AI Analysis Details Utilization Charts

#380640

Clicking on any row in the tabular report, in the upper pane, shows the utilization details in the Utilization Charts, in the lower pane.

If the lower pane has been collapsed, click the icon in the upper left corner to expand the pane.

Figure: Expanding the Lower Pane

Figure: Locating the AI Analysis Details Tab Charts

Utilization charts show hourly min/max and sustained activity for the selected system. The charts show the current and the recommended settings for selected workloads.

Use the left/right arrows to scroll through the workload charts. The following charts are available:

CPU Utilization in mCores - Busiest Container (Current and Recommended Settings)
CPU Utilization in mCores - Container Average (Current and Recommended Settings)
Memory (Total) Utilization - Busiest Container (Current and Recommended Settings)
Memory (Total) Utilization - Busiest Container - Container Average (Current and Recommended Settings)
No. of Restarts
In Service Instances

Disk I/O Bytes
Memory (RSS) Utilization - Busiest Container (Current and Recommended Settings)
Memory (RSS) Utilization - Container Average (Current and Recommended Settings)
Memory (Working Set) Utilization (Current and Recommended Settings)
Memory (Working Set) Utilization - Container Average (Current and Recommended Settings)

Note: If the required metrics are not available, the corresponding charts are not displayed.

Using the Utilization Charts

Utilization charts display current sustained activity in blue. The minimum and maximum utilization values are shown as vertical lines above and below the coloured block. The median is indicated as a grey line within the coloured block.

Figure: Utilization Chart Data

(hover over image to expand)

Figure: Utilization Chart Request/Limit

(hover over image to expand)

The Y-axis is scaled dynamically to match the range of data. The orange line indicating the current request setting is shown and the setting of the Y-axis scale also accounts for the request setting. The pink line indicates the current limit. If the request and the limit are the same, the dashed orange/pink line indicates the value. The units of the workload are also shown on the Y-axis.

Recommended CPU and memory settings are shown as dashed lines. Refer to the legend located below the charts for details.

Hover over any time or date on the current or recommended lines and click to see the detailed data.

An icon in the corner of each chart provides a zoom feature:

You can zoom in on the chart for easier viewing. All information indicated above is provided so you can review hourly data for the selected container. Arrows on each side of the selected chart allow to scroll through the other available charts.

Figure: Utilization Charts - Current and Recommended Values

Using the Legend

You can select or de-select the data displayed on each chart using the legend located below the charts.
Click an option to add or remove it from the displayed charts.

Figure: Utilization Charts - Using the Legend

Viewing the Other Charts

You can set the number of charts to view using the options in the lower right corner of the page. Select 1 to 4 charts to be displayed. Use the dots to the left of this selector to navigate left/right to see the other charts.

Video: Chart Selection

Expanded Modal View

Clicking the zoom button opens the chart in a modal view with additional options to view the collected and analyzed data.

The following options allow you to view all details your data:

Table: Modal View Data Aggregation Options

Option	Description
ML Model	The data is aggregated hourly for the last day (yesterday) and displayed using quartile data.
Historical Daily	The container's data, is aggregated for a single day. You can see the minimum, maximum, sustained and average values by clicking on any data point. Use the slider at the bottom to select the historical days of data to review. 90 days of data is retained.
Historical Hourly	The container's data, aggregated for a single hour. Like the historical daily view, you can see the minimum, maximum, sustained and average values by clicking on any data point. Use the slider at the bottom to select the range of data to view.
Historical (5 Min)	The container's data aggregated into 5-minute samples. This option is available for CPU, Working Set Memory, RSS Memory, Total Memory, and Restarts.

Using Policy Settings to Select Utilization Metrics

The Kubernetes Optimization policy settings allow you to analyze requests and limits based on different utilization metrics, e.g. Memory request analyzed based on Actual Memory Utilization and Memory Limit analyzed based on Raw Memory Utilization. If these policy settings have been enabled the corresponding charts will indicate the selected metric settings.

Contact Support@Densify.com for details on enabling these settings.

Understanding Memory Metrics

Densify collects 3 memory metric types from cAdvisor:

container_memory_usage_bytes

Provides the size of the overall memory used, regardless of whether it is used for mapping from disk or for memory allocation.
Includes both RSS (memory not used to map files from the disk) as well as the page cache (uses memory exclusively to map files from the disk), so in effect this metric counts memory access indiscriminately of its purpose.
Measures the current memory usage, including all memory regardless of when it was accessed.

Table: container_memory_usage_bytes

Usage

Description

Kubex Workload metric name:

Memory Utilization

Kubex uses this metric for:

Default policy settings for Requests and Limits:

Requests uses average container
Limits uses busiest container

container_memory_working_set

When this metric hits the MEM limit an OOM kill occurs.
WSS = Working Set Size, also called, current working set.
Represents the minimum amount of memory a process needs to keep running during a period of time.
Includes the active files cache.
Unlike fixed metrics like "cache" or "buffer", which have specific values in our operating systems, the working set is a calculated value determined in various ways.
cAdvisor calculates the working set as follows: total memory usage - inactive files
It includes all anonymous (non-file-backed) memory since Kubernetes does not support swap.
The metric typically also includes some cached (file-backed) memory, because the host OS cannot always reclaim such pages.
container_memory_usage_bytes is ≥ container_memory_working_set.

Table: container_memory_usage_bytes

Usage	Description
Workload metric name:	Memory Working Set
Kubex uses this metric for:	Not currently used. It will be used for detecting hidden process kills, in a future release.

container_memory_rss

Represents the amount of physical memory that a process has in the main memory that doesn’t correspond to anything on disk.
Typically includes stacks, heaps, and anonymous memory maps.
Size of memory not used for mapping files from the disk
Number of bytes of anonymous and swap cache memory (includes transparent pages).
Container_memory_rss container_memory_working_set