Basic Computing Operations: A Metric for User Computing Consumption in the Multi-Cloud

Multi-cloud computing, aggregating numerous heterogeneous resources, is increasingly common and, as a form of utility computing, offers services in a "pay-as-you-go" mode, where users pay for actual consumption. Existing leasing-time methods, however, fail to measure consumption accurately. Since resources are consumed through program execution, any valid metric must reflect it. We propose the metric, Basic Computing Operations (BCOPs), built on the RAEM-Machine model, which extends the classical RAM by characterizing consumption as two primitive operations: data calculations and data movements. We define BCOPs on programs that users execute as input by counting basic functions, and then design the theoretical model of a RAEM-Machine to formalize the program execution and develop a tool for measuring the BCOPs metric on real computers. Experiments confirm its effectiveness, with IR-level measurements showing less than 7.3% error compared to source-code level evaluation. BCOPs are further extended to BCOPS (BCOPs per second) for performance analysis, revealing, for example, that Matmul achieves 2.8 times higher efficiency than Sort.