Original source: Chain Tea House

1. Project Introduction

io.net is a distributed GPU system based on Solana, Render, Ray, and Filecoin, designed to use distributed GPU resources to solve computing challenges in the fields of AI and machine learning.

io.net solves the problem of insufficient computing resources by aggregating underutilized computing resources such as independent data computing centers, cryptocurrency miners, and excess GPUs from crypto projects such as Filecoin and Render, enabling engineers to obtain a large amount of computing power in an easily accessible, customizable, and low-cost system.

In addition, io.net introduces a distributed physical infrastructure network (depin), combining resources from various providers, enabling engineers to obtain a large amount of computing power in a customizable, cost-effective, and easy-to-implement manner.

io cloud now has more than 95,000 GPUs and more than 1,000 CPUs, supports rapid deployment, selects hardware, geographic locations, and provides a transparent payment process.

2. Core Mechanisms

2.1 Centralized Resource Aggregation

io.net's decentralized resource aggregation is one of its core features, which enables the platform to utilize decentralized GPU resources around the world to provide the necessary computing support for AI and machine learning tasks. The goal of this resource aggregation strategy is to optimize resource usage, reduce costs, and provide wider accessibility.

Here is a detailed introduction:

2.1.1 Advantages

Cost-effectiveness: By leveraging underutilized GPU resources in the market, io.net is able to provide lower-cost computing power than traditional cloud services. This is especially important for data-intensive AI applications, which often require a large amount of computing resources, and traditional methods can be costly. Scalability and flexibility: The decentralized model allows io.net to easily expand its resource pool without relying on a single vendor or data center. This model provides users with the flexibility to choose the resources that best suit their task needs.

2.1.2 How it works

Diversity of resource sources: io.net aggregates GPU resources from multiple sources, including independent data centers, individual cryptocurrency miners, and excess resources from other crypto projects such as Filecoin and Render. Technical implementation: The platform uses blockchain technology to track and manage these resources, ensuring transparency and fairness in resource allocation. Blockchain technology also helps automate payments and incentives allocated to users who contribute additional computing power to the network.

2.1.3 Specific steps

Resource discovery and registration: Resource providers (such as GPU owners) register their devices to the io.net platform. The platform verifies the performance and reliability of these resources to ensure that they meet specific standards and requirements. Resource pooling: Verified resources are added to the global resource pool and are available for rent by platform users. The distribution and management of resources are automatically executed through smart contracts, ensuring transparency and efficiency of the processing process. Dynamic resource allocation: When a user initiates a computing task, the platform dynamically allocates resources based on the task's requirements (e.g., computing power, memory, network bandwidth, etc.). Resource allocation takes into account cost efficiency and geographic location, optimizing task execution speed and cost.

2.2 Dual-Token Economic System

io.net's dual-token economic system is one of the core features of its blockchain network, designed to incentivize network participants and ensure the efficiency and sustainability of the platform's operations. This system includes two tokens: $IO and $IOSD, each of which plays a unique role. The structure and function of this economic system are described in detail below.

2.2.1 $IO Token

$IO is the main functional token of the io.net platform and is used for a variety of network transactions and operations. Its main uses include:

Payment and Fees: Users use $IO to pay for the rental of computing resources, including the use of GPUs. In addition, $IO is also used to pay for various services and fees on the network. Resource incentives: $IO tokens are issued as rewards to users who provide GPU computing power or participate in maintaining the network to encourage them to continue to contribute resources. Governance: $IO token holders can participate in the governance decisions of the io.net platform, including voting rights, and influence the future development direction and policy adjustments of the platform.

2.2.2 $IOSD Token

$IOSD is a stablecoin pegged to the US dollar, designed to provide a stable value storage and transaction medium for the io.net platform. The main functions are as follows:

Value stability: The value of $IOSD is fixed to the US dollar 1:1, providing users with a payment method that avoids the volatility of the crypto market. Easy transactions: Users can use $IOSD to pay platform fees, such as computing resources, to ensure the stability and predictability of transactions in value. Fee coverage: Certain network operations or transaction fees can be paid with $IOSD, which simplifies the fee settlement process.

