KR20200093089A - System and method for transaction of work requests and products based on blockchain - Google Patents

Abstract

The present invention relates to a system for blockchain-based work request and work product transaction. The blockchain manages a distributed ledger including blocks including work product evaluation data, work product, and work request including evaluation request, reward, and reward function and includes nodes including at least one computing device each. Of the nodes, a client node transmits the work request to the nodes and an evaluator node transmits the acceptance of the evaluation request to the client node in response to the work request. Of the nodes, at least one worker node transmits the acceptance of the work request to the client node in response to the work request and the worker node transmits the work product to the evaluator node. The evaluator node transmits the evaluation of the work product to the client node. The client node transmits the approval of the evaluation to the nodes. At least one of the nodes transmits reward to at least one of the evaluator node, the worker node, and the client node based on the reward function included in one of the blocks.

Description

Blockchain-based work request and outcome transaction method and system {SYSTEM AND METHOD FOR TRANSACTION OF WORK REQUESTS AND PRODUCTS BASED ON BLOCKCHAIN}

The present disclosure relates to a method and system for trading blockchain-based work requests and deliverables. More specifically, the present disclosure relates to a job transaction method and system that solves various jobs by efficiently using job execution resources such as storage and computing devices distributed and operated on a blockchain platform.

Most of the cloud-based managed computing platforms are run by large companies such as Amazon, Google, IBM, and Microsoft. Such a cloud-based computing platform provides companies with cost-effective computing solutions by utilizing economies of scale through centralization. However, such a centralized resource management method is difficult for individuals or small research groups to use, and thus, these small businesses may be a barrier to using the cloud system.

Difficulties for individuals or small businesses to use existing cloud-based computing platforms may arise, for example, in machine learning implementations of increasing interest. In other words, the problem of large-scale machine learning often requires complex system setup and expensive storage/compute resources. Because of this, even if the latest machine learning studies have been published, it can be very difficult or very long to reproduce or test the results.

Because of these issues, small research groups and companies may build private clouds, but this is cost-effective. In addition to the research field, private clouds are used in various fields that require machine learning. However, even if a private cloud is built, only a limited number of researchers use it, so the computing resources are not fully utilized.

Meanwhile, large AI companies prefer to use the most powerful resources for themselves. For example, new hardware devices are made available only to individually selected customers before they are released to the general public. In this case, small customers simply have to accept the hardware vendor's policy, even if it doesn't fit their needs, or wait until the vendor changes the policy.

Therefore, there is a need for a more neutral platform that can provide solutions in an efficient manner while sharing computing resources independent of a provider of centralized computing resources.

Blockchain technology can be used as a way to share computing resources without a centralized manager or provider. Blockchain technology is a technology that records and stores transactions made on network communications in a reliable and secure way. Blockchain network is a distributed environment system that can exchange digitized assets or transactions, and records the history of electronic transactions generated in peer-to-peer networks using a shared ledger. Record. The blockchain network uses a decentralized consensus mechanism. In particular, every validating node on the network approves (or disapproves) the transaction by executing the same (or agreed upon) consensus algorithm for the same transaction. Since the blockchain P2P network uses such a decentralized structure and consensus algorithm, it is virtually impossible to forge and falsify the transaction history by a third party, thereby guaranteeing the reliability and transparency of the transaction history.

One example of a blockchain solution that solves the problem of centralized computing is Ethereum. Ethereum, which can provide smart contracts without a central entity, provides a Turing-complete virtual machine, which can theoretically solve any computational problem, including machine learning. Ethereum offers smart contracts by decentralized logic, Swarm as decentralized storage, and Whisper as decentralized messaging to provide a fully decentralized platform.

However, distributed storage provided by Ethereum has a problem in that performance is lower than that of a centralized storage and a high operation cost is required. In areas where performance is important, such as machine learning, the efficiency of an empirically focused entity may be better. In addition, common blockchain models, including Ethereum, are not suitable for carrying large amounts of information, and the performance of Ethereum's smart contracts may not be sufficient to represent complex programs such as machine learning algorithms.

The present disclosure provides a method and system for trading work requests and work results between distributed terminals in a blockchain-based computing resource providing platform. In addition, the present disclosure provides a distributed computing method and system that is easy to apply to complex and large-scale tasks by evaluating and evaluating the results provided by a plurality of worker nodes by an evaluator node in a blockchain-based computing resource providing platform do.

In addition, the present disclosure provides a reliable, high-performance distributed computing platform through a hybrid approach that combines centralized tidal components, such as a secure runtime environment, mass storage, etc., on a distributed blockchain platform.

In a method of trading a block chain based work request and work result according to an embodiment of the present disclosure, the block chain includes a work request, wherein the work request includes an evaluation request, a reward and a compensation function, a work result and It manages a distributed ledger including a plurality of blocks including data related to evaluation of work results, and includes a plurality of nodes each including at least one computing device. The method comprises the steps of transmitting a job request from the plurality of nodes to the plurality of nodes by a client node; Sending an acceptance of the evaluation request to the client node, in response to the work request, by an evaluator node among the plurality of nodes; Sending, by at least one worker node among the plurality of nodes, an acceptance of the work request to the client node in response to the work request; Transmitting, by the worker node, the work result to the evaluator node; Sending, by the evaluator node, an evaluation of the work outcome to the client node; Sending, by the client node, an approval of the evaluation to the plurality of nodes; And transmitting, by at least one of the plurality of nodes, at least one of the evaluator node, the worker node, and the client node based on the compensation function included in any one of the plurality of blocks. do.

In a method of trading a block chain based work request and work result according to another embodiment of the present disclosure, the block chain includes a work request, wherein the work request includes an evaluation request, a reward and a reward function, a work result, and It manages a distributed ledger including a plurality of blocks including data related to evaluation of work results, and includes a plurality of nodes each including at least one computing device. The method includes generating a block including identification information of the client node and data related to the job request when a job request is transmitted to the plurality of nodes by a client node among the plurality of nodes; If an acceptance of the evaluation request is transmitted to the client node by an evaluator node among the plurality of nodes, in response to the operation request, a block including identification information of the evaluator node and data regarding acceptance of the evaluation request is generated. Generating; When the acceptance of the work request is transmitted to the client node by at least one worker node among the plurality of nodes, in response to the work request, includes identification information of the worker node and data regarding acceptance of the work request. Generating a block to be played; Generating, by the worker node, a block including identification information of the worker node and data on the work result when the work result is transmitted to the evaluator node; Generating, by the evaluator node, a block including identification information of the evaluator node and data on the evaluation when an evaluation of the work result is transmitted to the client node; Generating, by the client node, a block including identification information of the client node and data on the evaluation approval when the approval of the evaluation is transmitted to the plurality of nodes; And when at least one node among the plurality of nodes transmits a compensation to at least one of the evaluator node, the worker node, and the client node based on the compensation function included in any one of the plurality of blocks, the compensation And generating a block including identification information of the payment node and data on the reward.

