Analysis of Server Resource Consumption of Meteorological Satellite Application System Based on Contour Curve

Xiangang Zhao, Cunqun Fan, Manyun Lin, Lizi Xie, Xi ZhangAdvances 2018 · DOI

Abstract

In order to effectively analyze system server resource consumption, this paper proposes a server resource consumption analysis method based on the "contour curve." The methodology defines resource consumption analysis through several lenses: application resource consumption profiles, standard deviation profiles, frequency analysis, and monotonicity testing. By evaluating total consumption, mean values, kurtosis, skewness, and resource load, this framework provides a deep-dive into server behavior. The proposed method aims to improve server resource utilization and reduce hardware loss in high-demand meteorological satellite application systems.

Introduction

Meteorological satellite ground systems are data-intensive environments responsible for the continuous reception, processing, and transmission of massive data streams. This operational pressure places immense demand on server resources, including CPU, memory, I/O, and networking.

The primary challenge in managing these systems is that standard monitoring—which often relies on simple averages—can mask transient spikes, structural inefficiencies, or gradual performance degradation. To optimize these systems and prevent premature hardware wear-and-tear, engineers require a method to characterize the "statistical fingerprint" of an application’s resource usage over time. This paper introduces the "contour curve" framework to bridge the gap between raw telemetry and actionable system health insights.

Methodology: The Contour Curve Framework

The core innovation of this research is the Application Resource Consumption Profile, or the "contour curve." Rather than analyzing a single execution instance, the authors propose scheduling an application $N$ times and calculating the mean resource value at every discrete time point $j$ across all runs. This approach filters out stochastic noise and reveals the inherent resource demand pattern of the application.

Multi-Dimensional Analytical Layers

Beyond basic profiling, the paper introduces several sophisticated statistical layers to diagnose system health:

  • Standard Deviation Profiling: This measures the fluctuation of resource usage between different runs at the same time point. High standard deviation indicates unstable application behavior or unpredictable environmental interference.
  • Frequency Domain Analysis: By applying a Real Sequence Fast Fourier Transform (rFFT) to the contour curve, the system can identify periodicities or hidden instabilities in resource demand that are not visible in the time domain.
  • Monotonicity Testing: The framework employs the Mann-Kendall algorithm to detect if resource usage at a specific time point shows a consistent upward or downward trend across successive runs. This is a critical indicator for identifying issues like memory leaks or aggressive garbage collection cycles that worsen over time.
  • Higher-Order Statistics: The authors utilize Kurtosis to identify "bursty" consumption patterns and Skewness to categorize application behavior. A positive skewness indicates a "slow-start" (right-biased) application, while negative skewness suggests a "quick-and-slow" (left-biased) profile.

Resource Load and Health Metrics

The paper defines a weighted Resource Load metric that penalizes high-occupancy states more heavily than low-occupancy ones. This provides a more nuanced view of system stress than a raw average, as sustained high usage is more damaging to hardware longevity than intermittent spikes.

Based on these metrics, the authors establish heuristic thresholds for a "healthy" satellite application:

  • CPU Health: Profile average < 60% and a "CPU Load" score < 2000.
  • Memory Health: Profile average < 75% and a "Memory Load" score < 3000.
  • Consistency: A standard deviation mean near 0, indicating predictable behavior across every execution.

Discussion and Insights

One of the most significant insights of this framework is the application of the Mann-Kendall test across runs rather than within a single time series. If the $j$-th second of an application consistently consumes more CPU in the 50th run than in the 1st, it reveals a hidden state or resource exhaustion issue that traditional monitoring would likely miss.

Furthermore, the use of skewness for scheduling offers practical benefits for system administrators. By identifying whether an application is "slow first and then urgent," administrators can better interleave tasks to ensure that multiple "urgent" phases do not hit the hardware simultaneously, thereby leveling the total load on the server.

Conclusion and Future Work

The contour curve method allows for an in-depth analysis of server resource consumption, providing a quantitative basis for improving utilization and extending hardware life. While the mathematical framework is robust, the current study is primarily methodological.

Future research should focus on empirical validation through large-scale case studies to quantify specific performance gains. Additionally, automating this analysis into a real-time dashboard and applying it to cloud-native environments—where resource boundaries are more fluid—would represent a significant step forward in automated system operations (AIOps) for meteorological satellite systems.

Summary updated

Abstract

In order to effectively analyze the system server resource consumption, this paper proposed a server resource consumption analysis method based contour curve. Firstly, the definition of resource consumption analysis was given, and then several analysis methods of resource consumption were given, including: application resource consumption profile, application resource consumption profile standard deviation, application resource consumption frequency, application resource consumption monotonicity and so on. Finally, based on this, the total consumption, consumption mean, kurtosis, skewness, reload, resource load and so on were analyzed. The method proposed in this paper can effectively analyze the server resource consumption in depth, improve the server resource utilization and reduce the server hardware loss to a certain extent.
BibTeX
@inproceedings{zhao2018analysis,
  author = {Xiangang Zhao and Cunqun Fan and Manyun Lin and Lizi Xie and Xi Zhang},
  title = {{Analysis of Server Resource Consumption of Meteorological Satellite Application System Based on Contour Curve}},
  booktitle = {Advances},
  year = {2018},
  doi = {10.23977/acss.2018.21004},
}
Figure