2.2.3 How the dual token system works

io.net's dual token system interacts in several ways to support the operation and growth of the network:

Resource provider incentives: Resource providers (such as GPU owners) receive $IO tokens in return for contributing their devices to the network. These tokens can be used to further purchase computing resources, or traded on the market. Fee Payment: Users use $IO or $IOSD to pay for the use of computing resources. Choosing $IOSD can avoid the risks brought by cryptocurrency fluctuations. Economic Activity Incentive: Through the circulation and use of $IO and $IOSD, the io.net platform is able to stimulate economic activities and increase the liquidity and participation of the network. Governance Participation: $IO tokens also act as governance tokens, enabling holders to participate in the governance process of the platform, such as proposals and voting decisions.

2.3 Dynamic Resource Allocation and Scheduling

io.net's dynamic resource allocation and scheduling is one of the core functions of the platform. The key lies in efficiently managing and optimizing the use of computing resources to meet the diverse computing needs of users. This system ensures that computing tasks can be executed on the most appropriate resources in an intelligent and automated manner, while maximizing resource utilization and performance.

The following is a detailed introduction to each aspect of this mechanism:

2.3.1 Dynamic Resource Allocation Mechanism

1. Resource Identification and Classification:

When a resource provider connects its GPU or other computing resources to the io.net platform, the system first identifies and classifies these resources. This includes evaluating their performance indicators such as processing speed, memory capacity, network bandwidth, etc. These resources are then tagged and archived so that they can be dynamically allocated according to the needs of different tasks.

2. Demand matching:

When users submit computing tasks to io.net, they need to specify the requirements of the task, such as the required computing power, memory size, budget constraints, etc. The platform's scheduling system analyzes these requirements and selects matching resources from the resource pool.

3. Intelligent scheduling algorithms:

Advanced algorithms are used to automatically match the most suitable resources with the submitted tasks. These algorithms take into account the performance, cost efficiency, geographical location (to reduce latency) and user-specific preferences of the resources. The scheduling system also monitors the real-time status of resources, such as availability and load, to dynamically adjust resource allocation.

2.3.2 Scheduling and execution

1. Task queue and priority management:

All tasks are queued according to priority and submission time. The system processes the task queue according to preset or dynamically adjusted priority rules. Urgent or high-priority tasks can get a quick response, while long-term or cost-sensitive tasks may be executed during low-cost periods.

2. Fault tolerance and load balancing:

The dynamic resource allocation system includes a fault tolerance mechanism to ensure that even when some resources fail, tasks can be smoothly migrated to other healthy resources for continued execution. Load balancing technology ensures that no single resource is overloaded and optimizes the performance of the entire network by properly distributing task loads.

3. Monitoring and adjustment:

The system continuously monitors the execution status of all tasks and the operating status of resources. This includes real-time analysis of key performance indicators such as task progress and resource consumption. Based on this data, the system may automatically readjust resource allocation to optimize task execution efficiency and resource utilization.

2.3.3 User interaction and feedback

Transparent user interface: io.net provides an intuitive user interface that allows users to easily submit tasks, view task status, and adjust requirements or priorities. Feedback mechanism: Users can provide feedback on the results of task execution, and the system adjusts the resource allocation strategy for future tasks based on the feedback to better meet user needs.

3. System Architecture

3.1 IO Cloud

IO Cloud is designed to simplify the deployment and management of decentralized GPU clusters, providing machine learning engineers and developers with scalable and flexible access to GPU resources without significant hardware investment. This platform provides an experience similar to traditional cloud services, but with the advantages of a decentralized network.

Highlights:

Scalability and Affordability: Designed to be the most cost-effective GPU cloud, reducing AI/ML project costs by up to 90%. Integration with IO SDK: Enhances AI project performance through seamless integration, creating a unified high-performance environment. Global Coverage: Distributed GPU resources, optimized for machine learning services and inference, similar to CDN. RAY Framework Support: Scalable Python application development using the RAY distributed computing framework. Exclusive Features: Provides private access to the OpenAI ChatGPT plugin for easy deployment of training clusters. Crypto Mining Innovation: Seeks to revolutionize crypto mining by supporting the machine learning and AI ecosystem.

3.2 IO Worker

IO Worker is designed to simplify and optimize provisioning operations for WebApp users. This includes user account management, real-time activity monitoring, temperature and power consumption tracking, installation support, wallet management, security, and profitability analysis.

