Hotspot Comparison Analysis

更新时间:
复制 MD 格式

When application performance fluctuations cause system metrics to become abnormal, use time series charts and differential flame graphs to identify differences in process hotspots over time. This topic describes how to use the Hotspot Comparison Analysis feature.

Use cases

  • Compare hotspots on a single instance across different time periods.

  • Compare hotspots between two instances during the same or different time periods.

Limitations

  • Region restrictions

    This feature is currently available only in the Chinese mainland and China (Hong Kong).

  • Supported operating systems

    Architecture

    Operating system

    x86 architecture

    • Rocky Linux 9.5

    • Rocky Linux 9.1

    • Ubuntu 20.04

    • Alibaba Cloud Linux 3 Container-Optimized Edition

    • Rocky Linux 8.8

    • Ubuntu 22.04

    • Alibaba Cloud Linux 3 Pro

    • Alibaba Cloud Linux 2/3

    • CentOS 7.6 or later, or CentOS 8

    • Anolis OS 7/8

    • Ubuntu 24.04

Prerequisites

  • If you are using a RAM user, ensure that the Alibaba Cloud account (main account) has granted the AliyunECSReadOnlyAccess and AliyunSysomFullAccess system policies to the RAM user.

  • The Operating System Console service must be enabled.

    When you log on to the Operating System Console for the first time, click Enable Service.

  • The SysOM component is installed. For installation instructions, see Component Management.

Procedure

  1. Go to the Operating System Console.

  2. Install SysOM on the target ECS instance.

  3. In the left-side navigation pane, choose System Observation > Hotspot Comparison Analysis.

  4. Select the search criteria and click Perform Comparative Analysis.

    Note
    • To compare hotspots for a specific process, select the process in the PID (Optional) field. Otherwise, the analysis compares all processes on the two selected instances.

    • The supported hotspot types are ONCPU, memory, and lock. To compare memory or lock hotspots, you must configure Java hotspot tracking for the instances.

    image

Result analysis

  1. Analysis results

    A large model generates the analysis results. The results compare call stack hotspots between the two processes and provide optimization suggestions.

    image

  2. Process hotspot time series chart

    • This time series chart shows the change over time in ONCPU sample values for the selected processes, helping you quickly identify critical hotspot moments.

      image

    • You can select a time range in the panel's time series chart. Drag to select two time ranges that contain valid samples, and then click Perform Comparative Analysis again. The comparison results are regenerated based on the new time ranges.

      image

  3. Differential flame graph for function hotspots

    The differential flame graph provides three modes: Flame Graph, Function Table, and Both. These modes help you find hotspot differences: within a process over time, between the same process on different instances, or between different processes.

    1. Flame graph mode

      The flame graph shows the call stack hierarchy and time distribution. You can click to focus on specific parts of the call stack. In the graph, functions marked in red indicate an increase in the sample count for Process 2 compared to Process 1 during the selected time period. Functions marked in green indicate a decrease.

      image

    2. Function table mode

      The function table displays each function's name, its hotspot percentage in Process 1 and Process 2, and the difference between them. You can reorder the function rows based on the hotspot difference.

      image

    3. Both mode

      In this mode, clicking a function row in the function table also highlights the corresponding row in the flame graph.

      image