Metric

Metric

Metrics are data about system performance. By default, many services provide free metrics for resources (e.g., Virtual Server, File Storage, etc.), which are offered as basic monitoring through ServiceWatch. Detailed monitoring can be used for certain resources such as Virtual Server.

Indicator data is retained for 15 months (455 days), so you can view both the latest data and historical data.

Namespace

A namespace is a logical separation used to distinguish and group ServiceWatch metrics. In Samsung Cloud Platform services, namespaces are generally the same as the service name, and can be found in the ServiceWatch 연계 서비스 목록.

For custom metrics, users can define a namespace in ServiceWatch to distinguish them from other metrics, and can define it via the ServiceWatch Agent settings or OpenAPI. Detailed information about custom metrics and logs can be found in 사용자 정의 지표 및 로그.

Metric (metric)

A metric represents a set of data points sorted chronologically as they are collected in ServiceWatch. Each data point consists of a timestamp, the collected data, and the unit of the data.

For example, the CPU utilization of a specific Virtual Server is one of the basic monitoring metrics provided by Virtual Server. The data point itself can be generated by any application or activity that collects data.

By default, the Samsung Cloud Platform services integrated with ServiceWatch provide resource metrics for free. Detailed monitoring for some resources is offered as a paid service and can be enabled in each service.

Metrics can only be viewed in the region where they were created. Metrics cannot be arbitrarily deleted by users. However, if new data is not posted to ServiceWatch, they will automatically expire after 15 months. Data points older than 15 months (455 days) expire sequentially, and when new data points are added, data older than 15 months (455 days) is deleted.

Timestamp

The timestamp of a data point is the time information indicating when the data point was recorded. Each metric data point consists of a timestamp (time) and data.

A timestamp consists of hours, minutes, seconds, and a date.

Metric retention period

We maintain ServiceWatch metric data as follows.

• Data points with a collection interval set to 60 seconds (1 minute) are available for up to 15 days.
• Data points with a collection interval set to 300 seconds (5 minutes) are available for up to 63 days.
• Data points with a collection interval set to 3600 seconds (1 hour) are usable for up to 455 days (15 months).

Data points that were initially collected at a short interval are downsampled and stored for long-term retention.

For example, if data is collected at a 1‑minute interval, it is retained in 1‑minute granularity for 15 days. After 15 days, the data continues to be retained but can only be queried in 5‑minute intervals. After 63 days, the data is re‑aggregated and provided in 1‑hour intervals. If you need to retain metric data points longer than the metric retention period, you can archive them separately using the File Download or Object Storage Export features.

Dimension(Dimensions)

A key-value pair that serves as a unique identifier for a metric, allowing you to classify and filter data points.

For example, you can identify metrics for a specific server by using the resource_id dimension of the Virtual Server metrics.

Collection interval

It refers to the interval for collecting data points for each service’s metrics and is provided according to the collection interval predefined by each service.

Refer to each service’s ServiceWatch metrics page for the metric collection interval.

For example, Virtual Server provides a collection interval of 5 minutes during basic monitoring, and a 1‑minute interval when detailed monitoring is enabled.

Statistics

Statistics are a method of aggregating metric data over a specified period. ServiceWatch provides data aggregated as statistics based on metric data points supplied to ServiceWatch from each service. Aggregation is performed within the specified aggregation period using namespace, metric name, dimension, and data point units.

The provided statistics are sum, average, minimum, maximum.

• Total: sum of all data point values collected during the period
• Average: During the specified period, (sum of all data pointer values during that period) / (number of data pointers during that period) value
• Minimum: the lowest value observed during the specified period
• Maximum: the highest value observed during the specified period

unit

Each statistic has a measurement unit. Examples of units include Bytes, Second, Count, Percent, etc.

Aggregation period

Each statistic calculates the data points of the metric collected during the selected aggregation interval. The aggregation interval can be chosen from 1 minute, 5 minutes, 15 minutes, 30 minutes, 1 hour, 3 hours, 6 hours, 12 hours, or 1 day, with the default being 5 minutes. The aggregation interval is closely related to the collection frequency of metric data points, and to obtain correct aggregation results, the aggregation interval must be equal to or longer than the collection frequency.

For example, if you select average, choose 5 minutes as the aggregation period, and pick a metric with a 1‑minute collection interval, data points are collected every minute and the average is calculated over the data points collected during the 5‑minute period. Conversely, if the aggregation period is shorter than the collection interval, it means a valid aggregation result cannot be obtained.

Downsampling is applied for long-term storage of metric data. For example, if data is collected at a 1‑minute interval, after 15 days the data can only be queried in 5‑minute increments. If you set the aggregation period for such metrics from 5 minutes to 30 minutes, up to 5 minutes may be required to retrieve the downsampled data correctly. After 63 days, the data is re‑aggregated and provided in 1‑hour intervals. At that point, selecting an aggregation period from 1 hour to 1 day may take up to 1 hour to retrieve the data correctly. This occurs because aggregating the downsampled metric data takes time, which can cause aggregation delays.

Alert

When creating an alert policy, you can evaluate a single metric over the specified evaluation period, and if it meets the condition set based on the threshold, you can notify the user with an alert.

The alarm status is classified as Alert (alert), Normal (normal), Insufficient data (no data).

• Alert(Alert): when the metric meets the configured condition
• Normal (Normal): when the indicator does not meet the set conditions.
• Insufficient data(no data): when the metric data does not exist, is missing, or has not yet arrived

When the alarm status is Alert, after evaluating the alarm, if it deviates from the condition, the alarm status changes back to Normal.

For detailed information about alerts, refer to the 경보 entry.

Basic monitoring and detailed monitoring

ServiceWatch provides two types of monitoring: basic monitoring and detailed monitoring.

The Samsung Cloud Platform services integrated with ServiceWatch provide basic monitoring by publishing a default set of metrics to ServiceWatch for free. By default, if you use any of these services, basic monitoring is automatically enabled and can be viewed in ServiceWatch.

Detailed monitoring is available only for certain services and incurs charges. To use detailed monitoring, you must enable it in the service details.

Detailed monitoring options vary depending on the service provided.

• The default monitoring for Virtual Server has a collection interval of 5 minutes. When detailed monitoring is enabled, the metrics provided by default monitoring are collected at a 5 minutes → 1 minute interval.
• Basic monitoring of Object Storage is provided for basic metrics, and enabling replication metrics provides additional replication metrics.

The following includes services and guides that provide detailed monitoring.