Ultra-Heterogeneous Distributed Computing Network Of UtilityNet

The next computing revolution will be based on heterogeneous computing.

In the field of AI computing, both the growth rate of data scale and the diversification of data forms far exceed the current processing and computing capabilities. Architecture innovation will be the main driving force of innovation in the next decade, and heterogeneous computing will bring exponential expansion effect.

There are mainly the following requirements in AI computing: multiple computing acceleration, time to market considerations, scalability, development difficulty, data bandwidth delay, price, training speed, real-time deduction, energy efficiency ratio and volume. The diversified computing architecture includes scalar, vector, matrix and space, which are applied to CPU, GPU, AI, FPGA and ASIC products respectively. Performance, compatibility and learning cost among different development frameworks are important factors that affect development efficiency. Complex development environment and frameworks that can't be updated synchronously cause developers to spend a lot of energy to solve problems by themselves. These all depend on the construction of ecological chain. The integration of solutions of all parties and the support of languages, compilers, frameworks and runtime are the challenges faced by UtilityNet.

UtilityNet aims to realize the integration of different hardware frameworks and software programming frameworks in the second stage, and realize the hyper-heterogeneous distributed computing network based on blockchain.