How Trustworthy Computing Power Measurement Can Transform Cloud‑Native Services

As digital economies treat computing power like water and electricity, global capacity has surged to 650 EFLOPS by the end of 2022, and AI compute markets are projected to grow dramatically. This rapid diversification of compute resources creates a pressing demand for unified, standardized measurement methods to enable efficient allocation, optimization, and service‑oriented networking.

Overview of Computing Power Measurement Systems

The article first distinguishes three measurement perspectives:

Narrow compute‑network metrics : Focus on hardware processing units and system software platforms, separating compute and network metrics, which can lead to decoupled application awareness and sub‑optimal scheduling.

Broad compute‑network metrics : Extend narrow metrics by incorporating network and storage dimensions, providing a holistic view that aligns with modern multi‑resource workloads.

Telecom‑industry view : Highlights how major operators (China Unicom, China Telecom, China Mobile) propose their own grading and multi‑dimensional models, such as Unicom’s unified IT compute modeling, Telecom’s ITU‑T Y.2501 architecture, and Mobile’s “four‑sides‑three‑levels” framework covering compute, communication, memory, and storage.

Figures illustrate grading tables and network architectures proposed by these operators.

Technical Foundations for Trustworthy Measurement

The proposed solution combines a measurement suite, a WASM virtual machine, and blockchain technology:

WASM Virtual Machine : Provides cross‑platform execution, isolation, and bytecode‑level resource accounting. Unlike probe‑based or benchmark methods, WASM can precisely track resource consumption of each instruction during runtime.

Kubernetes Integration : The measurement suite is packaged as OCI‑compatible container images. During image build, application code is compiled to WASM via LLVM, then deployed with minimal changes to existing YAML manifests, preserving existing service, deployment, and volume configurations.

Measurement Workflow : The WASM VM loads and validates the bytecode, parses and executes instructions on a stack‑based model, and records a predefined Gas cost for each opcode. The total Gas consumption is aggregated and output after execution.

Figures depict the integration diagram, container image flow, and the Gas cost table for representative instructions.

Blockchain for Immutable Accounting

Blockchain introduces a decentralized, tamper‑proof ledger for recording measurement data. After each task, the collected Gas metrics are submitted to smart contracts, enabling transparent, consensus‑driven reconciliation between compute providers and network operators. Tokenization of Gas further clarifies ownership, improves traceability of data assets, and reduces settlement disputes.

Figures show the blockchain‑based accounting model and token management flow.

Trusted Measurement in Practice

A case study demonstrates deploying a face‑recognition application with the measurement suite and blockchain. The system tracks instruction‑level resource usage in real time, integrates seamlessly with cloud‑native environments, and provides fine‑grained metrics that support future runtime billing models.

Figures present the deployment architecture and the resulting compute‑power table for the face‑recognition workload.

Challenges and Future Directions

Two main challenges remain:

Absence of a defined conversion ratio between Gas units and CPU cycles limits broader applicability.

The current WASM instruction set does not fully cover specialized hardware such as GPUs and TPUs, restricting measurement for those accelerators.

Future research should focus on participating in standard‑setting initiatives (e.g., CCSA’s comprehensive compute evaluation framework) and exploring hybrid measurement approaches that combine WASM with FinOps or AI‑specific metrics to improve accuracy for heterogeneous hardware.

Conclusion

Integrating WASM and blockchain into compute‑power measurement offers precise, trustworthy, and cloud‑native compatible accounting, yet further standardization and hardware support are needed to fully realize its potential across the industry.