We find prodigy in crypto investment research
We find prodigy in crypto investment research
Subscribe to Brillian Research
Subscribe to Brillian Research
<100 subscribers
<100 subscribers
Share Dialog
Share Dialog
Fleek Network is a decentralized content and application delivery network (CDN) built by the Fleek team, specifically designed to accelerate the delivery of all Web3 content and applications. As an open source, trustless, censorship-resistant CDN, anyone can contribute bandwidth to the network by running a cache node. What makes Fleek Network unique is its raw neutrality: it accelerates the delivery of content regardless of its underlying storage protocol or provider. Fleek Network emphasizes transparency, every interaction on the network is publicly accessible, and all content is based on IPLD and internally addressable, thereby creating a public record of global broadband and content served, and smart contracts for The transparency that finance brings is comparable.
Fleek is a Web3 infrastructure platform that allows developers to host DApps, websites, or other front-ends, and utilize Web3 protocols to meet their storage and data needs, supporting protocols such as Fleek Network, IPFS, Arweave, Filecoin, POKT, ENS, and others.
Fleek initially collaborated with Web2 infrastructure providers like AWS and Cloudflare, while Fleek Network, a decentralized edge network launched by Fleek in 2023, aims to fill the gaps in the Web3 infrastructure stack.
According to available information, the team currently consists of 12 members. Founder Harrison Hines, besides Fleek, also served as CEO of Token Foundry. Co-founder and CTO Janison Sivarajah was formerly a software engineer at Token Foundry. The team is primarily composed of technical personnel.
Fleek has completed two rounds of financing:
[Seed Round] In October 2018, a seed round financing of $3.7 million was completed, with investments from Digital Currency Group, Coinbase Ventures, and Distributed Global.
[Series A] In December 2022, a Series A financing of $25 million was completed, led by Polychain Capital, with participation from Digital Currency Group, Protocol Labs, Arweave, The LAO, Coinbase Ventures, North Island Ventures, Distributed Global, Argonautic Ventures.
The goal of Fleek Network is to provide an efficient, trustless, and decentralized edge network, supporting a variety of edge services. Given the potential to build numerous edge services, the core purpose of Fleek Network is to serve as a foundational layer, allowing anyone to quickly develop and deploy new edge services without having to worry about network intelligent routing/work assignment, load balancing, or any other layer that is not a core feature/function of their service. Fleek Network achieves this by making the network core geographically aware and optimizing for speed.
The establishment of Fleek Network must be understood in the context of the evolution of the Web2 stack, which has undergone significant changes, shifting rapidly from "cloud" to "edge". This shift to the "edge" offers high security at a low cost, as well as substantial advantages in terms of performance and network latency. With the exponential growth of online gaming, and the rise of artificial intelligence, augmented reality, and virtual reality, the demand for low-latency optimized network infrastructure will only increase. Similar to the development of Web2, the trend in Web3 is towards modularity and composability, with an array of protocols and middleware with specialized functions emerging, providing specific services across the entire infrastructure stack. This has led to the decoupling of monolithic architecture, breaking it down into smaller, independently operable modules.
However, this trend in Web3 is still in its early stages, and the field is still rife with inefficient and duplicative work. Almost all Web3 protocols, middleware, services, and applications must choose between two suboptimal paths:
Leverage cloud infrastructure controlled by companies (such as AWS or Cloudflare) in their technology stack to achieve "Web2-like" performance (which is currently the demand of most developers and end-users), but this often introduces centralized issues into their system, significantly weakening their decentralized, censorship-resistant, and permissionless characteristics;
Attempt to build all of these performance optimizations into their network, economic model, nodes, or technology stack (such as specialized networks and peer-to-peer communications, geolocation and reputation-based routing, algorithms to decide which nodes should perform which tasks or provide which content, load balancing requests, etc.). However, for those projects attempting to build these functionalities on their own, this is not only extremely complex and time-consuming, but also inefficient.
To address the unmet needs of infrastructure in Web3, Fleek Network has emerged. Introducing a high-performance decentralized edge network as a shared infrastructure, orchestration, and performance layer into the Web3 stack is beneficial for nearly every Web3 project. It can help bring centralized performance to Web3, similar to the traditional web, without sacrificing the values of Web3. The shared edge infrastructure can also assist developers building Web3 infrastructure or middleware in reducing entry barriers and accelerating time to market. It does so by offloading parts of the stack to the decentralized edge, eliminating the need to build this functionality into their own networks. This enables developers to focus more of their time and resources on their unique products and services, while placing performance and latency optimization on the edge layer above, similar to modern networks. The foundation of the edge network is high performance, low latency, and geographically distributed load, characteristics that are challenging for most decentralized systems to handle at the base protocol layer. The Fleek Network addresses these performance requirements through the following technologies:
2.2.1 Geographical Awareness
Although the network does not have explicit geographical attributes, it can infer proximity relationships based on latency and hop count data (the number of intermediate nodes a data packet travels through from the source node to the destination node in a network. Each time a packet passes through an intermediate node, the hop count increases by one, typically used to measure the transmission distance or path length in a network). This also forms a part of the reputation system.
The three core elements influencing geographical awareness include:
Nodes with lower latency and fewer hops are considered to be closer to each other.
Each node regularly shares approximate information of its most popular (i.e., most frequently requested) content with the k nearest peer nodes.
A probabilistic data structure, the Cuckoo Filter, is adopted for approximate set representation. The Cuckoo Filter is a data structure based on cuckoo hashing, designed for efficient set membership tests, that is, to determine whether an element is a member of a set.
This approach enables nodes in the Fleek Network to efficiently distribute workloads among nearby or optimal nodes, achieving seamless integration with other system components.
2.2.2 Smart Routing and Work Allocation
Many decentralized systems operate like a leaderless team, where members work in an egalitarian manner. These systems rely on stake-weighted work allocation and commonly use gossip protocols for message passing and communication. The algorithms of these networks usually operate without any prior assumptions about the reliability or geographical location of other nodes in the network, similar to not considering the reliability or specific location of team members in advance. Even if data on reliability is collected at runtime, it is used to facilitate the system’s operation rather than to set rules.
In contrast to most other decentralized systems, the Fleek Network employs a unique setup:
The nodes in the network remain fixed throughout each epoch, with no new nodes joining until the next epoch begins.