In a system for trading blockchain-based work requests and work results according to another embodiment of the present disclosure, the block chain includes a work request, wherein the work request includes an evaluation request, a reward and a reward function, and a work result And a distributed ledger including a plurality of blocks including data regarding evaluation of work results, and a plurality of nodes each including at least one computing device. The plurality of nodes may transmit a work request to the plurality of nodes, and a client node capable of generating a block including identification information of the client node and data regarding the work request when the work request is transmitted; An evaluator node capable of transmitting an acceptance of the evaluation request to a client node, and generating a block including identification information of the evaluator node and data regarding acceptance of the evaluation request; And a worker node capable of sending an acceptance of the work request to the client node, and generating a block including identification information of the worker node and data regarding acceptance of the work request, wherein the worker node includes: When the work result is transmitted to the evaluator node, a block including identification information of the worker node and data related to the work result is generated, and the evaluator node transmits an evaluation of the work result to the client node. Then, if a block including identification information of the evaluator node and data related to the evaluation is generated, and the client node sends the approval of the evaluation to the plurality of nodes, the client node includes identification information of the client node. A block including data regarding evaluation approval is generated, and at least one node among the plurality of nodes is based on the compensation function included in any one of the plurality of blocks, the evaluator node, the worker node, and the client node When a reward is transmitted to at least one of the blocks, a block including identification information of the reward payment node and data related to the reward is generated.

According to various embodiments of the present disclosure, in a blockchain-based distributed computing resource providing platform, an evaluator node can identify and evaluate a work result provided by a plurality of worker nodes, and further execute a customized evaluation according to a type of work By doing so, it is possible to provide a distributed computing resource providing platform that is easy to apply to complex and large-scale tasks by efficiently using the minimum resources for each task.

In addition, according to some embodiments of the present disclosure, through a hybrid approach that combines a secure runtime environment and mass storage, which are centralized tidal components, to a blockchain-based distributed computing resource providing platform, a variety of various methods are provided through minimal centralized control. At the same time, it is possible to solve problems and to enable participating nodes to efficiently access and improve the solution to the task.

Specifically, a secure runtime environment in accordance with some embodiments of the present disclosure provides requested secure code to participating nodes (ie, members of the platform) and is suitable for various device combinations of working nodes. Allows tasks to be scheduled and assigned. In addition, the mass storage solves the problem that the existing decentralized storage is inefficient and expensive in the processing of large data, so that large-scale operations such as machine learning can be properly applied to the distributed computing environment.

Embodiments of the present disclosure will be described with reference to the accompanying drawings described below, where similar reference numerals indicate similar elements, but are not limited thereto.
1 is a schematic diagram showing a transaction system of a block chain based work request and work result according to an embodiment of the present disclosure.
2 is a block diagram showing the relationship between distributed terminals in a transaction system based on a blockchain-based work request and work result according to an embodiment of the present disclosure.
3 is a flowchart illustrating a method for trading a blockchain-based work request and a work result according to an embodiment of the present disclosure.
4 is a method of trading a block-based work request and a work product according to an embodiment of the present disclosure, wherein the manager node sends a work specification to a plurality of nodes to recruit members, and the work request and the work product include It is a flow chart showing the transaction method.
5 is a schematic diagram showing the operation of a host node and a public storage node in a transaction system based on a blockchain-based work request and work result according to an embodiment of the present disclosure.
6 is a schematic diagram illustrating an operation in which a host node according to an embodiment of the present disclosure sends an application including a secure runtime environment set from an administrator node to a worker node and an evaluator node.
7 is a schematic diagram showing a configuration in which content of a work result according to an embodiment of the present disclosure is stored in a common storage node, and a pointer to the content is stored in a block of the blockchain.
8 is a schematic diagram showing a configuration in which a common storage node according to an embodiment of the present disclosure distributes and stores contents of a result of a job in a plurality of shards, and a plurality of worker nodes communicate with each shard.
9 is a block diagram based on an embodiment of the present disclosure showing a configuration for evaluating a work result based on a change amount of a loss function value according to a parameter change in a transaction system for requesting a work and learning result by an artificial neural network based on a blockchain It is a schematic.

Hereinafter, with reference to the accompanying drawings, specific details for the practice of the present disclosure will be described in detail. However, in the following description, when there is a risk of unnecessarily obscuring the subject matter of the present disclosure, detailed descriptions of well-known functions or configurations will be omitted.

In the accompanying drawings, identical or corresponding elements are given the same reference numerals. In addition, in the following description of the embodiments, it may be omitted to describe the same or corresponding elements overlapping. However, although descriptions of components are omitted, it is not intended that such components are not included in any embodiment.

According to an embodiment, in a blockchain-based distributed computing resource providing platform, distributed storage and a worker that perform a task requiring large amounts of data and computation, such as learning an artificial neural network, requested by a client node, The results of the work executed by the nodes and executed by the worker nodes can be evaluated by the evaluator node. Based on the evaluation of the evaluator node with respect to the work result, rewards may be paid to nodes related to the task, such as a worker node. Through this method, the client node can expect high-quality work results while using resources distributed on the blockchain network in an optimal manner, and worker nodes participating in the blockchain network can reasonably evaluate the work results and accordingly You can expect compensation. Furthermore, since it is possible to execute a customized evaluation according to the type of work executed on the blockchain network, a distributed computing resource providing platform is provided that is easy to apply to complex large-scale tasks by efficiently using minimum resources for each task.

In addition, according to some embodiments of the present disclosure, a blockchain-based distributed computing resource providing platform includes a hybrid approach that combines a secure runtime environment, which is a centralized help component, and a large storage. Is provided. This minimal centralized control improves the inefficiencies of task execution that can occur in a common distributed cloud computing approach, solving many problems ranging from large jobs to tasks, while participating nodes provide solutions to tasks. It can be effectively approached and improved.

In the present disclosure, "blockchain (blockchain)" is a system of a distributed environment that can exchange digitized information, assets, or transactions in general, using a shared ledger (P2P) (peer- to-peer) may refer to a system or platform that records the history of electronic transaction history generated in a network. Here, the blockchain uses a decentralized consensus mechanism, and the validating node on the network executes the same (or agreed upon) consensus algorithm for the same transaction details. It may be a public blockchain that can approve (or disapprove), but is not limited thereto, and may include a private blockchain or a combination thereof. In addition, in the present disclosure, "node (node)" refers to a computing device capable of recording, maintaining, and storing a shared book in the blockchain, and having a computational power capable of recording and processing transaction history, such as creation of blocks. can do. In addition, in the present disclosure, the "node" may be any one of a manager node, a worker node, an evaluator node, and a client node described in detail below, but is not limited thereto. For example, the "node" is one of the nodes participating in the blockchain network, and may be a node other than the manager node, worker node, evaluator node, and client node.

