Greetings NuNetopians,

NuNet is worked up to announce that on NuNet will release our Public Alpha Testnet to our community. We will probably be launching multiple testing campaigns to check individual components with our community and will probably be rewarding testers for his or her participation, we are going to release more information on staged testing tomorrow with our official launch.

As previously announced, we will probably be releasing Public Alpha with our Decentralized ML use case which can allow users to onboard each latent CPU and GPU resources to be utilized by service providers to run compute jobs on the NuNet platform and be compensated in NTX (NuNet’s Utility Token) for the job.

Before jumping into the technical elements, we wanted to clarify the aim of releasing on test net and the core elements of the Public Alpha to raised help our community understand the method that will probably be explained.

As mentioned previously we will probably be launching Public Apha with Cardano Testnet integration. Our primary focus is a secure and robust platform which we plan to achieve with the assistance of the community and auditors. Specifically with this testnet release we’re constructing our documentation site and developer tooling pipeline in an effort to enable community testers to contribute most to the discharge Public Alpha on the mainnet.

The below table outlines the testing phases:

NuNet will launch Test Net Public Alpha using test NTX tokens. We’ve got implemented a wise contract on the PreProd Cardano Network to lock service provider NTX funds and reward compute providers for the usage of their resources.

Service Provider

In NuNet, the term ‘service provider’ refers back to the individual or group that wishes to provision compute jobs on NuNet’s Decentralized Community hardware. It could be an application, application developer or an independent use-case developer which provides access to most of the people to make use of its service that uses NuNet as an infrastructure to support its compute requirements.

Compute Provider

The term ‘compute provider’ refers to a person who has on-boarded their device onto the NuNet platform in an effort to provide computing power.

Oracle

​​NuNet Oracle will probably be liable for validating results of compute jobs and solving disputes with NTX.

DMS

The term DMS refers to ‘Device Management Service’, which is basically the NuNet platform itself. It’s the lightweight binary which establishes secure peer-to-peer connection between all computers on boarded onto the network.

Service Provider Dashboard

We seek advice from the Service Provider Dashboard because the interface through which service providers can deploy jobs on NuNet. NuNet Public Alpha on Testnet comes with one specific use-case for testing — ML on GPU and CPU. The provided Service Provider Dashboard implements user interface needed for deployment of related computing jobs via NuNet.

Compute Providers Dashboard

That is where providers can claim their tokens for work carried out on their devices.

The diagram below shows the sequence of the GPU ML (machine learning) use case, which is the first use-case for the Public Alpha Testnet. The business goal of implementing this sequence is to permit a user to run a ML service in a NuNet machine and pay NTX tokens to the machine provider for the computational resources used to execute the ML service.

the service provider, the next sequence will probably be used to run a job on the NuNet Platform. It’s broken into 5 sections to further help understand the method:

Section 1: User Requests to Run a Compute Job — Service API

1. The service provider will fill out the job information on NuNet Service Provider Dashboard and can subsequently request to run a compute ML job and submit all parameters.
2. The ML user authenticates on a Web UI, then inserts the URL of the ML model project link — PyTorch/TensorFlow (preferably a GitHub/GitLab/Git based repository link).
3. If the dataset for use with the ML project is stored individually, its URL could be specified as described within the ML code contribution workflow within the above example.
4. The ML user authenticates on a Web UI, then inserts the estimated time required for the ML job to run.
5. The service provider will then connect their Eternl or Nami Wallet and can determine the max amount of NTX tokens to be paid on the job through the Cardano network.

Section 2: Find an acceptable resource on NuNet — Device Management API

1. A request is shipped from the service provider’s DMS to the DMS on the compute provider’s machine to examine whether it is suitable.
2. The service provider is given an inventory of machines that may execute their job. This list is prioritized by soft constraints and selects essentially the most suitable device.
3. If no suitable device is found, the DMS will return a message to the service provider with a proof and can stop the sequence.

Section 3: Calculating the Price (COMPUTE API)

1. If there may be a connection, the DMS on the service provider’s machine will on the of the
2. This price is then declared on the compute provider’s machine.
3. The value is then returned to the service provider’s DMS and checks if the worth matches the overall max price allowed by the service provider.
4. If the worth is outside of the max price allowed by the service provider an error message will probably be sent to the service provider with a proof.

*Disclaimer although this is a component of the complete system we is not going to be implementing it for Public Alpha Testnet Testing. We’re only coping with rewards in successful cases. Price declaration will come after testnet launch.

Calculated price outside bounds

Calculated price inside bounds

1. If the worth is contained in the bounds of NTX offered, all information on price will probably be sent from the service provider’s DMS to the compute provider’s DMS.
2. This data can be returned to the Service Provider Dashboard to let the service provider know there was an acceptable connection.
3. The service provider will connect their wallet on the Service Provider Dashboard and sign the smart contract.
4. The NTX tokens will probably be held within the smart contract until the job has been accomplished.

Section 4: Running the Compute Job on Compute Hardware (DEVICE MANAGEMENT API)

1. Once NTX tokens have been held in escrow and the contract signed, the service provider’s DMS will routinely push the job to the compute provider’s machine.
2. The compute provider’s machine will routinely download the docker image containing the service.
3. The service will routinely run.
4. The telemetry data from the compute provider’s DMS will probably be sent to the Stats Database.

Section 5: Release funds to workflow constituents (TOKENOMICS API)

• Once the job has run, the compute provider’s DMS will confirm the outcomes on the Stats Database to find out if the job was successfully run.
• Funds within the escrow will probably be released for the machine which did the computational work as agreed if the work was verified appropriately.

• If the job has been unsuccessful, the funds will probably be returned to the service provider’s wallet with an error message.

• If the job was partially accomplished the % of tokens will probably be 1.) Returned to service provider for work not accomplished and a pair of.) Given to the compute provider for work accomplished.

• Disclaimer although this is a component of the complete system we is not going to be implementing it for Public Alpha Testnet Testing. We’re only coping with rewards in successful cases. Refunding in case of error or partial return usually are not yet implemented.

We look ahead to working with the community to check this technique in the approaching days.

NuNet currently has plenty of open positions for various roles throughout the team. If you could have the abilities and desire to hitch us in our journey, yow will discover more information and make contact with us through our profession page.

NuNet lets anyone share and monetize their computing resources, turning cloud computing power from a centralized service into an open protocol powered by blockchain. Discover more via: