This topic describes how to select specifications when you configure a Realtime Compute for Apache Flink cluster in exclusive mode. It also provides precautions during the configuration.
- Master nodes manage cluster resources and interactions among slave nodes, but are not used for computing.
- Slave nodes are used for computing.
Note A slave node cannot use all of its resources for computing because the operating system and interactions with other nodes also consume resources.
- The slave node specifications determine the scaling configuration. For example, if your slave node specifications are 8 CPU cores and 32 GB memory, you can add or remove only a specific number of nodes of this configuration for each scaling operation. In this case, the number of available compute units (CUs) increases or decreases by the value that is calculated by using the following formula: Number of nodes added or removed × 6 CUs.
- Three master nodes in a Realtime Compute for Apache Flink cluster support failover in case of a failure. This ensures cluster stability. If you configure a Realtime Compute for Apache Flink cluster with three master nodes, Alibaba Cloud provides you with a Service Level Agreement (SLA) guarantee.
- You cannot change the number of master nodes in a Realtime Compute for Apache Flink cluster.
- For simple operations such as single-stream filtering and string conversion, one CU can process 10,000 data records per second.
- For complex operations such as operations that use a JOIN clause, GROUP BY clause, or window function, one CU can process 1,000 to 5,000 data records per second.
- You must configure at least two slave nodes for a Realtime Compute for Apache Flink
cluster in exclusive mode. Therefore, the minimum computing capability of this cluster
is 6 CUs, which is calculated by using the following formula: 3 CUs × 2.
Slave node specifications Available CUs 4 CPU cores, 16 GB memory 3 8 CPU cores, 32 GB memory 6 16 CPU cores, 64 GB memory 13 24 CPU cores, 96 GB memory 21 32 CPU cores,128 GB memory 28 56 CPU cores, 224 GB memory 52 64 CPU cores, 256 GB memory 60Note The empirical data is for reference only.
- The master node specifications are restricted by the maximum number of CUs allowed
in a cluster. The following table lists empirical CU quantities for different master
Master node specifications Maximum number of CUs allowed in a cluster 4 CPU cores, 16 GB memory 80 8 CPU cores, 32 GB memory 160 16 CPU cores, 64 GB memory 800 24 CPU cores, 96 GB memory More than 800Note The empirical data is for reference only.