Reputation statistics are stored and broadcasted in an application state accessible to all nodes, making the reliability of nodes quantifiable and transparent.
All nodes have the same stake amount, and work allocation is strictly based on geographical location and reputation (i.e., the stake amount does not influence the workload assigned to a node), abandoning stake-weighted work allocation.
Leveraging these additional assumptions can significantly optimize the performance of the network's gossip and broadcast layers, enabling more efficient information transfer and reducing latency by routing work to the node closest to the end user making the request. Essentially, Fleek Network utilizes this information and employs deterministic algorithms to create an efficient and fault-tolerant connectivity graph for the entire network. At the start of each new epoch, nodes in the network can converge on the same overall network connectivity graph, quickly identifying their position and tasks within the network.
2.2.3 Stateless Execution
By default, the state of Fleek Network's execution environment is locally bound to the node running the service, making Fleek Network an optimal choice for creating stateless computations (where each executed task or operation does not depend on the state of previous tasks or operations) and for performing operations that can benefit the end user.
Maintaining a globally stateless (locally bounded) execution environment helps the network maintain maximum performance and low latency. However, a certain degree of state information can be maintained on individual nodes to facilitate local tasks and operations. An additional benefit of this approach is that it allows the system to randomly redistribute tasks to different computers at the end of each epoch, ensuring that no single computer becomes too important or controls too many resources. This maintains the decentralized nature of the system and significantly reduces the possibility of collusion or malicious attacks between nodes. This approach not only maintains the speed and security of the system but also allows the system to self-optimize, ensuring that the most suitable computers are found to perform different tasks at each stage.
2.2.4 VM-Less Core
Although developers can create or use virtual machines (VMs) at the service layer of Fleek Network, the core protocol of Fleek Network itself does not rely on virtual machines. The core design of not using virtual machines makes services more efficient in utilizing node resources and reduces restrictions on developers. For example, some non-EVM protocols, in order to achieve compatibility with EVM, run an EVM on their own native virtual machine, requiring two virtual machines to work together, leading to inefficiency. In contrast, the Fleek Network core protocol layer does not use VMs, while various virtual machines can be directly built as services, directly utilizing the resources of edge nodes, eliminating the unnecessary performance loss associated with running two layers of VMs.
2.2.5 Built-In (Externally Powered) File System
Fleek Network has built a built-in file system utilizing multiple external decentralized storage protocols (such as Filecoin, Arweave, IPFS, etc.), achieving modularity and thereby keeping network nodes lightweight, optimizing performance, and reducing latency. This also decreases the dependency risk for developers when it comes to storage choices. Services running on Fleek Network can efficiently access shared content through an SDK. While Fleek Network stores limited state information within the network, all other data storage needs depend on the built-in storage options or other external options chosen by the service.
2.2.6 Content-Addressable Core / Incremental Content Retrieval and Verification
Fleek Network efficiently implements content identification and verification using the Blake3 hash algorithm, with the entire network operating on a content-addressable basis. It also uses a Distributed Hash Table (DHT) to store mappings from immutable data pointers or IPLD (InterPlanetary Linked Data) to their corresponding Blake3 hash values.
Serving as a resource-rich caching layer, Fleek Network enhances the performance of many Web3 protocols and middleware services in data retrieval. It takes into account the geographical distribution of nodes and the popularity of files/data in specific regions, ensuring efficient data retrieval by replicating cached data between nearby nodes.
At the same time, the tree structure of Blake3 allows for the effective use of verifiable content streams. Considering the high-performance requirements of Fleek Network, the protocol has developed an alternative to a library named Bao, which has already been implemented by the authors of Blake3.
Incremental verification refers to the process of verifying only the changed parts of data or system state when alterations occur, rather than the entire system or dataset. This method aims to enhance efficiency and reduce the time and computational resources required for verification. The alternative library developed by Fleek Network utilizes precomputed data trees with larger block sizes and provides utility functions to efficiently prune these trees and create minimal subtrees while streaming content blocks. This approach enables significant performance optimization, making the cost of incremental verification for each node almost negligible.
Fleek Network is a PoS Ethereum sidechain that possesses its own edge node network, utilizing Ethereum to implement FLK tokens (ERC-20), staking, payments, governance, and other economic aspects of the protocol.
Node operators are required to stake FLK tokens (and possibly ETH through EigenLayer) to perform work on the network, providing resources such as bandwidth and CPU to the network. These resources are packaged into commodities and priced in USD at the network level. Developers, service creators, and customers use stablecoins to pay for the resources and commodities utilized on the network.
Clients request to run specific services, and nodes operate these services and perform work to earn fees. Fleek Network employs a combination of SNARKs (Succinct Non-Interactive Argument of Knowledge), Narwhal and Bullshark consensus, as well as other cryptographic, economic incentives and safeguards to create a trustless, decentralized, and sustainable environment.
2.3.1 Succinct Chain State
At the core of the protocol, only the following are kept in the state and all nodes need to replicate this data synchronously:
token balance ($FLK and stablecoins)
pledge information
node reputation
data about how much work the node has performed in a given epoch
2.3.2 Narwhal & Bullshark Consensus
To efficiently order transactions and achieve consensus, Fleek Network employs the Narwhal scheme (an open-source solution from Meta and MystenLabs), a DAG-based efficient mempool, in collaboration with Bullshark. This allows for consensus on global ordering to be achieved without network overhead.
In this network, not every node needs to participate in the ordering process; instead, a committee-based model is adopted. Any node with a qualifying stake is eligible to be part of the consensus committee. A new committee member is selected from a subset of participating nodes at the end of each epoch, using the decentralized random beacon present in the network. This committee, responsible for overseeing the transaction ordering process, passes its responsibilities to the next new committee, and so on. This regular rotation of committee members reduces the risk of collusion and ensures the robustness and reliability of the consensus mechanism. Other nodes in the network focus on processing work, batch-confirming transactions, and submitting the results to the committee for ordering and consensus.
2.3.3 Delivery Acknowledgement SNARKs
Narwhal/Bullshark primarily handles the ordering of a batch of "Delivery Acknowledgements." Nodes collect these acknowledgements while processing requests, and upon submission, the nodes are rewarded at the end of a cycle.
A "Delivery Acknowledgement" is a message signed by the client, verifying that the node has successfully delivered the task to the client. These acknowledgements are immediately finalized locally, and cannot be revoked by the client.
