andygrove opened a new pull request, #2864: URL: https://github.com/apache/datafusion-ballista/pull/2864
# Which issue does this PR close? N/A — small documentation improvement, no issue filed. # Rationale for this change The README introduction describes Ballista as an alternative to Spark that keeps the same execution model of stages, shuffles and AQE, then quotes a 3.4x speedup for "experimental pipelined execution" without saying what that is. Pipelined execution is only explained ~110 lines later, split across two overlapping sections. Readers get an incomplete picture of what Ballista is before they reach that point. # What changes are included in this PR? - The introduction now presents Ballista as an alternative to both Spark and Trino and briefly describes the staged (default) and pipelined (experimental) modes before the benchmark numbers, which now say "staged mode" and "pipelined mode". - Merges the "Execution Modes" and "One Engine for ETL and Interactive Queries" sections into one section that explains the fault-tolerance vs. latency trade-off, keeping the shared-capacity argument and the current limitations. - The Spark persona under "Who is Ballista for" now mentions pipelined execution, and a new persona covers teams running separate batch and interactive engines. - The architecture overview now describes how executors exchange data in each mode. - The Features list links to the execution modes section, and the Performance section says the queries are TPC-H-derived. - Fixes the `ballista-executor` Cargo features table: `mimalloc` is not a declared feature (it is enabled by `build-binary`), and the new `jemalloc` feature was missing. - Removes the terminal TUI screenshot from the collapsed Cargo features section, where it was out of place. # Are there any user-facing changes? Documentation only. 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
