All Products
Search
Document Center

Realtime Compute for Apache Flink:Limits

Last Updated:Sep 10, 2024

This topic describes the limits of Realtime Compute for Apache Flink, including performance limits and limits on the numbers of compute units (CUs), namespaces, and JAR packages. You can assess how the limits may affect your business and make necessary adjustments.

Usage limits

  • Browser: We recommend that you use browsers that use the Chrome kernel to access the development console of Realtime Compute for Apache Flink.

  • Regions and zones: Only specific regions and zones are supported. An Alibaba Cloud region has one or more zones. For more information, see Regions and zones.

  • CUs: You can purchase up to 1,000 CUs for a subscription workspace at a time. If you want to purchase more than 1,000 CUs, submit a ticket.

  • Network: By default, Realtime Compute for Apache Flink cannot access the Internet. For information about how to access the Internet, see the "How does Realtime Compute for Apache Flink access the Internet?" section of the Reference topic.

  • Performance: Under stress testing conditions for simple stream processing tasks, such as single-stream filtering and string conversion, one CU can process 40,000 to 55,000 data records per second. Under stress testing conditions for complex stream processing tasks, such as JOIN and GROUP BY operations and window aggregation, one CU can process 5,000 to 10,000 data records per second. For more information, see Performance white paper (Nexmark performance testing).

  • Others:

    • A workspace can contain up to 5,000 namespaces.

    • A namespace can contain up to 100 folders and the maximum folder depth is five.

    • A folder can contain up to 10,000 files and subfolders.

    • The compute instances on which your code runs are your assets. Therefore, you shall assume corresponding legal liabilities that arise from code running.

References