Once a node receives a "Delivery Acknowledgement" containing information about the resources provided to the client, it can add it to its local pool of acknowledgements and periodically send these messages to consensus (and hence to every other node) to claim fees. Once the acknowledgement is received, the client’s prepaid stablecoin balance is adjusted accordingly. Throughout a cycle, the deducted amounts from all clients are transferred to a payment pool and fairly distributed based on the work performed by nodes in that cycle.
This practice of periodically submitting "Delivery Acknowledgements" enables Fleek Network to leverage recursive SNARK proofs to aggregate these acknowledgements, eliminating the need to submit each transaction individually, thereby reducing network costs of broadcasting batches and significantly decreasing the computational power required to individually verify all the delivery acknowledgements.
2.3.4 Performance-Based Reputation System
The Fleek Network has implemented a reputation system based on mutual scoring, which determines the overall reputation score of each node in the network by collecting scores over a period of time and running an aggregation algorithm at the end of each cycle. The network draws inspiration from the EigenTrust algorithm, introducing the concept of transitive trust, meaning that nodes tend to trust their peers with positive interactions and higher scores, and weight the local trust values assigned based on the reputation of these peers.
To mitigate the impact of dishonest nodes and prevent Sybil attacks (i.e., manipulating the rating system by creating a large number of fake nodes), the network considers the reputation of nodes when allocating scores and further reduces the impact of false reports by performing outlier removal. Acquiring a high reputation score takes a significant amount of time and is calculated based on the exponential weighted moving average of the node's performance over all past periods. For specific calculation methods, please refer to: https://fleek.network/whitepaper.pdf (Appendix A).
In addition, reputation data can also be obtained by evaluating the service interactions between nodes, further enriching and optimizing the task allocation and traffic management in the network. These reputation data become a reliable source for optimizing task allocation and traffic management in the network, providing support for the network's efficient operation.
2.3.5 Roles in the Network
Clients - Developers - End Users - Node Operators
Clients are users who consume data from the network, such as developers interested in using CDNs in applications, media publishers who want to accelerate access to media resources, and so on.
Developers are users who provide business logic and end-to-end experience for end users through applications or services, such as protocol development contributors, service creators, etc.
End users are individuals to whom data or computational outputs are ultimately delivered, including static assets, server-side rendering, etc. Service outputs are targeted at end users, but the service fees are paid by the clients managing the applications or services that interact with the end users.
Node operators are responsible for building, configuring, installing, or maintaining nodes on servers.
Services on the Fleek Network are modular parts of edge-supportive software, running on edge nodes and leveraging the capabilities of these nodes to provide unique functionalities to their end users. These services can be seen as the user interface of the network.
Third-party developers have the autonomy to deploy decentralized Web services supporting edge computing on the network. These services run in an isolated sandbox environment, as mentioned previously in the VM-Less Core section. Initially, these core services will be statically linked to the node's binary files, but there are future plans to support dynamic loading of services, allowing developers to write and deploy services in any system-level programming language during node runtime.
From a permissions perspective, services can only access specific resources through the SDK, including file storage, cryptographic primitives, and metrics reporting APIs. Services cannot access other parts of the application or process, interact with other services (as they are mutually isolated), or directly access the edge node’s hardware or operating system kernel. Given the open and permissionless nature of the system, the code of the services will be managed by their respective developers in their code repositories, similar to the management of smart contracts.
A variety of edge services could be built on the Fleek Network. Here are some examples of traditional Web/edge services that could be developed:
● Hosting ● CDN ● Edgecompute/Serverless ●SSR/ISR(IncrementalStaticRegeneration)
● ImageOptimization ● Loadbalancing ●Gridcomputing ● NamingServices ● Analytics
● Databases(Document,KV,CRDT) ● Containerorchestration ●Queues/Events ● StepFunctions
2.4.1 Specific Use Cases for Web3 Services
Decentralized CDN: While most projects today use Cloudflare at the front of their stack for performance optimization and latency reduction, with the launch of Fleek Network, web3 projects can use decentralized CDN services as a direct alternative, gaining the advantages of web3 without sacrificing performance or cost. The decentralized CDN tracks the requests it serves, charges clients, and rewards nodes based on the bandwidth served. The reputation system tracks nodes that provide good CDN services, determining routing based on this.
Decentralized Website/App Hosting: Utilizes block storage and IPFS content addressing to track applications deployed with additional metadata. It essentially serves as the storage layer for frontends, similar to how platforms like S3 or Netlify operate. Hosting capabilities are further expanded through the use of CDN and compute services for server-side rendering. The costs for storage or processing for hosting and SSR can be directly charged to the client, with rewards distributed to node operators.
Decentralized IPFS Pinning: IPFS is a decentralized file storage and sharing network. Given the way Fleek Network operates, all files and content are addressed through Content Identifiers (CIDs), and the mapping from CID to source file is permanently stored. This means that even if a file disappears from the IPFS network, it can still be retrieved by the Fleek Network as long as it exists on at least one source. Moreover, Fleek Network integrates decentralized storage protocols (such as Filecoin and Arweave) and CDN services, providing a high-performance and economical IPFS pinning service that holds advantages over traditional centralized IPFS pinning services. Storing data on decentralized storage protocols is generally more economical than on cloud platforms, providing better data security and availability guarantees.
Other services include: Decentralized Edge Computing (based on Web3 protocols); Blockchain Snapshot as a Service; zkVM, EVM, and other VMs as a Service; Alternative Rollup Sequencer; Temporary Rollups; Edge Proofs of Presence; Verifiable Randomness; Web3 Querying/Events.
As of October 21, 2023, Fleek Network has submitted a total of 4578 code submissions. As the testing network was just opened on October 19, we can expect its future performance.
The testnet is divided into six phases, "0-5," and we are currently in phase 1. The tasks and progress of each phase are as follows:
Phase 0 (September 5): Launch of the node, initial network and node testing (performance, hardware specifications, clusters, cost, metrics, etc.)
Phase 1 (Ongoing): Launch of SDK/services, introduction of SDK, and testing of building and running services on the network, as well as some optimizations based on phase 0
Phase 2 (October): Launch of initial economics, introduction and testing of a more concrete version of economic algorithms, including staking, pricing, and using other elements/situations (valueless) tokens tested, as well as some optimizations based on phase 1
Phase 3 (November): Launch of Layer 2 contracts, introduction of test versions of various aspects of the protocol running on Ethereum L2 (staking, deposits, token contracts, communication between L2/FN, etc.)
