Hi,
Just want to share something I am working on in 3.2 if these matter.
* Shuffled hash join improvement (SPARK-32461)
* This is one of release notes JIRAs in 3.1, and major thing left is
sort-based fallback and code-gen for FULL OUTER join.
* Join and aggregation code-gen (SPARK-34287 and more to create)
* Add code-gen for all join types of sort merge join, object hash
aggregation and sort aggregation.
* Write Hive/Presto-compatible bucketed table (SPARK-19256)
* This is a long-standing issue and we made progress on plan during 3.1
development. We ideally want to finish the feature in 3.2.
For most of features here, we already developed internally and rolled out to
production.
Thanks,
Cheng Su
From: Dongjoon Hyun <[email protected]>
Date: Friday, February 26, 2021 at 4:06 PM
To: Hyukjin Kwon <[email protected]>
Cc: huaxin gao <[email protected]>, Xiao Li <[email protected]>, dev
<[email protected]>
Subject: Re: Apache Spark 3.2 Expectation
Sure, thank you, Hyukjin.
Bests,
Dongjoon.
On Fri, Feb 26, 2021 at 4:01 PM Hyukjin Kwon
<[email protected]<mailto:[email protected]>> wrote:
I have an idea which I'll send an email to discuss next or a week after the
next week. I did not have enough bandwidth to drive both together at the same
time. I would appreciate if we have some more time for 3.2.
In addition, It would also be great if we follow the schedule and catch
potential blockers quickly during QA instead of when we cut RCs. That will
considerably speed up the process and make it on time.
Thanks.
On Sat, 27 Feb 2021, 06:00 Dongjoon Hyun,
<[email protected]<mailto:[email protected]>> wrote:
Thank you for sharing your plan, Huaxin!
Bests,
Dongjoon.
On Fri, Feb 26, 2021 at 12:20 PM huaxin gao
<[email protected]<mailto:[email protected]>> wrote:
Thanks Dongjoon and Xiao for the discussion. I would like to add Data Source V2
Aggregate push down to the list. I am currently working on JDBC Data Source V2
Aggregate push down, but the common code can be used for the file based V2 Data
Source as well. For example, MAX and MIN can be pushed down to Parquet and Orc,
since they can use statistics information to perform these operations
efficiently. Quite a few users are interested in this Aggregate push down
feature and the preliminary performance test for JDBC Aggregate push down is
positive. So I think it is a valuable feature to add for Spark 3.2.
Thanks,
Huaxin
On Fri, Feb 26, 2021 at 11:13 AM Xiao Li
<[email protected]<mailto:[email protected]>> wrote:
Thank you, Dongjoon, for initiating this discussion. Let us keep it open. It
might take 1-2 weeks to collect from the community all the features we plan to
build and ship in 3.2 since we just finished the 3.1 voting.
3. +100 for Apache Spark 3.2.0 in July 2021. Maybe, we need `branch-cut` in
April because we took 3 month for Spark 3.1 release.
TBH, cutting the branch this April does not look good to me. That means, we
only have one month left for feature development of Spark 3.2. Do we have
enough features in the current master branch? If not, are we able to finish
major features we collected here? Do they have a timeline or project plan?
Xiao
Dongjoon Hyun <[email protected]<mailto:[email protected]>>
于2021年2月26日周五 上午10:07写道：
Thank you, Mridul and Sean.
1. Yes, `2017` was a typo. Java 17 is scheduled September 2021. And, of course,
it's a nice-to-have status. :)
2. `Push based shuffle and disaggregated shuffle`. Definitely. Thanks for
sharing,
3. +100 for Apache Spark 3.2.0 in July 2021. Maybe, we need `branch-cut` in
April because we took 3 month for Spark 3.1 release.
Let's update our release roadmap of the Apache Spark website.
> I'd roughly expect 3.2 in, say, July of this year, given the usual cadence.
> No reason it couldn't be a little sooner or later. There is already some good
> stuff in 3.2 and will be a good minor release in 5-6 months.
Bests,
Dongjoon.
On Thu, Feb 25, 2021 at 9:33 AM Sean Owen
<[email protected]<mailto:[email protected]>> wrote:
I'd roughly expect 3.2 in, say, July of this year, given the usual cadence. No
reason it couldn't be a little sooner or later. There is already some good
stuff in 3.2 and will be a good minor release in 5-6 months.
On Thu, Feb 25, 2021 at 10:57 AM Dongjoon Hyun
<[email protected]<mailto:[email protected]>> wrote:
Hi, All.
Since we have been preparing Apache Spark 3.2.0 in master branch since December
2020, March seems to be a good time to share our thoughts and aspirations on
Apache Spark 3.2.
According to the progress on Apache Spark 3.1 release, Apache Spark 3.2 seems
to be the last minor release of this year. Given the timeframe, we might
consider the following. (This is a small set. Please add your thoughts to this
limited list.)
# Languages
- Scala 2.13 Support: This was expected on 3.1 via SPARK-25075 but slipped out.
Currently, we are trying to use Scala 2.13.5 via SPARK-34505 and investigating
the publishing issue. Thank you for your contributions and feedback on this.
- Java 17 LTS Support: Java 17 LTS will arrive in September 2017. Like Java 11,
we need lots of support from our dependencies. Let's see.
- Python 3.6 Deprecation(?): Python 3.6 community support ends at 2021-12-23.
So, the deprecation is not required yet, but we had better prepare it because
we don't have an ETA of Apache Spark 3.3 in 2022.
- SparkR CRAN publishing: As we know, it's discontinued so far. Resuming it
depends on the success of Apache SparkR 3.1.1 CRAN publishing. If it succeeds
to revive it, we can keep publishing. Otherwise, I believe we had better drop
it from the releasing work item list officially.
# Dependencies
- Apache Hadoop 3.3.2: Hadoop 3.2.0 becomes the default Hadoop profile in
Apache Spark 3.1. Currently, Spark master branch lives on Hadoop 3.2.2's shaded
clients via SPARK-33212. So far, there is one on-going report at YARN
environment. We hope it will be fixed soon at Spark 3.2 timeframe and we can
move toward Hadoop 3.3.2.
- Apache Hive 2.3.9: Spark 3.0 starts to use Hive 2.3.7 by default instead of
old Hive 1.2 fork. Spark 3.1 removed hive-1.2 profile completely via
SPARK-32981 and replaced the generated hive-service-rpc code with the official
dependency via SPARK-32981. We are steadily improving this area and will
consume Hive 2.3.9 if available.
- K8s Client 4.13.2: During K8s GA activity, Spark 3.1 upgrades K8s client
dependency to 4.12.0. Spark 3.2 upgrades it to 4.13.2 in order to support K8s
model 1.19.
- Kafka Client 2.8: To bring the client fixes, Spark 3.1 is using Kafka Client
2.6. For Spark 3.2, SPARK-33913 upgraded to Kafka 2.7 with Scala 2.12.13, but
it was reverted later due to Scala 2.12.13 issue. Since KAFKA-12357 fixed the
Scala requirement two days ago, Spark 3.2 will go with Kafka Client 2.8
hopefully.
# Some Features
- Data Source v2: Spark 3.2 will deliver much richer DSv2 with Apache Iceberg
integration. Especially, we hope the on-going function catalog SPIP and
up-coming storage partitioned join SPIP can be delivered as a part of Spark 3.2
and become an additional foundation.
- Columnar Encryption: As of today, Apache Spark master branch supports
columnar encryption via Apache ORC 1.6 and it's documented via SPARK-34036.
Also, upcoming Apache Parquet 1.12 has a similar capability. Hopefully, Apache
Spark 3.2 is going to be the first release to have this feature officially. Any
feedback is welcome.
- Improved ZStandard Support: Spark 3.2 will bring more benefits for ZStandard
users: 1) SPARK-34340 added native ZSTD JNI buffer pool support for all IO
operations, 2) SPARK-33978 makes ORC datasource support ZSTD compression, 3)
SPARK-34503 sets ZSTD as the default codec for event log compression, 4)
SPARK-34479 aims to support ZSTD at Avro data source. Also, the upcoming
Parquet 1.12 supports ZSTD (and supports JNI buffer pool), too. I'm expecting
more benefits.
- Structure Streaming with RocksDB backend: According to the latest update, it
looks active enough for merging to master branch in Spark 3.2.
Please share your thoughts and let's build better Apache Spark 3.2 together.
Bests,
Dongjoon.