1 is a schematic diagram showing a transaction system of a block chain based work request and work result according to an embodiment of the present disclosure.

The blockchain-based work request and work result transaction system 100 is configured to allow a plurality of nodes to communicate with each other through the blockchain network 150. According to an embodiment, the blockchain stored and managed by the system 100 is a distributed ledger including a plurality of blocks including data regarding work requests, work results, and evaluation of work results between two or more nodes or terminals. (distributed ledger). In addition, the plurality of nodes constituting the blockchain may each include at least one computing device, and each of the plurality of blocks may be a hash reference for data regarding work requests, work results, and evaluation of work results. It may include.

Through the blockchain network 150, work requests and transactions of work results can be performed between nodes. The client node 110 may send a job request including compensation for a specific job, and the manager node 120 may create and manage a work environment executed by the worker node 140 and the evaluator node 130. It can perform the function of coordinating the operation and interconnection of other nodes. Meanwhile, the worker node 140 may perform a task requested by the client node 110 and provide the result, and the evaluator node 130 may check and evaluate the work result provided by the worker node 140 Can. Appropriate compensation for the worker node 140 may be paid according to the evaluation of the work result by the evaluator node 130.

In one embodiment, the job request includes information requesting that at least one of the plurality of worker nodes 140 execute the job and provide the result. For example, the job request may include an evaluation request, a reward for the job, and a reward function. In addition, the evaluation request includes information requesting at least one evaluator node 130 among the plurality of nodes to evaluate the work results provided by the worker node 140. For example, the evaluation request may include information requesting to rank a plurality of work results. Meanwhile, as described in detail below, the reward is a product that is finally paid to the worker node when the evaluator approves and approves the approved result according to the reward function. That is, the reward is a product to be paid as an object to be paid for the result of the work and/or evaluation, and may be an encrypted currency, but is not limited thereto. The compensation function may be a function that designates a worker node 140 and a method to pay compensation according to the evaluation result of the evaluator node 130. For example, the reward function may designate that the reward is equally distributed to all of the one or more worker nodes 140 that performed the work, or may be designated to be rewarded only to the worker node 140 having the highest evaluation rank. . The compensation function may be defined differently according to the preference of the client node 110 requesting the task or the type of the requesting task, and may be operated to optimize the compensation function to execute the requested task at a reasonable cost.

2 is a block diagram showing the relationship between distributed nodes or terminals in a transaction system based on a blockchain-based work request and work result according to an embodiment of the present disclosure.

As illustrated, the client node 110 may proceed with a transaction requesting work while depositing a reward to be paid on the blockchain network 150. In one embodiment, the client node 110 may be configured to send work requests to a plurality of nodes in the blockchain network 150. Here, the work request may include an evaluation request, compensation, and a compensation function. In this embodiment, the task may be a learning task by an artificial neural network such as machine learning. For example, when the client node 110 requests a work result and an evaluation to a plurality of nodes, a reward and a reward function may be transmitted together. As such, the rewards and reward functions transmitted to the blockchain network 150 are recorded in the blockchain book. After this, if there is approval of the work evaluation by the evaluator node 130, compensation is paid to a plurality of nodes related to the corresponding task, for example, at least one worker node 140 and the evaluator node 130 according to the compensation function. It can be configured as possible. The client node 110 may request to be executed by dividing one task into several tasks. At this time, the manager node 120, the evaluator node 130, and the worker node 140 related to each divided task may be different, and their organizational systems may also be different. In the present disclosure, "organization system" may refer to upper and lower hierarchical relationships between nodes participating in the blockchain network. In this organizational system, a vertical connection relationship between the manager node 120 and other nodes, worker nodes It may include a horizontal connection relationship between the 140 and the evaluator node 130. In FIG. 2, examples of various organizational systems between nodes 120 to 140 are shown for each task (eg, task 1, task 2, task 3). According to one embodiment, each task may have an organizational system between different nodes. Further, the client node 110 may be configured to generate a block including its own identification information and data related to the work request when the work request is transmitted.

The manager node 120 can create and manage a work environment executed by the worker node 140 and the evaluator node 130, and can perform functions to coordinate the operation and interconnection of other nodes, and challenges You can approve or reject the request. The manager node 120 may be a node that first defines a job specification or a node designated by the node as an administrator.

According to one embodiment, the manager node 120 transmits a job specification to a plurality of nodes of the blockchain network 150. Furthermore, when an appraiser node request is sent to the manager node 120 by at least one node among a plurality of nodes, the manager node 120 transmits the approval of the appraiser node 130 to at least one of the nodes that sent the appraiser node request. do. In addition, when a worker node request is transmitted to the manager node 120 by at least one node among a plurality of nodes, the manager node 120 may transmit a worker node acknowledgment to at least one of the nodes that have sent the worker node request.

In the above-described procedures, the manager node 120 may generate a block including data generated in each procedure, including its own identification information. In addition, the manager node 120 may recruit the evaluator node 130 and the worker node 140 to perform evaluation and execution of the corresponding job by broadcasting a job specification to a plurality of nodes. In response to the sending and recruiting of these job specifications, one or more nodes may send a request for evaluation and execution of the job. In response to this request, the manager node 120 may approve that the nodes that have transmitted the request become the evaluator node 130, the worker node 140, or another manager node. The manager node 120 may examine whether the evaluator node 130 or the worker node 140 has a resource capable of executing a task to determine whether to approve the approval. Here, the manager node 120 may be the evaluator node 130 or the worker node 140 itself.

The manager node 120 can coordinate the working environment. Specifically, the administrator node 120 defines a task schema for a specific task and implements the implementation method to the evaluator node 130 and the worker node 140 to evaluate and execute the task thereafter. It can be provided as a binary manager. The binary manager determines the batch of executable programs (ie binary code) of the jobs to be executed according to the job schema. In addition, the administrator node 120 may configure and provide a run-time environment in which a job is to be executed to be reliable and secure. According to an embodiment, the administrator node 120 transmits a security runtime environment setting to a host node among a plurality of nodes, and generates a block including data regarding its own identification information and security runtime environment setting. can do. In addition, the evaluator node 130 and the worker node 140 may receive an application including a set secure runtime environment from a host node and execute a task within the secure runtime environment.