Phase 4 (December): Final launch, introduction of the final form of the first generation network based on all the data/feedback and optimizations from all phases, and allowing testing of the actual mainnet environment
Phase 5 (Q1 2024): Mainnet launch
In the current phase 1 of the testnet, users can follow the guidelines to become node operators. For the specific process, please refer to: https://blog.fleek.network/post/fleek-network-testnet-phase-1/
Participants in the testnet can share 1% of $FLK tokens, and the reputation of nodes established on the testnet can be migrated to the mainnet. In addition, the foundation may provide additional grants to specific node operators, service builders, or users during the testnet period.
$FLK is the utility token of Fleek Network, issued as an ERC-20 token on Ethereum. It also serves as a governance token, with a maximum supply of 100 million. The specific allocation is as follows:
The primary income for token holders comes from a basket of stablecoins. For example, node operators settle their work rewards in stablecoins (pegged to the US dollar), ensuring a stable and predictable income stream.
$FLK is a staking token and an indispensable part of Fleek Network. Nodes running Fleek Network client software need to stake a certain amount of FLK to participate in the network and have the opportunity to earn fees by providing work. Below are some expected features of the $FLK token:
Work is allocated among nodes based on location (latency) and reputation (performance).
Resources on the network are priced and paid for by users in stablecoins pegged to the US dollar. Resource pricing occurs at the network level.
Nodes are only rewarded for the resources used/consumed in relation to the work they perform (including FLK rewards).
FLK rewards are paid per epoch (approximately 24 hours). $FLK rewards for each epoch are distributed proportionally among nodes that performed work during that period, calculated as the dollar income earned by each node compared to the total income earned during that period.
20% of the total supply of $FLK tokens is allocated for staking/rewards, using an inflationary mechanism to provide additional rewards to stakeholders. The actual reward/inflation rate will be controlled and updated algorithmically based on network usage, the market price of $FLK, and other factors.
Interestingly, the $FLK rewards for nodes will be dynamically adjusted. The system will take into account both the US dollar stablecoin fees earned by the node and the $FLK rewards. This means that if network usage/US dollar income increases, the $FLK rewards may decrease. Conversely, if the time-weighted average market price of $FLK rises, the absolute number of expected $FLK rewards for the node may decrease, and vice versa. With this design, the long-term income of nodes is expected to be more consistent.
Filecoin： A leader in the Web3 storage space, Filecoin is a decentralized storage network that transforms cloud storage into an algorithmic market. Miners earn Filecoin by providing storage to clients, who in turn spend Filecoin to hire miners for data storage or distribution. The project was created by the Protocol Labs team, which also founded IPFS (InterPlanetary File System). Filecoin employs a robust consensus mechanism based on Proof of Replication and Proof of Spacetime protocols. These proofs allow miners to verify the exact amount of storage space provided and the duration of data storage, ensuring data integrity and security. With the launch of Filecoin Virtual Machine (FVM), the project's applications extend beyond data storage and retrieval, aiming to restructure the underlying architecture of the internet. This results in a high overlap with the services provided by Fleek Network.
Storj：A decentralized, blockchain-based distributed cloud storage system, Storj offers functionalities similar to Filecoin but with a more focused application scope. While it doesn't have the groundbreaking innovation found in Filecoin and IPFS, Storj's advantages become clear when compared to traditional centralized storage solutions.
Arweave: Unlike Filecoin and similar platforms that use traditional blockchain architecture for transaction and off-chain storage, Arweave employs on-chain storage. Each node stores only a portion of the blocks, avoiding the need to sync the entire blockchain. Blocks in Arweave not only link to the previous block, like traditional blockchain protocols, but also point to a random previous recall chunk, creating a spool-like block structure called blockweave. Additionally, there is no size limit for each block. Arweave has a well-developed ecosystem, including infrastructure, middleware, and applications, as well as its own NFT standard that addresses the issue of off-chain NFT metadata storage, providing a comprehensive value proposition and application scenarios.
Competitive Risk: Being relatively young compared to established storage leaders, Fleek Network may face challenges in customer acquisition due to the first-mover advantages of its competitors.
Economic Model Risk: The network adopts a dynamic model to adjust rewards for nodes, with critical parameters determined by network proposals. This could lead to proposal rejections due to conflicts of interest, potentially destabilizing the network.
Fleek Network is poised to play a significant role in decentralizing CDN and edge layers, allowing users to leverage its caching and computational capabilities. Beyond storage and data retrieval, Fleek Network can be used for computational tasks and operations, such as server-side rendering and data processing. A major advantage of Fleek Network is its support for various infrastructures, products, and features available on the Fleek platform, enhancing the developer experience. Backed by a mature platform and leading investment institutions, Fleek Network holds a competitive edge in the market.
In terms of tokenomics design, the model of Fleek Network is yet to be finalized, with a focus on value capture and ecosystem incentives rather than a large share for the team.
Considering its competitiveness, economic model, and associated risks, Fleek Network is rated 4 stars.
— Reference —
Website：https://fleek.network/
Project Docs：https://docs.fleek.network/docs/
Whitepaper：https://fleek.network/whitepaper.pdf
— Investment Risks and Disclaimer —
This report is provided for reference only and should not be used as a basis for investment advice or decisions. Investors should not make any investment actions based on this content. Brillian Research and the authors of the report assume no responsibility for the investment consequences arising from this.
The report is effective as of the date of its publication. However, as time progresses, market and economic conditions may fluctuate, and the information in the report may not be able to reflect these changes in real-time. The graphics, charts, and other visual presentation materials provided in this report are for reference only. They do not imply that Brillian Research will guide investment decisions, nor can they comprehensively display the various factors that need to be considered in investment decisions.
Some views or assumptions in the report may reflect Brillian Research’s expectations or predictions for the future. However, numerous known and unknown risks and uncertainties may cause actual conditions to deviate from the views or assumptions in this report.
Any expectations, predictions, or estimates in this text are based on specific assumptions and are subject to certain uncertainties. These forward-looking statements may be affected by inaccurate assumptions or unpredictable risk factors and may differ from actual conditions. Readers should note that some or all of the forward-looking assumptions may not be realized.
