Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improve the scalability of the join between the LHS and GroupBys by breaking up the join #621

Open
wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Improve the scalability of the join between the LHS and GroupBys by breaking up the join #621

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
eeb0ef6
Break up joins
mears-stripe Nov 22, 2023
9b787bc
Fix addJoinBreak
mears-stripe Nov 22, 2023
5096e4a
PR feedback
mears-stripe Nov 22, 2023
d1ec5d8
Empty-Commit
mears-stripe Nov 22, 2023
93ecd2c
Empty-Commit
mears-stripe Nov 22, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion spark/src/main/scala/ai/chronon/spark/Join.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,7 @@ class Join(joinConf: api.Join,
extends JoinBase(joinConf, endPartition, tableUtils, skipFirstHole, mutationScan, showDf) {

private val bootstrapTable = joinConf.metaData.bootstrapTable
private val joinsAtATime = 8
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we make this consume a spark conf param - via tableUtils?


private def padFields(df: DataFrame, structType: sql.types.StructType): DataFrame = {
structType.foldLeft(df) {
Expand Down Expand Up @@ -263,7 +264,13 @@ class Join(joinConf: api.Join,
// a bootstrap source can cover a partial date range. we combine the columns using coalesce-rule
rightResults
.foldLeft(bootstrapDf) {
case (partialDf, (rightPart, rightDf)) => joinWithLeft(partialDf, rightDf, rightPart)
case (partialDf, ((rightPart, rightDf), i)) =>
val next = joinWithLeft(partialDf, rightDf, rightPart)
if (((i + 1) % joinsAtATime) == 0) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if we have 24 parts - there will be 3 cache points - at 8, 16, 24

16 should evict the 8 cache. 24 shouldn't cache since it is the last one.

tableUtils.addJoinBreak(next)
} else {
next
}
}
// drop all processing metadata columns
.drop(Constants.MatchedHashes, Constants.TimePartitionColumn)
Expand Down
3 changes: 3 additions & 0 deletions spark/src/main/scala/ai/chronon/spark/TableUtils.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -324,6 +324,9 @@ case class TableUtils(sparkSession: SparkSession) {
df
}

def addJoinBreak(dataFrame: DataFrame): DataFrame =
dataFrame.cache()
Copy link
Contributor

@nikhilsimha nikhilsimha Nov 22, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TableUtils has a cache_level param and a wrap with cache method that does exception handling to release the resources claimed by the cache. I think we should use that here.


def insertUnPartitioned(df: DataFrame,
tableName: String,
tableProperties: Map[String, String] = null,
Expand Down