In addition, the manager node 120 may approve or reject the challenge request, and generate a block including data regarding the identity of the challenge request and the approval or rejection of the challenge request. Here, the challenge request is a request for the client node 110 or the worker node 140 to disagree with the evaluation of the work product, and to obtain a trial panel for evaluation with respect to the upper node. In one embodiment, the compensation function may include a compensation function for at least one of approval and rejection of a challenge request. Also, the client node 110 or at least one worker node 140 may transmit a challenge request to the manager node 120. The manager node 120 may determine whether the evaluation of the work result transmitted by the evaluator node 130 is valid, and transmit approval or rejection of a challenge request to a plurality of nodes. In one embodiment, the reward function may be set such that rewards are paid differently when the challenge request is approved and rejected.

The evaluator node 130 may evaluate the work result of the worker node 140 and transmit it to the client node 110. According to an embodiment, a node that transmits acceptance among a plurality of nodes receiving an evaluation request from the client node 110 may be configured to be the evaluator node 130. According to another embodiment, for the job specification broadcast by the manager node 120 to a plurality of nodes, at least one node among the plurality of nodes requests to be the evaluator node 130, and the manager node 120 approves it By doing so, the corresponding application node may be configured to be the evaluator node 130.

The evaluator node 130 may receive a work result from the worker node 140. The evaluator node 130 validates and evaluates the results. The evaluator node 130 may transmit an evaluation of the work results to the client node 110. According to this evaluation, or according to the compensation function according to the evaluation, the worker node 140 may be configured to receive differentiated compensation. In addition, the evaluator node 130 may generate a block including data related to the identification information and work result evaluation.

The worker node 140 executes the task and transmits the result of the task. According to an embodiment, a node that transmits acceptance among a plurality of nodes receiving a work request from the client node 110 may be configured to be the worker node 140. According to another embodiment, for the job specification broadcast by the manager node 120 to a plurality of nodes, at least one node among the plurality of nodes requests to become the worker node 140, and the manager node 120 approves it. It may be configured to be an evaluator node 140.

The worker node 140 may perform the requested operation. According to an embodiment, the task may be a learning task using an artificial neural network, in which case the task may be a process of generating a learning process through input data as a code. In addition, the worker node 140 may be configured to transmit the results of the work to the evaluator node 130. The output of the work may include content such as a file or code, and may include a content address such as a link to stored content, a pointer, and the like. According to an embodiment, the worker node 140 may transmit the work result content to a public storage node that provides a large storage. In this case, the common storage node may be configured to generate an address of the corresponding work result content and transmit the address of the work result content to the worker node 140. At this time, the worker node 140 may be configured to generate a block including data regarding its own identification information and work result (or its address).

Each node, including the client node 110, the manager node 120, the evaluator node 130, and the worker node 140, may include a computing device. For example, each node may be a portable computer such as a smart phone or a tablet PC, a desktop computer, or a server device, but is not limited thereto.

In this embodiment, a fee may be required when nodes generate blocks. That is, to create a block for each transaction and record it in the blockchain network 150, it can be configured to provide a fee to nodes of the blockchain network. According to an example, the fee may be configured to be equal to or higher than the fee at the first request of work by the client node 110.

According to an embodiment, a reputation for interaction between nodes may be recorded in the blockchain as a ledger. Each node can grasp the reputation of nodes by tracking information such as the approval of the manager, evaluator, or worker nodes recorded on the blockchain, requests for tasks, acceptance of tasks, submission of work results, and distribution of rewards. Through this reputation, nodes can form higher trust in the transaction process or receive higher rewards. For example, the client node 110 may refuse to perform work by the poorly reputed worker node 140. As another example, the worker node 140 may reject the acceptance of the work request when a highly competitive worker node 140 is included, and thus fierce competition is expected in accepting the work.

3 is a flowchart illustrating a method for trading a blockchain-based work request and a work result according to an embodiment of the present disclosure. The execution results of each step of the method 300 of transaction request based on the blockchain and the work result described in detail below may be generated and recorded as a block in the blockchain network 150. However, hereinafter, for convenience of description, description of a block generation process for an execution result of each step may be omitted.

According to an embodiment of the present invention, a method 300 for trading a blockchain-based work request and a work result is transmitted by a client node among a plurality of nodes to a blockchain network (or a plurality of nodes connected to the network). Initiated by step S310. The work request may include an evaluation request, a reward for the job, and a reward function. Specifically, the job request may include an identification number of the job, information on input data, a time limit, a list of desired worker or evaluator nodes, and information that the evaluation procedure is omitted when the worker node 140 is trusted. It may include. As a result of the execution of step S310, the client node 110 generates a block including identification information of the client node 110 and data regarding a work request.

In response to the transmission of the job request, the evaluation request acceptance is transmitted to the client node 110 by the evaluator node among the plurality of nodes (S312). In this step, the evaluator node 130 generates a block including identification information of the evaluator node 130 and data regarding acceptance of the evaluation request.

Next, when the work request acceptance is transmitted to the client node 110 by at least one worker node 140 among the plurality of nodes (S314), the identification information of the worker node 140 and data on accepting the work request are included. Block.

After the above steps S310 to S314 are executed, the worker node 140 executes a job according to the job request. The job can be executed on a job environment set by the administrator node 120. For example, as an application that includes a secure runtime environment, tasks can be executed using dockers, sandbox wrappers, programming languages, and libraries. As a result of the task execution, when the task result is transmitted to the evaluator node 130 by the worker node 140 (S316), a block including identification information of the worker node 140 and data on the task result is generated.

Next, the evaluator node 130 performs an evaluation on the work results provided by the worker node 140. For example, the evaluator node 130 drives the work results on the work environment set from the manager node 120, measures the evaluation value for the work results, or the relative rank (rank) and/or meets the criteria (pass) or fail). As a result, when the evaluation of the work result is transmitted to the client node by the evaluator node (S318), a block including identification information and evaluation data of the evaluator node is generated.

The client node can review whether the evaluation provided by the evaluator node is appropriate. As a result, when the approval of the evaluation (S330) is transmitted to the plurality of nodes by the client node, a block including identification information of the client node and data on the evaluation approval is generated.

Next, at least one of the evaluator node, the worker node, and the client node based on the compensation function included in the block (for example, a block generated based on the client node's work request) by at least one node among the plurality of nodes. When one reward is transmitted (S332, S334), a block including identification information of the node that transmitted the reward and data related to the reward is generated. Here, the data related to the reward may include the size of the reward, the object paid, and the like.

In the above-described embodiment, the compensation is described to be paid to the operator and the evaluator, but compensation may be paid in various ways according to the compensation function. For example, when the job output is failing evaluation by the evaluator node, it may be configured to return the reward to the client node without paying the reward to the worker node that executed the job.