Highlights:

Worker Homepage: Provides a dashboard for real-time monitoring of connected devices, with the ability to delete and rename devices. Device Details Page: Displays comprehensive device analytics, including traffic, connection status, and work history. Earnings & Rewards Page: Tracks earnings and work history, with transaction details accessible on SOLSCAN. Add New Device Page: Simplifies the device connection process, enabling fast and easy integration.

3.3 IO Explorer

IO Explorer is designed as a comprehensive platform to provide users with deep insights into the operations of the io.net network, similar to how blockchain explorers provide transparency into blockchain transactions. Its main goal is to enable users to monitor, analyze, and understand the details of the GPU Cloud, ensuring full visibility into network activity, statistics, and transactions, while protecting the privacy of sensitive information.

Benefits:

Browser Home: Provides insights into provisioning, verified vendors, active hardware counts, and real-time market pricing. Cluster Page: Displays public information about clusters deployed in the network, along with real-time metrics and booking details. Device Page: Displays public details of devices connected to the network, providing real-time data and transaction tracking. Real-time Cluster Monitoring: Provides instant insights into cluster status, health, and performance, ensuring users have the latest information.

3.4 IO-SDK

IO-SDK is the foundational technology of Io.net, derived from a branch of Ray technology. It enables tasks to run in parallel and handle different languages, and is compatible with major machine learning (ML) frameworks, making IO.NET flexible and efficient for a variety of computing needs. This setup, coupled with a set of well-defined technologies, ensures that IO.NET Portal can meet today's needs and adapt to future changes.

Application of multi-layer architecture

· User interface: Serves as the visual front end for users, including the public website, customer area, and GPU provider area. The design is intuitive and user-friendly.

· Security layer: Ensures the integrity and security of the system, including network protection, user authentication, and activity logging.

· API layer: Serves as a communication center for websites, providers, and internal management, facilitating data exchange and operations.

· Backend layer: the core of the system, handling operations such as cluster/GPU management, customer interaction, and automatic scaling.

· Database layer: stores and manages data, with primary storage for structured data and cache for temporary data.

· Task layer: manages asynchronous communication and tasks, ensuring efficiency of execution and data flow.

· Infrastructure layer: infrastructure, including GPU pools, orchestration tools, and execution/ML tasks, equipped with powerful monitoring solutions.

3.5 IO Tunnels

Use reverse tunneling technology to create a secure connection from the client to the remote server, allowing engineers to bypass firewalls and NAT for remote access without complex configuration. Workflow: IO Worker connects to the intermediate server (io.net server). Then, the io.net server listens for connections from IO Worker and engineer machines, facilitating data exchange through reverse tunnels.

Application in io.net

Engineers connect to IO Workers through the io.net server, simplifying remote access and management without network configuration challenges. Benefits: Accessibility: Direct access to IO Workers, eliminating network barriers. Security: Ensure protected communications and maintain data privacy. Scalability and flexibility: Effectively manage multiple IO Workers in different environments.

3.6 IO Network

IO Network uses a mesh VPN architecture to provide ultra-low latency communication between antMiner nodes.

Mesh VPN Network:

Decentralized Connectivity: Unlike the traditional star model, mesh VPN directly connects nodes, providing enhanced redundancy, fault tolerance, and load distribution. Advantages: Strong resistance to node failures, strong scalability, low latency, and better traffic distribution.

Benefits of io.net:

Direct connections reduce latency and optimize application performance. No single point of failure, the network can still operate even if a single node fails. Enhanced user privacy by making data tracking and analysis more challenging. The addition of new nodes does not affect performance. Resource sharing and processing are more efficient between nodes.

4. $IO Tokens

4.1 Basic Framework of $IO Tokens

1. Fixed Supply:

The maximum supply of $IO tokens is fixed at 800 million. This supply is designed to ensure the stability of the token value and prevent inflation.

2. Distribution and Incentives:

Initially, 300 million $IO tokens will be issued. The remaining 500 million tokens will be issued as rewards to suppliers and their shareholders, a process that is expected to last for 20 years. Rewards are released hourly and follow a decreasing model (starting at 8% in the first year, decreasing by 1.02% per month, or about 12% per year) until the total issuance cap of 800 million tokens is reached.

3. Destruction Mechanism:

$IO uses a programmatic token destruction system, which uses the revenue generated by io.net from the IOG network to purchase and destroy $IO tokens. The destruction mechanism adjusts the amount destroyed based on the price of $IO to create deflationary pressure on the token.

