Currently, the maximum instance type provided by Time Series Database (TSDB) is Ultimate Edition II. If you need instances with larger instance types, submit a ticket.
To protect instances, TSDB restricts the usage of resources depending on the instance types. When you use a TSDB instance, do not exceed the corresponding limits to avoid system errors. If an error is returned to the application layer, indicating that the limit is exceeded, upgrade the instance type in the TSDB console. The following table describes the details about the limits on use.
|Instance type||Maximum number of tags in a timeline||Maximum number of timelines covered in a subquery (10,000)||Maximum number of data points covered in a subquery (10,000)||Maximum number of data points returned by a subquery (10,000)||Maximum QPS|
|Basic Edition I||16||10||700||15||200|
|Basic Edition II||20||15||700||20||1,000|
|Basic Edition III||20||20||700||20||1,200|
|Standard Edition I||20||80||2,000||80||4,000|
|Standard Edition II||24||120||4,000||160||8,000|
|Ultimate Edition I||24||160||6,000||180||24,000|
|Ultimate Edition II||24||200||8,000||200||50,000|
- Subquery: each time series query can contain multiple subqueries to implement batch queries.
- Number of timelines covered in a subquery: the total number of hit timelines after each subquery is issued to the engine layer.
- Number of data points covered in a subquery: the total number of data points to be scanned and calculated after each subquery is issued to the storage layer.
- Number of data points returned by a subquery: the number of data points returned to the application layer by a subquery after time series aggregation.
- Maximum QPS: the query benchmark for queries per second (QPS) measurement is that a timeline, on which a data record (containing only one tag) is reported every 10 seconds, is scanned within a time range of 5 minutes.
If a query contains multiple subqueries and one of the subqueries violates the subquery limits for the instance, the entire query fails and an error is returned. In this case, you can modify the query conditions, or upgrade the instance type.