[GitHub] [arrow-datafusion] yjshen commented on a diff in pull request #2226: Introduce new optional scheduler, using Morsel-driven Parallelism + rayon (#2199)

[GitBox <gi...@apache.org>]

yjshen commented on code in PR #2226:
URL: https://github.com/apache/arrow-datafusion/pull/2226#discussion_r863796230


##########
datafusion/core/src/scheduler/mod.rs:
##########
@@ -0,0 +1,454 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+//! A [`Scheduler`] maintains a pool of dedicated worker threads on which
+//! query execution can be scheduled. This is based on the idea of [Morsel-Driven Parallelism]
+//! and is designed to decouple the execution parallelism from the parallelism expressed in
+//! the physical plan as partitions.
+//!
+//! # Implementation
+//!
+//! When provided with an [`ExecutionPlan`] the [`Scheduler`] first breaks it up into smaller
+//! chunks called pipelines. Each pipeline may consist of one or more nodes from the
+//! [`ExecutionPlan`] tree.
+//!
+//! The scheduler then maintains a list of pending [`Task`], that identify a partition within
+//! a particular pipeline that may be able to make progress on some "morsel" of data. These
+//! [`Task`] are then scheduled on the worker pool, with a preference for scheduling work
+//! on a given "morsel" on the same thread that produced it.
+//!
+//! # Rayon
+//!
+//! Under-the-hood these [`Task`] are scheduled by [rayon], which is a lightweight, work-stealing
+//! scheduler optimised for CPU-bound workloads. Pipelines may exploit this fact, and use [rayon]'s
+//! structured concurrency primitives to express additional parallelism that may be exploited
+//! if there are idle threads available at runtime
+//!
+//! # Shutdown
+//!
+//! Queries scheduled on a [`Scheduler`] will run to completion even if the
+//! [`Scheduler`] is dropped
+//!
+//! [Morsel-Driven Parallelism]: https://db.in.tum.de/~leis/papers/morsels.pdf
+//! [rayon]: https://docs.rs/rayon/latest/rayon/
+//!
+//! # Example
+//!
+//! ```rust
+//! # use futures::TryStreamExt;
+//! # use datafusion::prelude::{CsvReadOptions, SessionConfig, SessionContext};
+//! # use datafusion_scheduler::Scheduler;
+//!
+//! # #[tokio::main]
+//! # async fn main() {
+//! let scheduler = Scheduler::new(4);
+//! let config = SessionConfig::new().with_target_partitions(4);
+//! let context = SessionContext::with_config(config);
+//!
+//! context.register_csv("example", "../core/tests/example.csv", CsvReadOptions::new()).await.unwrap();
+//! let plan = context.sql("SELECT MIN(b) FROM example")
+//!     .await
+//!    .unwrap()
+//!    .create_physical_plan()
+//!    .await
+//!    .unwrap();

Review Comment:
   nit: the above four lines are of 1 space less indent.



-- 
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: github-unsubscribe@arrow.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org