Fleek Network is a decentralized content and application delivery network (CDN) built by the Fleek team, specifically designed to accelerate the delivery of all Web3 content and applications. As an open source, trustless, censorship-resistant CDN, anyone can contribute bandwidth to the network by running a cache node. What makes Fleek Network unique is its raw neutrality: it accelerates the delivery of content regardless of its underlying storage protocol or provider. Fleek Network emphasizes transparency, every interaction on the network is publicly accessible, and all content is based on IPLD and internally addressable, thereby creating a public record of global broadband and content served, and smart contracts for The transparency that finance brings is comparable.
Fleek is a Web3 infrastructure platform that allows developers to host DApps, websites, or other front-ends, and utilize Web3 protocols to meet their storage and data needs, supporting protocols such as Fleek Network, IPFS, Arweave, Filecoin, POKT, ENS, and others.
Fleek initially collaborated with Web2 infrastructure providers like AWS and Cloudflare, while Fleek Network, a decentralized edge network launched by Fleek in 2023, aims to fill the gaps in the Web3 infrastructure stack.
According to available information, the team currently consists of 12 members. Founder Harrison Hines, besides Fleek, also served as CEO of Token Foundry. Co-founder and CTO Janison Sivarajah was formerly a software engineer at Token Foundry. The team is primarily composed of technical personnel.
Fleek has completed two rounds of financing:
[Seed Round] In October 2018, a seed round financing of $3.7 million was completed, with investments from Digital Currency Group, Coinbase Ventures, and Distributed Global.
[Series A] In December 2022, a Series A financing of $25 million was completed, led by Polychain Capital, with participation from Digital Currency Group, Protocol Labs, Arweave, The LAO, Coinbase Ventures, North Island Ventures, Distributed Global, Argonautic Ventures.
The goal of Fleek Network is to provide an efficient, trustless, and decentralized edge network, supporting a variety of edge services. Given the potential to build numerous edge services, the core purpose of Fleek Network is to serve as a foundational layer, allowing anyone to quickly develop and deploy new edge services without having to worry about network intelligent routing/work assignment, load balancing, or any other layer that is not a core feature/function of their service. Fleek Network achieves this by making the network core geographically aware and optimizing for speed.
The establishment of Fleek Network must be understood in the context of the evolution of the Web2 stack, which has undergone significant changes, shifting rapidly from "cloud" to "edge". This shift to the "edge" offers high security at a low cost, as well as substantial advantages in terms of performance and network latency. With the exponential growth of online gaming, and the rise of artificial intelligence, augmented reality, and virtual reality, the demand for low-latency optimized network infrastructure will only increase. Similar to the development of Web2, the trend in Web3 is towards modularity and composability, with an array of protocols and middleware with specialized functions emerging, providing specific services across the entire infrastructure stack. This has led to the decoupling of monolithic architecture, breaking it down into smaller, independently operable modules.
However, this trend in Web3 is still in its early stages, and the field is still rife with inefficient and duplicative work. Almost all Web3 protocols, middleware, services, and applications must choose between two suboptimal paths:
Leverage cloud infrastructure controlled by companies (such as AWS or Cloudflare) in their technology stack to achieve "Web2-like" performance (which is currently the demand of most developers and end-users), but this often introduces centralized issues into their system, significantly weakening their decentralized, censorship-resistant, and permissionless characteristics;
Attempt to build all of these performance optimizations into their network, economic model, nodes, or technology stack (such as specialized networks and peer-to-peer communications, geolocation and reputation-based routing, algorithms to decide which nodes should perform which tasks or provide which content, load balancing requests, etc.). However, for those projects attempting to build these functionalities on their own, this is not only extremely complex and time-consuming, but also inefficient.
To address the unmet needs of infrastructure in Web3, Fleek Network has emerged. Introducing a high-performance decentralized edge network as a shared infrastructure, orchestration, and performance layer into the Web3 stack is beneficial for nearly every Web3 project. It can help bring centralized performance to Web3, similar to the traditional web, without sacrificing the values of Web3. The shared edge infrastructure can also assist developers building Web3 infrastructure or middleware in reducing entry barriers and accelerating time to market. It does so by offloading parts of the stack to the decentralized edge, eliminating the need to build this functionality into their own networks. This enables developers to focus more of their time and resources on their unique products and services, while placing performance and latency optimization on the edge layer above, similar to modern networks. The foundation of the edge network is high performance, low latency, and geographically distributed load, characteristics that are challenging for most decentralized systems to handle at the base protocol layer. The Fleek Network addresses these performance requirements through the following technologies:
2.2.1 Geographical Awareness
Although the network does not have explicit geographical attributes, it can infer proximity relationships based on latency and hop count data (the number of intermediate nodes a data packet travels through from the source node to the destination node in a network. Each time a packet passes through an intermediate node, the hop count increases by one, typically used to measure the transmission distance or path length in a network). This also forms a part of the reputation system.
The three core elements influencing geographical awareness include:
Nodes with lower latency and fewer hops are considered to be closer to each other.
Each node regularly shares approximate information of its most popular (i.e., most frequently requested) content with the k nearest peer nodes.
A probabilistic data structure, the Cuckoo Filter, is adopted for approximate set representation. The Cuckoo Filter is a data structure based on cuckoo hashing, designed for efficient set membership tests, that is, to determine whether an element is a member of a set.
This approach enables nodes in the Fleek Network to efficiently distribute workloads among nearby or optimal nodes, achieving seamless integration with other system components.
2.2.2 Smart Routing and Work Allocation
Many decentralized systems operate like a leaderless team, where members work in an egalitarian manner. These systems rely on stake-weighted work allocation and commonly use gossip protocols for message passing and communication. The algorithms of these networks usually operate without any prior assumptions about the reliability or geographical location of other nodes in the network, similar to not considering the reliability or specific location of team members in advance. Even if data on reliability is collected at runtime, it is used to facilitate the system’s operation rather than to set rules.
In contrast to most other decentralized systems, the Fleek Network employs a unique setup:
The nodes in the network remain fixed throughout each epoch, with no new nodes joining until the next epoch begins.
Reputation statistics are stored and broadcasted in an application state accessible to all nodes, making the reliability of nodes quantifiable and transparent.
