References → Custom SparkX Sizes
For on-prem installations, there are minimum or recommended sizes (for both memory and CPU) to use with custom Spark depending on your cluster size.
The following table illustrates the recommended sizes to use:
Cluster size/ Service | XSmall | Small | Medium | Large | XLarge | 2XLarge |
---|---|---|---|---|---|---|
Kyuubi Server Pod | Default for both CPU and Memory | Default for both CPU and Memory | Default for both CPU and Memory | Default for both CPU and Memory | Default for both CPU and Memory | Default for both CPU and Memory |
SparkX Driver Pod | ● Memory: 16 ● CPU Cores: 4 ● MaxResultSize: 200 MBs | ● Memory: 16 ● CPU Cores: 4 ● MaxResultSize: 200 MBs | ● Memory: 32 ● CPU Cores: 8 ● MaxResultSize: 400 MBs | ● Memory: 64 ● CPU Cores: 16 ● MaxResultSize: 800 MBs | ● Memory: 64 ● CPU Cores: 16 ● MaxResultSize: 800 MBs | ● Memory: 64 ● CPU Cores: 16 ● MaxResultSize: 800 MBs |
SparkX Executors | ● Memory: 8 ● CPU Cores: 2 ● Min Executors: 0 ● Max Executors: 1 | ● Memory: 8 ● CPU Cores: 2 ● Min Executors: 0 ● Max Executors: 4 | ● Memory: 32 ● CPU Cores: 8 ● Min Executors: 0 ● Max Executors: 6 | ● Memory: 32 ● CPU Cores: 8 ● Min Executors: 0 ● Max Executors: 8 | Memory: 32 ● CPU Cores: 8 ● Min Executors: 0 ● Max Executors: 8 | ● Memory: 32 ● CPU Cores: 8 ● Min Executors: 0 ● Max Executors: 8 |
Engine Partitioning/Data Transfer Parallelism | 8 | 16 | 32 | 64 | 64 | 64 |