With the widespread application of blockchain technology across variousindustries, detecting and analyzing performance bottlenecks is crucialfor evaluating and optimizing blockchain system performance. How-ever, current research lacks general performance metrics for detectingand analyzing performance bottlenecks, and few studies focus on thisaspect within blockchain systems. To address this, this paper first pro-poses 18 fine-grained performance metrics to comprehensively evaluateperformance across various layers of blockchain systems. Subsequently,We introduce a generalized loosely-coupled performance measurementframework to capture these metrics and construct the causal relation-ship between them, i.e., the mesoscopic performance structure. Thisapproach allows for the detection and analysis of performance bot-tlenecks. Finally, numerous experimental results demonstrate that thecausality between the relevant performance metrics disappears whenthe system reaches a performance bottleneck. Additionally, the frame-work has a performance impact of less than 15% on ChainMaker.