All nodes have the same stake amount, and work allocation is strictly based on geographical location and reputation (i.e., the stake amount does not influence the workload assigned to a node), abandoning stake-weighted work allocation.
Leveraging these additional assumptions can significantly optimize the performance of the network's gossip and broadcast layers, enabling more efficient information transfer and reducing latency by routing work to the node closest to the end user making the request. Essentially, Fleek Network utilizes this information and employs deterministic algorithms to create an efficient and fault-tolerant connectivity graph for the entire network. At the start of each new epoch, nodes in the network can converge on the same overall network connectivity graph, quickly identifying their position and tasks within the network.
2.2.3 Stateless Execution
By default, the state of Fleek Network's execution environment is locally bound to the node running the service, making Fleek Network an optimal choice for creating stateless computations (where each executed task or operation does not depend on the state of previous tasks or operations) and for performing operations that can benefit the end user.
Maintaining a globally stateless (locally bounded) execution environment helps the network maintain maximum performance and low latency. However, a certain degree of state information can be maintained on individual nodes to facilitate local tasks and operations. An additional benefit of this approach is that it allows the system to randomly redistribute tasks to different computers at the end of each epoch, ensuring that no single computer becomes too important or controls too many resources. This maintains the decentralized nature of the system and significantly reduces the possibility of collusion or malicious attacks between nodes. This approach not only maintains the speed and security of the system but also allows the system to self-optimize, ensuring that the most suitable computers are found to perform different tasks at each stage.
2.2.4 VM-Less Core
Although developers can create or use virtual machines (VMs) at the service layer of Fleek Network, the core protocol of Fleek Network itself does not rely on virtual machines. The core design of not using virtual machines makes services more efficient in utilizing node resources and reduces restrictions on developers. For example, some non-EVM protocols, in order to achieve compatibility with EVM, run an EVM on their own native virtual machine, requiring two virtual machines to work together, leading to inefficiency. In contrast, the Fleek Network core protocol layer does not use VMs, while various virtual machines can be directly built as services, directly utilizing the resources of edge nodes, eliminating the unnecessary performance loss associated with running two layers of VMs.
2.2.5 Built-In (Externally Powered) File System
Fleek Network has built a built-in file system utilizing multiple external decentralized storage protocols (such as Filecoin, Arweave, IPFS, etc.), achieving modularity and thereby keeping network nodes lightweight, optimizing performance, and reducing latency. This also decreases the dependency risk for developers when it comes to storage choices. Services running on Fleek Network can efficiently access shared content through an SDK. While Fleek Network stores limited state information within the network, all other data storage needs depend on the built-in storage options or other external options chosen by the service.
2.2.6 Content-Addressable Core / Incremental Content Retrieval and Verification
Fleek Network efficiently implements content identification and verification using the Blake3 hash algorithm, with the entire network operating on a content-addressable basis. It also uses a Distributed Hash Table (DHT) to store mappings from immutable data pointers or IPLD (InterPlanetary Linked Data) to their corresponding Blake3 hash values.
Serving as a resource-rich caching layer, Fleek Network enhances the performance of many Web3 protocols and middleware services in data retrieval. It takes into account the geographical distribution of nodes and the popularity of files/data in specific regions, ensuring efficient data retrieval by replicating cached data between nearby nodes.
At the same time, the tree structure of Blake3 allows for the effective use of verifiable content streams. Considering the high-performance requirements of Fleek Network, the protocol has developed an alternative to a library named Bao, which has already been implemented by the authors of Blake3.
Incremental verification refers to the process of verifying only the changed parts of data or system state when alterations occur, rather than the entire system or dataset. This method aims to enhance efficiency and reduce the time and computational resources required for verification. The alternative library developed by Fleek Network utilizes precomputed data trees with larger block sizes and provides utility functions to efficiently prune these trees and create minimal subtrees while streaming content blocks. This approach enables significant performance optimization, making the cost of incremental verification for each node almost negligible.
Fleek Network is a PoS Ethereum sidechain that possesses its own edge node network, utilizing Ethereum to implement FLK tokens (ERC-20), staking, payments, governance, and other economic aspects of the protocol.
Node operators are required to stake FLK tokens (and possibly ETH through EigenLayer) to perform work on the network, providing resources such as bandwidth and CPU to the network. These resources are packaged into commodities and priced in USD at the network level. Developers, service creators, and customers use stablecoins to pay for the resources and commodities utilized on the network.
Clients request to run specific services, and nodes operate these services and perform work to earn fees. Fleek Network employs a combination of SNARKs (Succinct Non-Interactive Argument of Knowledge), Narwhal and Bullshark consensus, as well as other cryptographic, economic incentives and safeguards to create a trustless, decentralized, and sustainable environment.
2.3.1 Succinct Chain State
At the core of the protocol, only the following are kept in the state and all nodes need to replicate this data synchronously:
token balance ($FLK and stablecoins)
pledge information
node reputation
data about how much work the node has performed in a given epoch
2.3.2 Narwhal & Bullshark Consensus
To efficiently order transactions and achieve consensus, Fleek Network employs the Narwhal scheme (an open-source solution from Meta and MystenLabs), a DAG-based efficient mempool, in collaboration with Bullshark. This allows for consensus on global ordering to be achieved without network overhead.
In this network, not every node needs to participate in the ordering process; instead, a committee-based model is adopted. Any node with a qualifying stake is eligible to be part of the consensus committee. A new committee member is selected from a subset of participating nodes at the end of each epoch, using the decentralized random beacon present in the network. This committee, responsible for overseeing the transaction ordering process, passes its responsibilities to the next new committee, and so on. This regular rotation of committee members reduces the risk of collusion and ensures the robustness and reliability of the consensus mechanism. Other nodes in the network focus on processing work, batch-confirming transactions, and submitting the results to the committee for ordering and consensus.
2.3.3 Delivery Acknowledgement SNARKs
Narwhal/Bullshark primarily handles the ordering of a batch of "Delivery Acknowledgements." Nodes collect these acknowledgements while processing requests, and upon submission, the nodes are rewarded at the end of a cycle.
A "Delivery Acknowledgement" is a message signed by the client, verifying that the node has successfully delivered the task to the client. These acknowledgements are immediately finalized locally, and cannot be revoked by the client.