4.2 Fees and Revenues

Usage Fees:

io.net charges users and suppliers a variety of fees, including reservation fees and payment fees when booking computing power. These fees are set to maintain the financial health of the network and support the market circulation of $IO.

Payment Fees:

A 2% fee is charged for payments made in USDC and no fee is charged for payments made in $IO.

Supplier Fees:

Similar to users, suppliers also pay fees when they receive payment, including booking fees and payment fees.

4.3 Ecosystem

GPU renters (also known as users), such as machine learning engineers who want to purchase GPU computing power on the IOG Network. These engineers can use $IO to deploy GPU clusters, cloud gaming instances, and build Unreal Engine 5 (and similar) pixel streaming applications. Users also include individual consumers who want to perform serverless model inference on BC8.ai and the hundreds of applications and models that io.net will host in the future. GPU owners (also known as suppliers), such as independent data centers, crypto mining farms, and professional miners, want to provide underutilized GPU computing power on the IOG Network and profit from it. IO coin holders (also known as the community) participate in providing cryptoeconomic security and incentives to coordinate mutual benefits and penalties between parties to promote the development and adoption of the network.

4.4 Specific Allocation

Community: 50% of the total allocation, this part of the token is mainly used to reward community members and incentivize platform participation and growth. R&D Ecosystem: 16%, used to support the platform's R&D activities and ecosystem construction, including partners and third-party developers. Initial Core Contributors: 11.3%, rewarding team members who made key contributions in the early stages of the platform. Early Backers: Seed: 12.5%, this part of the token is allocated to early seed investors to reward their trust and financial support for the project in its early stages. Early Backers: Series A: 10.2%, allocated to Series A investors in return for their investment of funds and resources in the early stages of the project development.

4.5 Halving Mechanism

2024-2025: During these two years, 6,000,000 $IO tokens will be released each year. 2026-2027: Starting in 2026, the annual release will be halved to 3,000,000 $IO tokens. 2028-2029: The release will continue to halve, with 1,500,000 $IO tokens released each year.

5. Team / Cooperation / Financing

io.net has a leadership team with diverse skills and experience, who have decades of experience in the technology field and have contributed to the company's success.

Tory Green is the COO of io.net, and was previously the COO of Hum Capital and the Director of Corporate Development and Strategy of Fox Mobile Group.

Ahmad Shadid is the founder and CEO of io.net, and was previously a quantitative system engineer at WhalesTrader.

Garrison Yang is the Chief Strategy Officer and Chief Marketing Officer of io.net, and was previously the Vice President of Growth and Strategy at Ava Labs. He graduated from the University of California, Santa Barbara with a degree in Environmental Health Engineering.

In March this year, io.net received a $30 million Series A financing round led by Hack VC, with participation from Multicoin Capital, 6th Man Ventures, M13, Delphi Digital, Solana Labs, Aptos Labs, Foresight Ventures, Longhash, SevenX, ArkStream, Animoca Brands, Continue Capital, MH Ventures and OKX, as well as industry leaders including Solana founder Anatoly Yakovenk, Aptos founders Mo Shaikh and Avery Ching, Animoca Brands' Yat Siu and Perlone Capital's Jin Kang.

6. Project Evaluation

6.1 Track Analysis

io.net is a decentralized computing network based on the Solana blockchain, focusing on providing powerful computing power by integrating underutilized GPU resources. This project is mainly in the following track areas:

1. Decentralized Computing

io.net has built a decentralized physical infrastructure network (Depin) that leverages GPU resources from different sources (e.g., independent data centers, crypto miners). This decentralized approach aims to optimize the utilization of computing resources and reduce costs while increasing accessibility and flexibility.

2. Cloud Computing

Despite its decentralized approach, io.net provides services similar to traditional cloud computing, such as GPU cluster management and scalability for machine learning tasks. io.net aims to create an experience similar to traditional cloud services, but leverages the advantages of decentralized networks to provide more efficient and cost-effective solutions.

3. Blockchain Applications

As a project based on blockchain technology, io.net uses the characteristics of blockchain, such as security and transparency, to manage resources and transactions in the network.