According to one embodiment, shared computer resources can be allocated to each node and tasks for efficient use and evaluation on the blockchain network. Ethereum, a kind of existing blockchain, uses the Ethereum Virtual Machine (EVM). In this environment, all lines of code (line-by-line) are verified to verify that the provided code is executed well. Such a code verification method is inefficient for verification of a large-scale application program. In particular, the machine learning code is generally large and complex, in which case the blockchain-based worker/evaluator model according to the present embodiment is advantageous in terms of performance. In one example, the evaluator node 130 identifies the work output provided by the plurality of worker nodes 140 and determines the rating or rating of the work. The method of evaluation can be different depending on the type of task. For some tasks, a simple process of checking whether the result value satisfies a specific threshold or boundary can be sufficiently evaluated. For other tasks, it is possible to adopt the one with the lowest loss as a result of machine learning learning through competition between work results executed by a plurality of worker nodes 140. Through the customized evaluation according to the work as described above, it is possible to consume only the minimum amount of resources available on the blockchain network.

In one embodiment, the client node 110 or the worker node 140 may be configured to apply for a challenge. For example, there may be a case where the client node 110 wants to obtain a judge for evaluation by dissatisfied with the evaluation of the work results transmitted by the evaluator node 130. At this time, the client node 110 may request a judgment as to whether the evaluation is valid for an upper node such as the manager node 120. When such a challenge application is possible, the compensation function may include a function related to compensation according to at least one of approval and rejection of the challenge request.

Steps S340 to S364 illustrated in FIG. 3 represent a challenge and reward procedure for evaluation of work results in a blockchain network according to an embodiment. According to the present embodiment, the client node or at least one worker node transmits a challenge request to the manager node (S340). In response to the transmission of the challenge request, the manager node sends approval or rejection of the challenge request to a plurality of nodes (S350, S360). The compensation function used in the steps (S352, S354, S362, and S364) in which compensation is executed may be set such that compensation is differently paid when the challenge request is approved and rejected. For example, when the challenge request is approved, the reward to be paid to the evaluator node 140 is configured to be paid to the client node 110, and when the challenge request is rejected, it is configured to be paid according to a normal compensation function. Can.

In FIG. 3, the rewards are shown in steps S352, S354, S362, and S364, which are paid to the evaluator node 130 and the worker node 140, or to the client node 110 and the worker node 140, The reward method is not limited thereto, and rewards may be paid in various ways according to the reward function. For example, when the manager node 120 approves the challenge request for the challenge requested by the client node 110, that is, when it is determined that the evaluation of the work result by the evaluator node 130 is wrong, the evaluator node 130 ) And the worker node 140 may be configured to return the reward to the client node 110 without paying the reward.

According to an embodiment, when the evaluator node 130 accepts an evaluation request from the client node 110, a certain deposit may be configured to be attracted to the blockchain network. In addition, when the challenge request for evaluation is approved by the manager node 120, that is, when it is determined that the evaluation of the evaluator node 130 is wrong, a deposit is transmitted to the client node 110. On the other hand, when the evaluation is approved by the client node 110, the deposit may be configured to be returned to the evaluator node 130. The deposit may be configured to be larger than the fee required for the challenge request. When such a deposit system is introduced, it is possible to expect a more reliable evaluation from the evaluator node 130.

4 is a flowchart illustrating a procedure in which a manager node sends a job specification to a plurality of nodes to recruit members in a method of trading a blockchain-based job request and a job result according to an embodiment of the present disclosure. The method 400 for trading a block-based work request and a work result according to an embodiment may be initiated by step S410 in which a work specification is transmitted to a plurality of nodes by the manager node 120. The manager node 120 may recruit the evaluator node 130 and the worker node 140 by broadcasting job specifications to notify other participating nodes about the job. Here, the job specification indicates contents related to the job, and may include test data on the job, requirements required by the evaluator, and the operator. As a result of the execution of step S410, the manager node 120 generates a block including data related to the job specification including identification information of the manager node 120.

In response to the transmission of the job specification, the evaluator node request is transmitted to the manager node 120 by at least one node among the plurality of nodes (S412). In this step, a block including identification information of the node that sent the evaluator node request and data regarding the evaluator node request is generated.

In addition, in response to the appraiser node request, the approval of the appraiser node is transmitted to at least one of the nodes that have sent the appraiser node request by the manager node 120 (S414). According to an embodiment, in this process, the manager node 120 performs performance of processing the test data by the corresponding node using a dashboard in order to determine whether the requested node is eligible as the evaluator node 130 ( Performance), or examine the reputation of the node. As a result of the execution of step S414, the manager node 120 generates a block including identification information of the manager node 120 and data related to appraiser node approval.

On the other hand, in response to the job specification transmission, a worker node request is transmitted to the manager node 120 by at least one node among the plurality of nodes (S416). In this step, a block including identification information of the node that sent the worker node request and data related to the worker node request is generated.

In addition, in response to the worker node request, the approval of the worker node is transmitted to at least one of the nodes that have sent the worker node request by the manager node 120 (S418). According to an embodiment, in this process, the manager node 120 may use a dashboard and a reputation of the corresponding node in order to determine whether the requested node is qualified as the worker node 130. As a result of the execution of step S418, the manager node 120 generates a block including identification information of the manager node 120 and data regarding worker node approval.

As described above, among the plurality of nodes, designated as the evaluator node 130 and the worker node 140 are at least among the plurality of nodes for the job specification broadcast by the manager node 120 to the plurality of nodes. One node may be configured to apply by being the evaluator node 130 or the worker node 140, and the administrator node 120 approving it. The administrator node 120 through the approval/rejection process of the appraiser node 130 and the worker node 140 application, the pool of the appraiser node 130 and the worker node 140 before requesting work from the client node 110 ( pool). The client node may be configured to make work requests and evaluation requests to evaluator nodes and worker nodes in the pool configured as described above.

The job specification broadcast by the manager node 120 may include test data and the like. Therefore, the node requesting to be the evaluator node 130 and/or the worker node 140 executes test data included in the job specification, and the manager node 120 sees and approves the execution performance of the nodes appearing in the dashboard. / Decide to refuse. In the present disclosure, "dashboard" refers to the overall state of the participating nodes (for example, the state of the transaction, the progress of the task, the performance of the node executing the task, etc.) to the users of the blockchain network 150. ) May refer to a data structure or interface capable of displaying. In one example, the dashboard can be expressed in the form of a web page, and each node can access the dashboard through an API. Through this configuration, a pool can be configured and managed with nodes that have the necessary resources to execute a job according to the job specification, and the client can request a job from verified nodes included in the pool. This method of managing pools of nodes is important for executing tasks that require large-scale resources such as machine learning. That is, according to the method according to the present embodiment, it is possible to prevent an unnecessary evaluation and task execution process that may occur by applying a node lacking a large amount of data processing capability of machine learning as a worker or evaluator node.

The resource level of each node in the pool managed by the manager node 120 may vary with time. For example, in general, the level of available resources is lowered because the resources of the node are driven for the operation of the service in the daytime, while the service level may have a high level of resource utilization in the nighttime when the service is not operated. In this case, the nodes in the pool are assigned to a task that requires a small resource and receive a high evaluation, instead of being put on a task that requires a lot of resources during the daytime when the resource utilization level is low, and receiving a low evaluation. It can be advantageous to get a high reward.

5 is a schematic diagram showing the operation of the host node 510 and the common storage node 520 in the transaction system of a blockchain-based work request and work result according to an embodiment of the present disclosure.

The system 500 according to the present exemplary embodiment may provide trust and security to members or nodes connected to the blockchain network 150 and manage various combinations of computing device types. In addition, generally distributed storage (storage) has a problem that is more inefficient and costly to operate than a centralized storage. To solve this problem of the existing distributed storage, the system 500 is a hybrid, which is a combination of a centralized tidal component, a secure runtime environment and a large storage and decentralized blockchain. Implement reliable, high-performance distributed computing through an approach.

As shown, the host node 510 may provide a secure runtime environment and an application that can be executed in the environment. In the present disclosure, a “runtime environment” refers to a state of a virtual machine or a virtual machine that provides software services for a process or program while a computer such as a node is running, a process or process while the computer is running. It can be a collection of software services for a program. For example, the runtime environment may refer to an environment in which a file in the form of an image generated to be executable on a computing device can be safely executed. An image file executable in the runtime environment is an image of a docker hub (for example, see https://hub.docker.com/search/?image_filter=official&type=image ), or an image where a specific application is installed on a virtual machine. It may be (for example, see https://aws.amazon.com/marketplace/b/6297422013?ref_=header_nav_category_6297422013).

In addition, the public storage node 520 functions as a mass storage, and may be any one of a plurality of nodes of the blockchain network 150. Various types of tasks can be solved by executing centralized control such as the common storage node 520 to a minimum. Corresponding to the existing phone Neumann architecture in general, the central processing unit (CPU), memory, input/output (I/O) and communication between them as the main components of computing, by the system 500 of the present embodiment. The distributed computing environment provided can be composed of a secure runtime environment, a mass storage, and a blockchain as main components.

6 is a schematic diagram showing a configuration in which a host node according to an embodiment of the present disclosure transmits an application including a secure runtime environment to a worker node and an evaluator node based on the secure runtime environment setting received from the administrator node. According to one embodiment, the secure runtime environment may be set by the administrator node 120. The manager node 120 may configure and provide a runtime environment in which a job is to be executed to be reliable and secure.

According to an embodiment, the manager node 120 transmits a security runtime environment setting to a host node 510 among a plurality of nodes, and a block including identification information of the manager node 120 and data on the security runtime environment setting Produces In addition, the host node transmits an application including a set security runtime environment to at least one of a worker node and an evaluator node, and generates a block including identification information of the host node and data about the application including the security runtime environment. can do.

Learning tasks by artificial neural networks, such as machine learning, may require various combinations of memory or storage size and computing power of computing devices such as GPUs. Accordingly, the manager node 110 may schedule and allocate the task so that the task properly requested by other computing nodes is executed. That is, the administrator node 110 sets a secure runtime environment suitable for the execution of the requested operation, and the worker node 140 or the evaluator node ( By sending it to 130), it is possible to allocate work to a reliable computer at a reasonable operating cost. The secure runtime environment enables smooth transactions between nodes on the blockchain network by confirming that the code to be executed in the environment is harmless to worker nodes and the like.

The secure runtime environment according to an embodiment may use a sandboxing structure and a resource control method to execute code in various environments. The secure runtime environment may include a reference node that provides a convenient backend for programming learning tasks by artificial neural networks (hereinafter referred to as "backend node"). The secure runtime environment can use dockers and sandbox wrappers as applications for safe execution of code. Backend nodes can provide efficient backend application programming interface (API) servers and popular AI frameworks that host different programming languages. Backend nodes can support a variety of programming languages and runtime libraries, such as Python 2/3, R, PHP, and C/C++. Runtime libraries supported by the backend nodes can include TensorFlow, Keras, Caffe and MXNet in addition to Java, Javascript, Julia, Octave, Haskell, Lua, NodeJS. The secure runtime environment may include a binary manager-responsible for authentication, deployment, and update of the binary. When a new binary is distributed, the binary manager may be configured to notify the evaluator node 130 and the worker node 140 to download the new binary in a secure runtime environment.

7 is a schematic diagram showing a configuration in which contents of a work result according to an embodiment of the present disclosure are stored in a common storage node, and pointers to contents are stored in a block. According to an embodiment, the worker node 140 transmits the content of the work result to the public storage node 520, the public storage node 520 stores the content, and generates an address of the stored content to generate the worker node 140 ). In this case, the worker node 140 generates a block including the first data regarding the identification information and the content address of the worker node. On the other hand, when the worker node 140 transmits the work result to the evaluator node 130 after the task is executed, the corresponding work result may include a content address stored in the common storage node 520. Also, the data related to the work result included in the block generated by the evaluator node 130 may include second data regarding the content address.

In general, the operation of machine learning entails the problem of processing large-scale input/output data, and thus requires efficient data transfer between nodes of the blockchain network according to the present embodiment. However, blocks of general blockchains are designed to have a limited size for efficiency. To solve this problem, the blockchain network according to the present embodiment can share data through a centralized mass storage such as a common storage node 520, thereby efficiently transmitting data between nodes. . For example, the client node 110 may share input data related to the requested operation with the worker node 140 through the common storage node 520. In addition, the evaluator node 130 and the worker node 140 may share the work results, including output data related to the requested operation, with the client node 110 through the public storage node 520. In this data sharing process, each node can use the API provided by the system, for example, a blockchain API can be used to allow calls by the data path of the storage space.

Each node of the blockchain network can secure a suitable space in the public storage node 520 before the actual work execution starts. For example, the administrator node 120 reserves space in the public storage node 520 in the process of coordinating the work, and the client node 110 reserves space for input data and test data to be learned, and the evaluator node ( 130) and the worker node 140 may secure temporary space to be used during job execution and job outcome evaluation.

According to an embodiment, the mass storage managed by the public storage node 520 may be content-address-based. That is, the address of the content stored in the storage may be an encrypted hash function for the uploaded content-for example, SHA-2. Nodes in the blockchain network can share the uploaded data through the public storage node 520, so there is no need to execute updates to the same data on each node, and the data stored in the public storage node 520 is the data owner. It may not contain personal information such as. In one embodiment, in order to encrypt data uploaded to the public storage node 520, the encryption/decryption function and the encryption/decryption key may be provided by the public storage node 520 and the blockchain protocol. In one embodiment, when a file is created and stored in the public storage node 520, the file owner (or file creation node) specifies a time to live (TTL) of the generated file, and the file is automatically generated after a specified period of time. It can be configured to be deleted. This helps to efficiently manage the public storage node 520.

8 is a schematic diagram showing a configuration in which a common storage node according to an embodiment of the present disclosure distributes and stores contents of a result of a job in a plurality of shards, and a plurality of worker nodes communicate with each shard. As illustrated, the common storage node may distribute and store the contents of the work result provided by the worker node in a plurality of shards. In this case, the address of the content may include addresses for a plurality of shards.

In addition, parallel processing of data can be implemented using a client node as a parameter server. That is, when the client node is used as a parameter server, the parameter server executes multiple iterative learning processes while adjusting parameters for multiple neural network models to execute artificial neural network learning in a blockchain-based distributed processing environment. Can. As an example, as illustrated in FIG. 8, when a client node requests a learning operation of an artificial neural network, parameters required for training a plurality of artificial neural network models are divided, and each divided parameter is a plurality of worker nodes and an evaluator node It can be processed by parallel processing by. Here, the client node may repeatedly request the blockchain network based on the improved dataset and parameters. In response to this work request, the worker node, the evaluator node, and the manager node can cooperate with each other and repeatedly execute training of a plurality of artificial neural network models until a satisfactory work result is generated.

In the case of an artificial neural network learning task, the common storage node has a plurality of data shards for input data required for the task. The public storage node can copy and store data shards to multiple storages, thereby enabling workers to efficiently perform data communication with the public storage node. In one embodiment, the client node 110 may request, from an artificial neural network learning task, a task execution and evaluation based on a plurality of weights, an artificial neural network model copy address, and a data shard address from a plurality of evaluator nodes. have. In response, the evaluator node may transmit the learned result, and the client node 110 may update the weight based on the learned result and transmit the adjusted weight back to the evaluator node.

9 is a schematic diagram showing a configuration for evaluating a work result based on a change amount of a loss function value according to a parameter change in a transaction system of a work request and a work result by a blockchain-based artificial neural network according to an embodiment of the present disclosure to be. According to an embodiment, the work request includes a request for learning work by an artificial neural network, and the evaluation may include an evaluation based on a change amount of a loss function value according to a parameter change.

For example, it is assumed that when the weight w is an initial weight vector having an arbitrary value, the operation of optimizing the local convex loss function (L(w)) is assumed. In this case, if the worker node uses gradient descents, it will randomly convert the weight w so that the loss function L(w) is zero. In this case, the weight w can be updated using w=w-a(dL(w)/dw). The evaluator node can evaluate the loss function L(w) to provide the highest compensation for the worker node that converged closest to zero.

The preceding description of the present disclosure is provided to enable those skilled in the art to make or use the present disclosure. Various modifications of the present disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to various modifications without departing from the spirit or scope of the present disclosure. Accordingly, the present disclosure is not intended to be limited to the examples described herein, but is intended to be given the broadest scope consistent with the principles and novel features disclosed herein.

100, 500: work request and work product trading system
110: client node
120: manager node
130: Evaluator node
140: worker node
150: blockchain network
300, 400: How to request work and trade results
510: host node
520: public storage node
700: Blockchain
710: block

Claims (20)