Once a node receives a "Delivery Acknowledgement" containing information about the resources provided to the client, it can add it to its local pool of acknowledgements and periodically send these messages to consensus (and hence to every other node) to claim fees. Once the acknowledgement is received, the client’s prepaid stablecoin balance is adjusted accordingly. Throughout a cycle, the deducted amounts from all clients are transferred to a payment pool and fairly distributed based on the work performed by nodes in that cycle.
This practice of periodically submitting "Delivery Acknowledgements" enables Fleek Network to leverage recursive SNARK proofs to aggregate these acknowledgements, eliminating the need to submit each transaction individually, thereby reducing network costs of broadcasting batches and significantly decreasing the computational power required to individually verify all the delivery acknowledgements.
2.3.4 Performance-Based Reputation System
The Fleek Network has implemented a reputation system based on mutual scoring, which determines the overall reputation score of each node in the network by collecting scores over a period of time and running an aggregation algorithm at the end of each cycle. The network draws inspiration from the EigenTrust algorithm, introducing the concept of transitive trust, meaning that nodes tend to trust their peers with positive interactions and higher scores, and weight the local trust values assigned based on the reputation of these peers.
To mitigate the impact of dishonest nodes and prevent Sybil attacks (i.e., manipulating the rating system by creating a large number of fake nodes), the network considers the reputation of nodes when allocating scores and further reduces the impact of false reports by performing outlier removal. Acquiring a high reputation score takes a significant amount of time and is calculated based on the exponential weighted moving average of the node's performance over all past periods. For specific calculation methods, please refer to: https://fleek.network/whitepaper.pdf (Appendix A).
In addition, reputation data can also be obtained by evaluating the service interactions between nodes, further enriching and optimizing the task allocation and traffic management in the network. These reputation data become a reliable source for optimizing task allocation and traffic management in the network, providing support for the network's efficient operation.
2.3.5 Roles in the Network
Clients - Developers - End Users - Node Operators
Clients are users who consume data from the network, such as developers interested in using CDNs in applications, media publishers who want to accelerate access to media resources, and so on.
Developers are users who provide business logic and end-to-end experience for end users through applications or services, such as protocol development contributors, service creators, etc.
End users are individuals to whom data or computational outputs are ultimately delivered, including static assets, server-side rendering, etc. Service outputs are targeted at end users, but the service fees are paid by the clients managing the applications or services that interact with the end users.
Node operators are responsible for building, configuring, installing, or maintaining nodes on servers.
Services on the Fleek Network are modular parts of edge-supportive software, running on edge nodes and leveraging the capabilities of these nodes to provide unique functionalities to their end users. These services can be seen as the user interface of the network.
Third-party developers have the autonomy to deploy decentralized Web services supporting edge computing on the network. These services run in an isolated sandbox environment, as mentioned previously in the VM-Less Core section. Initially, these core services will be statically linked to the node's binary files, but there are future plans to support dynamic loading of services, allowing developers to write and deploy services in any system-level programming language during node runtime.
From a permissions perspective, services can only access specific resources through the SDK, including file storage, cryptographic primitives, and metrics reporting APIs. Services cannot access other parts of the application or process, interact with other services (as they are mutually isolated), or directly access the edge node’s hardware or operating system kernel. Given the open and permissionless nature of the system, the code of the services will be managed by their respective developers in their code repositories, similar to the management of smart contracts.
A variety of edge services could be built on the Fleek Network. Here are some examples of traditional Web/edge services that could be developed:
● Hosting ● CDN ● Edgecompute/Serverless ●SSR/ISR(IncrementalStaticRegeneration)
● ImageOptimization ● Loadbalancing ●Gridcomputing ● NamingServices ● Analytics
● Databases(Document,KV,CRDT) ● Containerorchestration ●Queues/Events ● StepFunctions
2.4.1 Specific Use Cases for Web3 Services
Decentralized CDN: While most projects today use Cloudflare at the front of their stack for performance optimization and latency reduction, with the launch of Fleek Network, web3 projects can use decentralized CDN services as a direct alternative, gaining the advantages of web3 without sacrificing performance or cost. The decentralized CDN tracks the requests it serves, charges clients, and rewards nodes based on the bandwidth served. The reputation system tracks nodes that provide good CDN services, determining routing based on this.
Decentralized Website/App Hosting: Utilizes block storage and IPFS content addressing to track applications deployed with additional metadata. It essentially serves as the storage layer for frontends, similar to how platforms like S3 or Netlify operate. Hosting capabilities are further expanded through the use of CDN and compute services for server-side rendering. The costs for storage or processing for hosting and SSR can be directly charged to the client, with rewards distributed to node operators.
Decentralized IPFS Pinning: IPFS is a decentralized file storage and sharing network. Given the way Fleek Network operates, all files and content are addressed through Content Identifiers (CIDs), and the mapping from CID to source file is permanently stored. This means that even if a file disappears from the IPFS network, it can still be retrieved by the Fleek Network as long as it exists on at least one source. Moreover, Fleek Network integrates decentralized storage protocols (such as Filecoin and Arweave) and CDN services, providing a high-performance and economical IPFS pinning service that holds advantages over traditional centralized IPFS pinning services. Storing data on decentralized storage protocols is generally more economical than on cloud platforms, providing better data security and availability guarantees.
Other services include: Decentralized Edge Computing (based on Web3 protocols); Blockchain Snapshot as a Service; zkVM, EVM, and other VMs as a Service; Alternative Rollup Sequencer; Temporary Rollups; Edge Proofs of Presence; Verifiable Randomness; Web3 Querying/Events.
As of October 21, 2023, Fleek Network has submitted a total of 4578 code submissions. As the testing network was just opened on October 19, we can expect its future performance.
The testnet is divided into six phases, "0-5," and we are currently in phase 1. The tasks and progress of each phase are as follows:
Phase 0 (September 5): Launch of the node, initial network and node testing (performance, hardware specifications, clusters, cost, metrics, etc.)
Phase 1 (Ongoing): Launch of SDK/services, introduction of SDK, and testing of building and running services on the network, as well as some optimizations based on phase 0
Phase 2 (October): Launch of initial economics, introduction and testing of a more concrete version of economic algorithms, including staking, pricing, and using other elements/situations (valueless) tokens tested, as well as some optimizations based on phase 1
Phase 3 (November): Launch of Layer 2 contracts, introduction of test versions of various aspects of the protocol running on Ethereum L2 (staking, deposits, token contracts, communication between L2/FN, etc.)