Projects similar to io.net in terms of functions and goals include:

Golem: It is also a decentralized computing network where users can rent or lease unused computing resources. Golem is committed to creating a global supercomputer. Render: It uses a decentralized network to provide graphics rendering services. Render uses blockchain technology to enable content creators to access more GPU resources, thereby accelerating the rendering process. iExec RLC: This project creates a decentralized market that allows users to rent out their computing resources. iExec supports various types of applications, including data-intensive applications and machine learning workloads, through blockchain technology.

6.2 Project Advantages

Scalability: io.net has designed a highly scalable platform specifically to meet customers' bandwidth needs and enable teams to easily scale workloads on GPU networks without large-scale adjustments. Batch Inference and Model Serving: The platform supports parallelized inference on data batches, allowing machine learning teams to deploy workflows on distributed GPU networks.

Parallel Training: To overcome memory limitations and sequential workflows, io.net leverages distributed computing libraries to parallelize training tasks across multiple devices. Parallel Hyperparameter Tuning: Leveraging the inherent parallelism of hyperparameter tuning experiments, io.net optimizes scheduling and search patterns. Reinforcement Learning (RL): Leveraging open source reinforcement learning libraries, io.net supports highly distributed RL workloads and provides a simple API.

Instant Accessibility: Unlike the long deployment of traditional cloud services, io.net Cloud provides instant access to GPU provisioning, enabling users to launch their projects in seconds.

Cost Efficiency: io.net is designed to be an affordable platform that is suitable for different categories of users. Currently, the platform is about 90% more cost-effective than competing services, providing significant savings for machine learning projects.

High Security and Reliability: The platform promises to provide best-in-class security, reliability, and technical support, ensuring a secure and stable environment for machine learning tasks. Ease of Implementation: io.net Cloud eliminates the complexity of building and managing infrastructure, enabling any developer and organization to seamlessly develop and scale AI applications.

6.3 Project Challenges

1. Technical Complexity and User Adoption

Challenges: While decentralized computing offers significant cost and efficiency advantages, the complexity of its technology can pose a large barrier to entry for non-technical users. Users need to understand how to operate a distributed network and how to effectively utilize distributed resources. Impact: This may limit the widespread adoption of the platform, especially among user groups that are less familiar with blockchain and distributed computing.

2. Cybersecurity and Data Privacy

Challenges: Although blockchain provides enhanced security and transparency, the openness of decentralized networks may make them more vulnerable to cyber attacks and data leaks. Impact: This requires io.net to continuously strengthen its security measures to ensure the confidentiality and integrity of user data and computing tasks, which is key to maintaining user trust and platform reputation.

3. Performance and Reliability

Challenges: Although io.net strives to provide efficient computing services through decentralized resources, coordinating between hardware resources in different geographical locations and of different quality may bring challenges in performance and reliability. Impact: Any performance issues caused by hardware mismatch or network latency may affect customer satisfaction and the overall effectiveness of the platform.

4. Scalability

Challenges: Although io.net has designed a highly scalable network, it is still a huge technical challenge to effectively manage and expand distributed resources around the world in actual operations. Impact: This requires continuous technical innovation and management improvements to keep the network stable and responsive in the face of rapidly growing user and computing needs.

5. Competition and Market Acceptance

Challenges: io.net is not without competition in the blockchain and decentralized computing market. Other platforms such as Golem, Render, and iExec are also providing similar services, and rapid changes in the market may quickly change the competitive situation. Impact: To remain competitive, io.net needs to continue to innovate and improve the uniqueness and value of its services to attract and retain users.

7. Conclusion

In summary, io.net has set a new benchmark in the field of modern cloud computing with its innovative decentralized computing network and blockchain-based architecture. By aggregating underutilized GPU resources around the world, io.net provides unprecedented computing power, flexibility, and cost-efficiency for machine learning and artificial intelligence applications. This platform not only makes it faster and more economical to deploy large-scale machine learning projects, but also provides strong security and scalable solutions for all types of users.

Faced with challenges such as technical complexity, network security, performance stability, and market competition, if IO.Net can overcome challenges and cultivate a vibrant ecosystem, it has the potential to fundamentally reshape the way we access and utilize computing power in the Web3 era. However, as with any emerging technology, it is important to realize that its long-term success will depend on continued development, adoption, and its ability to navigate the evolving landscape of blockchain-based infrastructure.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia