| Parameters | Default | Note |
|---|
| spark.blaze.enable | true | Enable/disable blaze engine. |
| spark.blaze.batchSize | 10000 | Suggested batch size for arrow batches. |
| spark.blaze.memoryFraction | 0.6 | Suggested fraction of off-heap memory used in native execution. Actual off-heap memory usage is expected to be spark.executor.memoryOverhead * fraction. |
| spark.blaze.tokio.num.worker.threads | 1 | Number of worker threads used in tokio runtime, 0 to use default available parallelism value. For CPUs those support hyperthreading, it is recommended to set this value to the number of available physical cores. |
| spark.blaze.enableInputBatchStatistics | true | Enable extra metrics of input batch statistics. |
| spark.blaze.partialAggSkipping.enable | true | Enable partial aggregate skipping. (see https://github.com/blaze-init/blaze/issues/327) |
| spark.blaze.partialAggSkipping.ratio | 0.8 | Partial aggregate skipping ratio. |
| spark.blaze.partialAggSkipping.minRows | 20000 | Minimum number of rows to trigger partial aggregate skipping. |
| spark.blaze.parquet.enable.pageFiltering | false | Parquet enable page filtering. |
| spark.blaze.parquet.enable.bloomFilter | false | Parquet enable bloom filter. |
| spark.blaze.forceShuffledHashJoin | false | Replace all sort-merge join to shuffled-hash join, only used for special benchmarking. |
| Parameters | Default |
|---|
| spark.blaze.enable.scan | true |
| spark.blaze.enable.project | true |
| spark.blaze.enable.filter | true |
| spark.blaze.enable.sort | true |
| spark.blaze.enable.union | true |
| spark.blaze.enable.smj | true |
| spark.blaze.enable.shj | true |
| spark.blaze.enable.bhj | true |
| spark.blaze.enable.bnlj | true |
| spark.blaze.enable.local.limit | true |
| spark.blaze.enable.global.limit | true |
| spark.blaze.enable.take.ordered.and.project | true |
| spark.blaze.enable.aggr | true |
| spark.blaze.enable.expand | true |
| spark.blaze.enable.window | true |
| spark.blaze.enable.generate | true |
| spark.blaze.enable.local.table.scan | true |
| spark.blaze.enable.data.writing | false |
