Google Cloud performance use cases

The following sections describe various ways that Performance Dashboard can help you. This page describes Google Cloud performance use cases. For project performance, see Project performance use cases.

Performance diagnostics: Is it the project or Google Cloud?

You observe latency or packet loss values in your project, and you want to validate and compare against reference data outside of your projects. The Performance for all of Google Cloud view in Performance Dashboard lets you correlate per-project network performance, such as VM-to-VM packet loss or latency, to performance over all of Google Cloud, to show packet loss and latency across all relevant Google Cloud infrastructure.

In the following example, you can see that Google Cloud traffic experienced packet loss recently. By holding the pointer over the chart, you can see that the peak loss occurred between asia-east1-a and asia-northeast3-b at 2:26 PM. If your project's issues correspond to this time and location, you save troubleshooting time because the issues impact all of Google Cloud.

Packet loss summary.
Packet loss summary (click to enlarge)

Holding the pointer over a time series shows data for the source and destination. To show the chart legend, you click the Show Legend button on the time series chart. Holding the pointer over a zone pair in the legend highlights the time series data for that pair.

Packet loss time series highlights.
Packet loss time series highlights (click to enlarge)

To see all the areas where the packet loss issue occurred at 2:26 PM, you update the heatmap by moving the time selector on the time series chart.

Average packet loss.
Average packet loss (click to enlarge)

Scenario: Validate if the project performance differs from Google Cloud

Using the per-project Performance Dashboard in the Network Intelligence Center, you identified a region or zone pair with high latency or packet loss. You want to see if they differ significantly from the Google Cloud average.

Hold the pointer over the region or zone pair that you want to validate. You can see the value of latency or packet loss for your project and the value of the Google Cloud average for the same pair underneath. If they are similar, it indicates that the performance observed in your projects is similar to the average across all projects within the same region or zone pair.

Average packet loss in regions.
Average packet loss in regions (click to enlarge)

Scenario: Investigate current high latency or packet loss issue

You identified the region or zone pairs with high packet loss or latency, and you want to compare those with the overall Google Cloud values. You can refine the Google Cloud performance view of Performance Dashboard by using the source and destination region selection capability. There are two ways in which you can refine the Google Cloud performance view:

  • To view data for specific regions, you select the regions from the Locations menu. You can select up to five regions.

    Latency in regions.
    Latency in regions (click to enlarge)
  • To view zone-to-zone data for a region pair, you click the square in the regional heatmap that corresponds to the region pair.

    Regional heatmap.
    Regional heatmap (click to enlarge)

With both selection methods, the time series data and heatmap for the selected pairs are displayed. In the following example, you selected asia-east1 and europe-north1.

Timeseries and heatmap for selected pairs.
Timeseries and heatmap for selected pairs (click to enlarge)

Scenario: Investigate historic high latency or packet loss issue

If you want to investigate a specific time slot in which you observed abnormal latency or packet loss values, you can use the time zoom in the time series. In this example, you have selected 30 days.

Latency summary.
Latency summary (click to enlarge)

You can further refine the time selection by clicking and dragging on the time series chart to select the days or hours related to your investigation.

The chart displays the data for Monday the 9th through Friday the 13th. You can pinpoint a modest spike in latency on November 11 at 8:15 AM.

Latency summary time zoom.
Latency summary time zoom (click to enlarge)

Workload optimization: Planning for performance

The Google Cloud performance view provides an average performance view for selected region/zone pairs. You can use the Google Cloud performance view in the Network Intelligence Center in the Google Cloud Console to plan the performance management of your applications. The Google Cloud performance view helps you to identify the best regions and zones to deploy new resources or to optimize your existing workloads. For example, you are currently running in specific zones and regions, but you know they are accessed by users or other workloads from other regions.

Scenario: Understand performance in new zones and regions

You want to add workloads to a region, and you want to know which zone best serves your requirements.

You use the source and destination region selection menu to select source-destination pairs where you want to see the global average metrics for latency or packet loss.

To view the zones within a region pair, you select the region pair from the heatmap. In this example, you clicked the heatmap square for the region pair asia-east1 and asia-northeast3 and adjusted the time selector to 30 days.

Packet loss region pair heatmap.
Packet loss region pair heatmap (click to enlarge)

Scenario: View historical data for regions or zone pairs

Use the time series chart capability to display historical data for selected zone pairs. In this example, you can see the heatmap square for the zone pair asia-east1-a and asia-northeast3-a and the time selector adjusted to 30 days.

Zone pairs historical data.
Zone pairs historical data (click to enlarge)

What's next