashgraph voting algorithm does not require any additional messages be sent in order for nodes to vote on validating transactions (or those votes to be counted) beyond those messages by which the network nodes learned of the transaction itself. THROUGHPUT The hashgraph is fast. It is limited only by the bandwidth. If each network node has enough bandwidth to download and upload a given number of transactions per second, the network as a whole can handle close to that many transactions per second. Even a fast home internet connection could enable a hashgraph node to be fast enough to handle transaction volume equal to that of the entire global VISA card network. 12PERFORMANCE THE FOLLOWING CHARTS GIVE PERFORMANCE RESULTS FOR HASHGRAPH TECHNOLOGY UNDER TEST CONDITIONS. The charts on the previous pages show the results of tests performed using hashgraph nodes hosted by Amazon AWS m4.4xlarge instances. Figure 1 shows performance for a single region (Virginia), Figure 2 shows results for two regions 2,000 miles apart on opposite sides of the continental United States (Virginia and Oregon), and Figure 3 shows results for 8 regions (Virginia, Oregon, Canada, Sao Paulo, Australia, Seoul, Tokyo, and Frankfurt). Each line on the chart is for a different number of instances (computers), which is shown to the right of the line. In every case, the instances were distributed evenly across the number of regions being used. The horizontal axis is the number of 100-byte transactions per second for which the ledger achieved consensus. In these experiments, this throughput ranges from less than 50,000 tps up to almost 500,000 tps. On most of the lines, the second dot from the left is 10,000 tps. The vertical axis is the average number of seconds from when a node first creates a transaction until it knows the exact consensus order and consensus timestamp for it. This isn’t just a time to a first confirmation — it is the time until a 100% certain finality is reached. In all of the experiments, this latency was under 11 seconds. Various experiments had latencies down to less than 0.04 seconds. In the graphs, there are clear tradeoffs between throughput, latency, number of computers, and geographic distribution. For 32 computers running at 50,000 transactions per second, consensus finality is reached in 3 seconds when the network is spread across 8 regions spanning the globe. When the network stretches only 2,000 miles across the U.S., this drops to 1.5 seconds. In a single region, it drops to 0.75 seconds. If it is desired to keep the latency under the 7 seconds required by credit cards, while still achieving 200,000 transactions per second, it is possible to use 32 computers in eight regions, or use 64 computers in two regions, or use 128 computers in one region. It is important to note that these tests are purely for achieving consensus on transaction order and timestamps. They do not include the time to process transactions. For example, if every transaction is digitally signed, then these results suggest that a great deal of processing power might be needed to verify hundreds of thousands of digital signatures per second. It is possible that GPU implementations could be helpful. In addition, if a transaction is a request to store a large file, then bandwidth limitations would slow down the network’s ability to process that transaction. Network pricing for file storage transactions will be structured to make it relatively expensive to store large files on the ledger, which should disincentivize any efforts to slow down the network in this way. 0.01 0.1 1 10 100 0 500,000400,000300,000200,000100,00050,000 150,000 250,000 350,000 450,000 64 computers 32 computers 16 computers 8 computers 4 computers 2 computers 128 computers 400,000 tps Figure 1 Hashgraph Latency vs Throughput 1 region, m4.4xlarge Latency (seconds) Throughput (100-byte transactions per second) 13PERFORMANCE 0.01 0.1 1 10 100 0 500,000400,000300,000200,000100,00050,000 150,000 250,000 350,000 450,000 Figure 2 Hashgraph Latency vs Throughput 2 regions, m4.4xlarge Latency (seconds) Throughput (100-byte transactions per second) 64 computers 32 computers 16 computers 8 computers 4 computers 128 computers 14PERFORMANCE 0.01 0.1 1 10 100 0 500,000400,000300,000200,000100,00050,000 150,000 250,000 350,000 450,000 Figure 3 Hashgraph Latency vs Throughput 8 regions, m4.4xlarge Throughput (100-byte transactions per second) Latency (seconds) 64 computers 32 computers 16 computers8 computers 15PERFORMA
프로젝트로 돌아가기