How to Call C-end Computing Power
Last updated
Last updated
To effectively utilize the computing power provided by C-end nodes, the DMCN protocol implements a slot-based allocation mechanism. The number of slots assigned to each node is determined by its hardware capabilities and network conditions. Specifically, DMCN considers the following factors:
CPU Processing Power: Nodes with more powerful CPUs can handle more complex tasks and are allocated more slots accordingly.
Network Bandwidth: Nodes with higher network bandwidth can transfer data more efficiently and are allocated more slots to facilitate faster task completion.
Operating System: Different operating systems may have varying levels of efficiency in managing computational resources. DMCN takes this into account when allocating slots to ensure optimal performance across diverse node configurations.
When a B-end user submits a computing task, Robot Smarter receives the task and breaks it down into smaller units. This approach draws inspiration from Google's MapReduce framework. Once the task is divided into units, Robot Smarter estimates the computing cost and duration and provides this information to the B-end user. Robot Smarter is responsible for splitting, merging, and monitoring the completion of the task.
Upon B-end user confirmation, Job Manager distributes the task units to available slots for execution. Unlike traditional data centers, decentralized computing cannot guarantee that individual nodes remain operational for extended periods. Task splitting effectively addresses this challenge. Job Manager monitors node heartbeats, indicating whether nodes are in a working state. Node scheduling and task distribution follow the proximity and idle-first principles. C-end users earn rewards based on the formula: (unit price of task units) * (number of units).
For large-scale computing tasks, DMCN utilizes the IPFS interface to achieve decentralized storage of intermediate data. DMCN only performs cache calculations on the CPU and does not store any data. It operates in a sandbox environment. Therefore, storage solutions like IPFS are external options, and users can freely choose where to store their data (locally or on the cloud).
