DxChain Network Design Principle & Architecture
(
)
The architecture of DxChain is adapted from the IPFS, Hadoop HDFS, IOTA, Golem, Plasma and others! The entire system is designed based on the principle of incentivisation, which means that miners cam maximize the utilization of unused storage resources and facilitate decentralized big data computations with lower costs and much conveniences. The following section will describe the architecture of the DxChain. The Chains-on-chain design, storage and computing with consensus protocol as well as the incentive mechanism will be explained in the subsequent section!
DxChain architecture
DxChain provides decentralized big data, machine learning computation with the support of decentralized data storage. To achieve this complex project, the DxChain team is designing a specific chains-on-chain structure — which manages the master chain, storage chain and computation chain, to reach the consensus and to provide the incentive mechanism in a Byzantine environment.
A peer-to-peer storage network, such as Inter Planetary File System (IPFS), Swarm provides a flexible and extensible file system where the data-frame and schema can be build. The storage chain is built on top of the storage data model, in which Proof of Spacetime (POST) validates whether the storage provider really stores the files.
DxChain will start with Hadoop, the industry-proven big data platform, as its computation engine. The Hadoop running elements, such as the job and task tracker and workers, communicate with one another through using a computational chain to synchronize running jobs. With two consensus mechanisms: verification game and Provable Data Computation (PDC), this chain would incentivize nodes that provides computation power. The master chain orchestrates the storage chain and the computation chain to maintain the master chain blocks, where the participating nodes may provide storage and computation power to obtain incentives.
The figire shows the high-level DxChain infrastructure. From the figure, one can see how the master chain, storage chain and computation chain work together.
Chains-on-chain structure
DxChain team is designing the three-chain structure with two unique properties:
- side and master chains structure;
- immutable master chain and elastic side chains.
Property 1 splits the chains according to their functions.
Property 2 solves the problems of scalability, throughput and latency of the blockchain.
The master chain is in charge of maintaining transactions. Based on experience from BTC and ETH, this master chain is kept immutable. The side chains in the DxChain are in charge of storage and computation. These functions are time bounded, so that the blocks do not need to indefinitely store expired data. The DxChain team is designing a chain shortening algorithm which could remove unused data to save valuable block space.
Computing
Most of the popular blockchains are used for financial activity transactions, so their computations are simply checking transactional records on the chains; this type of computation does not require a lot of resources. However, the computations running on the DxChain Network are intended to support more general-purpose processing, such as database query and Map Reduce work. BTC uses blockchain to store all its transaction records to achieve consensus in the network and this consensus needs more than half of the active nodes to agree. It is impossible to store all computation states in a blockchain to ensure computation consensus; therefore, the DxChain team uses two mechanisms to ensure the correctness of the computations: verification game algorithm and Provable Data Computation (PDC).
Verification game algorithm is designed as a system with three main types of roles: solver, verifier and judge. This interactive system can testify to the correctness of a computation procedure without wasting too much computation power. Provable Data Computation (PDC) is a statistical framework to testify to the correctness of computation results with few copies of redundant computing.
Using verification game algorithm and PDC, DxChain Network is envisioned to solve the problems of outsourced computation and decentralized mining. Therefore, Map Reduce work could potentially be performed on this platform.
Hadoop
Hadoop is an industry-proven big data platform which includes HDFS for file storage, YARN or Map Reduce for job scheduling, as well as a collection of tools to run distributed 6 jobs. Since Hadoop is a centralized system, it needs coordinators to schedule jobs. The DxChain Network is designed to adapt the Hadoop ecosystem to a decentralized environment. In order to orchestrate the job tracker, task tracker and workers, a computation chain is designed to keep the computation states on the DxChain Network — it is used to validate the correctness of the computations and manage computation tasks. With the control of the computation states, a Map Reduce pipeline can be implemented on the DxChain. Built on Map Reduce, a collection of tools such as Pig, Hive and Mahout is intended to be implemented for database operations and machine learning algorithms. Some business intelligence operations may be established through using these tools.
DxChain Links
White Paper: docsend.com/view/8w3fman
Telegram: https://t.me/dxchain
Twitter: https://twitter.com/DxChainNetwork
Website: https://www.dxchain.com/
My referral link: https://t.me/DxChainBot?start=nhtvmv-nhtvmv![DX-Banner.jpg]
Coins mentioned in post:
✅ @dgsm, I gave you an upvote on your post! Please give me a follow and I will give you a follow in return and possible future votes!
Thank you in advance!
Congratulations @dgsm! You received a personal award!
Click here to view your Board of Honor
Congratulations @dgsm! You received a personal award!
You can view your badges on your Steem Board and compare to others on the Steem Ranking
Vote for @Steemitboard as a witness to get one more award and increased upvotes!