In the method of trading blockchain-based work requests and work results,
The blockchain includes a distributed ledger including a plurality of blocks including work request-the work request includes an evaluation request, a reward and a reward function-and data related to the evaluation of the work result and the work result. Management, and includes a plurality of nodes, each comprising at least one computing device,
The above method,
Sending a job request to the plurality of nodes by a client node among the plurality of nodes;
Sending an acceptance of the evaluation request to the client node, in response to the work request, by an evaluator node among the plurality of nodes;
Transmitting, by at least one worker node among the plurality of nodes, an acceptance of the work request to the client node in response to the work request;
Transmitting, by the worker node, the work result to the evaluator node;
Sending, by the evaluator node, an evaluation of the work outcome to the client node;
Sending, by the client node, an approval of the evaluation to the plurality of nodes; And
And transmitting, by at least one of the plurality of nodes, a compensation to at least one of the evaluator node, the worker node, and the client node based on the compensation function included in any one of the plurality of blocks. , Way.
According to claim 1,
The plurality of nodes further includes an administrator node, and the compensation function includes a compensation function for at least one of approval and rejection of a challenge request,
The above method,
Sending, by the client node or the worker node, a challenge request to the manager node; And
And sending, by the administrator node, approval or rejection of the challenge request to the plurality of nodes.
According to claim 2,
Sending, by the manager node, a job specification to the plurality of nodes;
Sending, by at least one of the plurality of nodes, an application for an evaluator node to the manager node;
Sending, by the manager node, an approval of an evaluator node to at least one of the nodes that have sent the request for the evaluator node;
Transmitting, by at least one of the plurality of nodes, an application for a worker node to the manager node; And
And sending, by the manager node, an acknowledgment of a worker node to at least one of the nodes that sent the request for the worker node.
The method of claim 3
Sending, by the administrator node, a secure runtime environment setting from the plurality of nodes to a host node; And
And sending, by the host node, the established secure runtime environment and application to at least one of the worker node and the evaluator node.
The method of claim 1
Further comprising the step of transmitting, by the worker node, the content of the work result to a common storage node, so that the address of the content is generated,
The work result transmitted to the evaluator node by the worker node includes the address of the content,
The data related to the work result included in the generated block includes second data regarding the address of the content.
The method of claim 5
Further comprising, by the common storage node, the contents of the work results are distributed and stored in a plurality of shards
The address of the content includes addresses for a plurality of shards.
The method of claim 1
The work request includes a learning work request by an artificial neural network,
The evaluation includes an evaluation based on a change amount of a loss function value according to a change in parameters of the artificial neural network.
In the method of trading blockchain-based work requests and work results,
The blockchain includes a distributed ledger including a plurality of blocks including work request-the work request includes an evaluation request, a reward, and a reward function-and data related to evaluation of the work result and the work result. Management, and includes a plurality of nodes, each comprising at least one computing device,
The above method,
Generating a block including identification information of the client node and data related to the job request when a job request is transmitted to the plurality of nodes by a client node among the plurality of nodes;
If an acceptance of the evaluation request is transmitted to the client node by an evaluator node among the plurality of nodes, in response to the operation request, a block including identification information of the evaluator node and data regarding acceptance of the evaluation request is generated. Generating;
When the acceptance of the work request is transmitted to the client node by at least one worker node among the plurality of nodes, in response to the work request, includes identification information of the worker node and data regarding acceptance of the work request. Generating a block to be played;
Generating, by the worker node, a block including identification information of the worker node and data on the work result when the work result is transmitted to the evaluator node;
Generating, by the evaluator node, a block including identification information of the evaluator node and data on the evaluation when an evaluation of the work result is transmitted to the client node;
Generating, by the client node, a block including identification information of the client node and data on the evaluation approval when the approval of the evaluation is transmitted to the plurality of nodes; And
When at least one of the plurality of nodes transmits a reward to at least one of the evaluator node, the worker node, and the client node based on the reward function included in any one of the plurality of blocks, the reward is paid And generating a block including the identification information of the node and data related to the compensation.
The method of claim 8,
The plurality of nodes further includes an administrator node, and the compensation function includes a compensation function for at least one of approval and rejection of a challenge request,
The above method,
Generating a block including identification information of the client node or the worker node and data regarding the challenge request when a challenge request is transmitted to the manager node by the client node or the worker node; And
And when the approval or rejection of the challenge request is transmitted to the plurality of nodes by the manager node, generating a block including identification information of the manager node and data regarding approval or rejection of the challenge request. , Way.
The method of claim 9,
Generating, by the manager node, a block including identification information of the manager node and data on the job specification when a job specification is transmitted to the plurality of nodes;
Generating, by at least one node among the plurality of nodes, an application for an evaluator node to the manager node, generating a block including identification information of the node and data regarding the application for the evaluator node;
Generating, by the manager node, a block including identification information of the manager node and data regarding approval of the evaluator node when an approval of the evaluator node is transmitted to at least one of the nodes that have transmitted the request for the evaluator node;
Generating a block including identification information of the node and data regarding the application of the worker node when an application of the worker node is transmitted to the manager node by at least one node among the plurality of nodes; And
Generating, by the manager node, a block including identification information of the manager node and data on the approval of the worker node when the approval of the worker node is transmitted to at least one of the nodes that have sent the request for the worker node. The method, which includes more.
The method of claim 10
Generating, by the manager node, a block including identification information of the manager node and data on the security runtime environment setting when a security runtime environment setting is transmitted from the plurality of nodes to a host node; And
When the set secure runtime environment and application are transmitted to the at least one node of the worker node and the evaluator node by the host node, a block including identification information of the host node and data regarding the secure runtime environment and application The system further comprising the step of generating.
The method of claim 8
When the content of the work result is transmitted to the public storage node by the worker node, and the address of the content is generated, a block including identification information of the worker node and first data regarding the address of the content is generated. Further comprising steps,
The work result transmitted to the evaluator node by the worker node includes the address of the content,
The data related to the work result included in the generated block includes second data regarding the address of the content.
The method of claim 12
Further comprising, by the common storage node, the contents of the work results are distributed and stored in a plurality of shards
The address of the content includes addresses for a plurality of shards.
The method of claim 8
The work request includes a learning work request by an artificial neural network,
The evaluation includes an evaluation based on a change amount of a loss function value according to a change in parameters of the artificial neural network.
In the transaction system of blockchain-based work requests and work results,
The blockchain includes a distributed ledger including a plurality of blocks including work request-the work request includes an evaluation request, a reward and a reward function-and data related to the evaluation of the work result and the work result. Management, and includes a plurality of nodes, each comprising at least one computing device,
The plurality of nodes,
A client node capable of transmitting a job request to the plurality of nodes, and generating a block including identification information of the client node and data regarding the job request when the job request is transmitted;
An evaluator node capable of transmitting an acceptance of the evaluation request to a client node, and generating a block including identification information of the evaluator node and data regarding acceptance of the evaluation request; And
And a worker node capable of transmitting an acceptance of the work request to the client node, and generating a block including identification information of the worker node and data regarding acceptance of the work request,
When the worker node transmits the work result to the evaluator node, the worker node generates a block including identification information of the worker node and data on the work result,
When the evaluator node transmits an evaluation of the work result to the client node, it generates a block including identification information of the evaluator node and data on the evaluation,
When the client node transmits the approval of the evaluation to the plurality of nodes, the client node generates a block including data on the evaluation approval including identification information of the client node,
If at least one node among the plurality of nodes transmits a reward to at least one of the evaluator node, the worker node, and the client node based on the reward function included in any one of the plurality of blocks, the reward payment node A system for generating a block including identification information and data related to the compensation.
The method of claim 15,
The plurality of nodes further includes an administrator node.
The compensation function includes a compensation function for at least one of approval and rejection of a challenge request,
When the client node or the worker node sends a challenge request to the manager node, the client node or the worker node generates a block including identification information of the client node or the worker node and data regarding the challenge request.
When the administrator node sends an approval or rejection of the challenge request to the plurality of nodes, the system generates a block including identification information of the administrator node and data regarding approval or rejection of the challenge request.
The method of claim 16
When the administrator node transmits a secure runtime environment setting to a host node among the plurality of nodes, a block including identification information of the administrator node and data regarding the secure runtime environment setting is generated,
The host node, when transmitting the set security runtime environment and application to at least one of the worker node and the evaluator node, blocks the block including identification information of the host node and data about the security runtime environment and application. Generated, system.
The method of claim 15
The worker node transmits the content of the work result to a common storage node, and when the address of the content is generated, generates a block including identification information of the worker node and first data regarding the address of the content,
The work result transmitted to the evaluator node by the worker node includes the address of the content,
The data related to the work result included in the generated block includes second data regarding the address of the content.
The method of claim 18
The common storage node distributes and stores the contents of the work result in a plurality of shards,
The address of the content includes addresses for a plurality of shards.
As a non-transitory computer-readable storage medium storing instructions for the blockchain-based work request and the transaction of work results,
The blockchain includes a distributed ledger including a plurality of blocks including work request-the work request includes an evaluation request, a reward, and a reward function-and data related to evaluation of the work result and the work result. Management, and includes a plurality of nodes, each comprising at least one computing device,
The instructions, when executed by a processor, cause the processor to:
An operation request is transmitted from the plurality of nodes to the plurality of nodes by a client node;
An acceptance of the evaluation request to the client node by an evaluator node among the plurality of nodes in response to the operation request;
At least one worker node among the plurality of nodes, in response to the job request, sending an acceptance of the job request to the client node;
Transmitting, by the worker node, the work result to the evaluator node;
Sending, by the evaluator node, an evaluation of the work outcome to the client node;
Sending, by the client node, an approval of the evaluation to the plurality of nodes; And
At least one node among the plurality of nodes is configured to perform an operation of transmitting a compensation to at least one of the evaluator node, the worker node, and the client node based on the compensation function included in any one of the plurality of blocks. A computer readable storage medium.

Images