Phase 4 (December): Final launch, introduction of the final form of the first generation network based on all the data/feedback and optimizations from all phases, and allowing testing of the actual mainnet environment
Phase 5 (Q1 2024): Mainnet launch
In the current phase 1 of the testnet, users can follow the guidelines to become node operators. For the specific process, please refer to: https://blog.fleek.network/post/fleek-network-testnet-phase-1/
Participants in the testnet can share 1% of $FLK tokens, and the reputation of nodes established on the testnet can be migrated to the mainnet. In addition, the foundation may provide additional grants to specific node operators, service builders, or users during the testnet period.
$FLK is the utility token of Fleek Network, issued as an ERC-20 token on Ethereum. It also serves as a governance token, with a maximum supply of 100 million. The specific allocation is as follows:
The primary income for token holders comes from a basket of stablecoins. For example, node operators settle their work rewards in stablecoins (pegged to the US dollar), ensuring a stable and predictable income stream.
$FLK is a staking token and an indispensable part of Fleek Network. Nodes running Fleek Network client software need to stake a certain amount of FLK to participate in the network and have the opportunity to earn fees by providing work. Below are some expected features of the $FLK token:
Work is allocated among nodes based on location (latency) and reputation (performance).
Resources on the network are priced and paid for by users in stablecoins pegged to the US dollar. Resource pricing occurs at the network level.
Nodes are only rewarded for the resources used/consumed in relation to the work they perform (including FLK rewards).
FLK rewards are paid per epoch (approximately 24 hours). $FLK rewards for each epoch are distributed proportionally among nodes that performed work during that period, calculated as the dollar income earned by each node compared to the total income earned during that period.
20% of the total supply of $FLK tokens is allocated for staking/rewards, using an inflationary mechanism to provide additional rewards to stakeholders. The actual reward/inflation rate will be controlled and updated algorithmically based on network usage, the market price of $FLK, and other factors.
Interestingly, the $FLK rewards for nodes will be dynamically adjusted. The system will take into account both the US dollar stablecoin fees earned by the node and the $FLK rewards. This means that if network usage/US dollar income increases, the $FLK rewards may decrease. Conversely, if the time-weighted average market price of $FLK rises, the absolute number of expected $FLK rewards for the node may decrease, and vice versa. With this design, the long-term income of nodes is expected to be more consistent.
Filecoin： A leader in the Web3 storage space, Filecoin is a decentralized storage network that transforms cloud storage into an algorithmic market. Miners earn Filecoin by providing storage to clients, who in turn spend Filecoin to hire miners for data storage or distribution. The project was created by the Protocol Labs team, which also founded IPFS (InterPlanetary File System). Filecoin employs a robust consensus mechanism based on Proof of Replication and Proof of Spacetime protocols. These proofs allow miners to verify the exact amount of storage space provided and the duration of data storage, ensuring data integrity and security. With the launch of Filecoin Virtual Machine (FVM), the project's applications extend beyond data storage and retrieval, aiming to restructure the underlying architecture of the internet. This results in a high overlap with the services provided by Fleek Network.
Storj：A decentralized, blockchain-based distributed cloud storage system, Storj offers functionalities similar to Filecoin but with a more focused application scope. While it doesn't have the groundbreaking innovation found in Filecoin and IPFS, Storj's advantages become clear when compared to traditional centralized storage solutions.
Arweave: Unlike Filecoin and similar platforms that use traditional blockchain architecture for transaction and off-chain storage, Arweave employs on-chain storage. Each node stores only a portion of the blocks, avoiding the need to sync the entire blockchain. Blocks in Arweave not only link to the previous block, like traditional blockchain protocols, but also point to a random previous recall chunk, creating a spool-like block structure called blockweave. Additionally, there is no size limit for each block. Arweave has a well-developed ecosystem, including infrastructure, middleware, and applications, as well as its own NFT standard that addresses the issue of off-chain NFT metadata storage, providing a comprehensive value proposition and application scenarios.
Competitive Risk: Being relatively young compared to established storage leaders, Fleek Network may face challenges in customer acquisition due to the first-mover advantages of its competitors.
Economic Model Risk: The network adopts a dynamic model to adjust rewards for nodes, with critical parameters determined by network proposals. This could lead to proposal rejections due to conflicts of interest, potentially destabilizing the network.
Fleek Network is poised to play a significant role in decentralizing CDN and edge layers, allowing users to leverage its caching and computational capabilities. Beyond storage and data retrieval, Fleek Network can be used for computational tasks and operations, such as server-side rendering and data processing. A major advantage of Fleek Network is its support for various infrastructures, products, and features available on the Fleek platform, enhancing the developer experience. Backed by a mature platform and leading investment institutions, Fleek Network holds a competitive edge in the market.
In terms of tokenomics design, the model of Fleek Network is yet to be finalized, with a focus on value capture and ecosystem incentives rather than a large share for the team.
Considering its competitiveness, economic model, and associated risks, Fleek Network is rated 4 stars.
— Reference —
Website：https://fleek.network/
Project Docs：https://docs.fleek.network/docs/
Whitepaper：https://fleek.network/whitepaper.pdf
— Investment Risks and Disclaimer —
This report is provided for reference only and should not be used as a basis for investment advice or decisions. Investors should not make any investment actions based on this content. Brillian Research and the authors of the report assume no responsibility for the investment consequences arising from this.
The report is effective as of the date of its publication. However, as time progresses, market and economic conditions may fluctuate, and the information in the report may not be able to reflect these changes in real-time. The graphics, charts, and other visual presentation materials provided in this report are for reference only. They do not imply that Brillian Research will guide investment decisions, nor can they comprehensively display the various factors that need to be considered in investment decisions.
Some views or assumptions in the report may reflect Brillian Research’s expectations or predictions for the future. However, numerous known and unknown risks and uncertainties may cause actual conditions to deviate from the views or assumptions in this report.
Any expectations, predictions, or estimates in this text are based on specific assumptions and are subject to certain uncertainties. These forward-looking statements may be affected by inaccurate assumptions or unpredictable risk factors and may differ from actual conditions. Readers should note that some or all of the forward-looking assumptions may not be realized.
No activity yet