Shulou(Shulou.com)06/03 Report--

This is the second in a series of articles on monitoring data visualization. This article focuses on the summary picture.

In the first part of this series, we discussed time series diagrams-visualization of how infrastructure metrics change over time. In this article, we will introduce summary diagrams that flatten out specific time periods to provide visualization of the infrastructure summary window:

Single value summary ranking list change chart host map distribution version

For each graphic type, we will explain how it works and when it is used. But first, we will quickly discuss the two concepts necessary to understand the infrastructure summary diagram: cross-temporal aggregation (which you can think of as "time flattening" or "snapshot") and cross-spatial aggregation.

Cross-time summary

In order to provide a summary view of metrics, visualization must flatten the time series to a single value by compressing the time dimension out of sight. This cross-time aggregation may mean displaying only the latest values returned by the metric query, or more complex aggregations to return calculated values within the moving time window.

For example, instead of displaying the latest reported values for metrics queries, you might want to display the maximum values reported by each host in the past 60 minutes to address problematic spikes:

[

Cross-spatial aggregation

Not all metrics queries are meaningful and can be divided by host, container, or other infrastructure unit. Therefore, you usually need to do some aggregation across spaces to create metric visualization that reasonably reflects your infrastructure. This aggregation can take many forms: aggregate metrics through message queues, database tables, some properties of the application or host itself (operating system, availability zone, hardware profile, etc.).

Cross-spatial aggregation allows you to slice and segment the infrastructure to accurately isolate metrics from observable critical systems.

It may be more useful to view the peak latency for each internal service built on Redis than the host-level peak latency listed in the example above. Alternatively, you can only display the maximum reported by any host in the infrastructure:

! [Redis latency graph] Cross-spatial aggregation: grouping hosts by service name (top) or compressing the list of hosts into a single value (bottom)

Cross-spatial aggregation in time series diagrams is also useful. For example, it is difficult to understand the host-level chart of an Web request, but when summarizing metrics by availability zone, you can easily interpret the same data:

! [Redis delay graph] aggregation from never aggregated (line chart, top) to cross-space (stacked area map, bottom)

The main reason for marking metrics is to enable cross-spatial aggregation.

Single value summary

A single-valued summary uses a conditional format, such as a green / yellow / red background, to display the current value of a given metric query to convey whether the value is within the expected range. The values displayed in the single-valued summary do not have to represent instantaneous measurements. The widget can display the latest values of the report, or the summary values calculated based on all query values throughout the time window. These visualizations provide a narrow but clear window for your infrastructure.

[

When to use a single-valued summary why a given system's working indicators make key indicators immediately visible Web server requests key resource indicators overview of resource status and health load balancers healthy host error indicators quickly cause concern for potential problems fatal database anomalies compared to previous values The calculated metrics change clearly conveys the key trends of hosts in use compared to Toplists a week ago.

A ranking is an ordered list that allows you to rank hosts, clusters, or any other network segment of the infrastructure by their indicator values. Because they are easy to interpret, top-level lists are particularly useful in the advanced status panel.

Compared to a single-valued summary, the top list has an additional aggregation layer in space because the values of the metric query are divided by group. Each group can be a single host or a collection of related hosts.

[

When to use the list what why work or resource metrics from different hosts or groups are clearly found at a glance, people with poor performance or excessive consumption of resources the integral custom metrics processed by each application server are returned as a list of values to communicate the Datadog agent version change chart Change graphs in use by KPI (for example, for status boards on wall-mounted monitors) in an easy-to-read format

The top list provides you with a summary of the values of the most recent metrics, while the change chart compares the current values of the metrics with their values at some point in the past.

The main difference between a change diagram and other visualization is that it takes two different time ranges as parameters: one for evaluating the size of the window and the other for setting the backtracking window.

[

When to use change diagrams and why why cyclical metrics rise and fall daily, weekly or monthly to separate indicator trends from regular benchmarks database write throughput, compared with the same period last week [advanced infrastructure indicators quickly identify large-scale trend host total, compared with yesterday's same period [mainframe map Host maps

The host map is a unique way to see the entire infrastructure or any part of it at a glance. However, if you slice and slice the infrastructure (by data center, by service name, by instance type, and so on), you will see that each host in the selected group is hexagonal and color-coded and resized according to any metrics reported by these hosts.

This particular type of visualization is unique to Datadog. In this way, it is specifically designed for infrastructure monitoring, as opposed to the general visualization described elsewhere in this article.

[

When to use host maps what why resource utilization metrics clearly identify overloaded components the load of each application host (grouped by cluster) [identify misallocation of resources (e.g. Whether any instance is too large or too small) CPU utilization of each EC2 instance type [error or other work metrics quickly identify degraded host HAProxy 5xx error of each server [related metrics * View the throughput and memory used of the relevant application server in a single diagram [release version]

The distribution map shows a histogram of the metric values across the infrastructure portion. Each bar chart in the chart represents a range of merged values whose height corresponds to the number of entities within which values are reported.

The distribution map is closely related to the heat map. The main difference between the two is that the heat map shows changes over time, while the distribution is a summary of the time window. Like heat maps, distributions can easily visualize a large number of entities that report specific metrics, so they are often used to map metrics at a single host or container level.

[

When to use the distribution, why, a single indicator reported by a large number of entities clearly conveys the overall health of each host's network latency [view differences among team members, uptime of each host]

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.