Google Git
Sign in
apache / impala / 0dff5ef3621a0be907c6968096b643a5b2d9f30c
commit0dff5ef3621a0be907c6968096b643a5b2d9f30c[log] [tgz]
authorTim Armstrong <tarmstrong@cloudera.com>Sun Oct 06 22:55:38 2019 -0700
committerImpala Public Jenkins <impala-public-jenkins@cloudera.com>Thu Oct 10 04:10:51 2019 +0000
treefc7bd0ce30c306ade830ffe19e62aef9613592ef
parent950973a1680432735319f0992620b6425ba1ec57 [diff]
IMPALA-9015: improve mt_dop scan scheduling

Implement longest-processing time algorithm for assigning scan ranges
to instances within a host. This is a standard algorithm that works
well in practice and solves some specific bugs in the current
algorithm.

The previous approach tended to assign multiple ranges to the
first instance and induce skew. E.g. if the ranges were
[3, 4, 5, 6] and it had 4 instances, it would assign
[3, 4], [5], [6], []. This also had the unfortunate consequence
that not all instances actually got allocated scan ranges,
making scheduling hard to reason about.

Testing:
Added a unit test for the core algorithm that checks directly
that it fixes that above problems.

Perf:
The algorithm is O(n log n) instead of O(n), where n is the
number of scan ranges allocated to a backend. This seems
worthwhile to get more even work distribution because
the payoff can be significant in may cases.

I ran a single node perf run of TPC-H scale 30 with
mt_dop=4 and saw significant perf improvements:

Report Generated on 2019-10-07
Run Description: "ccd741856d0e18be9c89087da71d9fc59f1c75ad vs 37d3df99e588b0e9080660c472c3bb06a08361ac"

Cluster Name: UNKNOWN
Lab Run Info: UNKNOWN
Impala Version:          impalad version 3.4.0-SNAPSHOT RELEASE ()
Baseline Impala Version: impalad version 3.4.0-SNAPSHOT RELEASE ()

+----------+-----------------------+---------+------------+------------+----------------+
| Workload | File Format           | Avg (s) | Delta(Avg) | GeoMean(s) | Delta(GeoMean) |
+----------+-----------------------+---------+------------+------------+----------------+
| TPCH(30) | parquet / none / none | 8.51    | -3.78%     | 5.70       | -4.07%         |
+----------+-----------------------+---------+------------+------------+----------------+

+----------+----------+-----------------------+--------+-------------+------------+-----------+----------------+-------+----------------+---------+--------+
| Workload | Query    | File Format           | Avg(s) | Base Avg(s) | Delta(Avg) | StdDev(%) | Base StdDev(%) | Iters | Median Diff(%) | MW Zval | Tval   |
+----------+----------+-----------------------+--------+-------------+------------+-----------+----------------+-------+----------------+---------+--------+
| TPCH(30) | TPCH-Q11 | parquet / none / none | 3.08   | 3.04        |   +1.11%   |   0.72%   |   1.09%        | 5     |   +1.52%       | 0.58    | 1.88   |
| TPCH(30) | TPCH-Q2  | parquet / none / none | 1.83   | 1.83        |   -0.01%   |   3.68%   |   1.25%        | 5     |   +2.00%       | 0.00    | -0.00  |
| TPCH(30) | TPCH-Q9  | parquet / none / none | 36.29  | 36.17       |   +0.32%   |   2.22%   |   2.94%        | 5     |   +0.56%       | 0.00    | 0.20   |
| TPCH(30) | TPCH-Q8  | parquet / none / none | 5.80   | 5.93        |   -2.06%   |   0.61%   |   0.94%        | 5     |   -1.74%       | -2.02   | -4.14  |
| TPCH(30) | TPCH-Q3  | parquet / none / none | 8.91   | 9.15        |   -2.62%   |   3.54%   |   3.96%        | 5     |   -2.17%       | -0.87   | -1.12  |
| TPCH(30) | TPCH-Q15 | parquet / none / none | 3.38   | 3.46        |   -2.30%   |   0.82%   |   0.68%        | 5     |   -2.54%       | -2.31   | -4.89  |
| TPCH(30) | TPCH-Q6  | parquet / none / none | 1.62   | 1.67        |   -2.73%   |   2.52%   |   0.12%        | 5     |   -2.88%       | -1.15   | -2.49  |
| TPCH(30) | TPCH-Q20 | parquet / none / none | 3.67   | 3.78        |   -3.06%   |   1.12%   |   1.69%        | 5     |   -2.69%       | -2.31   | -3.41  |
| TPCH(30) | TPCH-Q4  | parquet / none / none | 3.08   | 3.18        |   -3.05%   |   1.22%   |   0.32%        | 5     |   -2.86%       | -2.31   | -5.57  |
| TPCH(30) | TPCH-Q14 | parquet / none / none | 5.52   | 5.72        |   -3.56%   |   1.35%   |   1.02%        | 5     |   -3.60%       | -2.31   | -4.81  |
| TPCH(30) | TPCH-Q16 | parquet / none / none | 2.32   | 2.41        |   -3.86%   |   1.04%   |   1.64%        | 5     |   -3.99%       | -2.02   | -4.50  |
| TPCH(30) | TPCH-Q7  | parquet / none / none | 24.22  | 25.27       |   -4.15%   |   0.40%   |   2.01%        | 5     |   -3.74%       | -2.31   | -4.55  |
| TPCH(30) | TPCH-Q18 | parquet / none / none | 8.64   | 8.99        |   -3.97%   |   1.02%   |   0.47%        | 5     |   -4.27%       | -2.31   | -8.15  |
| TPCH(30) | TPCH-Q12 | parquet / none / none | 3.38   | 3.53        |   -4.28%   |   1.67%   |   1.39%        | 5     |   -4.26%       | -2.31   | -4.51  |
| TPCH(30) | TPCH-Q5  | parquet / none / none | 11.77  | 12.28       |   -4.18%   |   0.72%   |   0.94%        | 5     |   -4.62%       | -2.31   | -8.06  |
| TPCH(30) | TPCH-Q17 | parquet / none / none | 7.65   | 8.05        |   -4.98%   |   1.03%   |   2.09%        | 5     |   -4.68%       | -2.31   | -4.82  |
| TPCH(30) | TPCH-Q21 | parquet / none / none | 27.97  | 29.52       | I -5.27%   |   0.78%   |   0.39%        | 5     | I -5.16%       | -2.31   | -14.17 |
| TPCH(30) | TPCH-Q1  | parquet / none / none | 4.92   | 5.24        | I -6.05%   |   3.59%   |   2.32%        | 5     | I -6.79%       | -1.73   | -3.30  |
| TPCH(30) | TPCH-Q19 | parquet / none / none | 4.56   | 4.92        | I -7.33%   |   1.55%   |   0.90%        | 5     | I -8.50%       | -2.31   | -9.64  |
| TPCH(30) | TPCH-Q22 | parquet / none / none | 2.09   | 2.28        | I -8.40%   |   1.95%   |   1.85%        | 5     | I -9.21%       | -2.31   | -7.29  |
| TPCH(30) | TPCH-Q13 | parquet / none / none | 10.83  | 11.84       | I -8.55%   |   0.53%   |   0.57%        | 5     | I -9.34%       | -2.31   | -25.56 |
| TPCH(30) | TPCH-Q10 | parquet / none / none | 5.63   | 6.22        | I -9.53%   |   0.75%   |   0.73%        | 5     | I -10.46%      | -2.31   | -21.37 |
+----------+----------+-----------------------+--------+-------------+------------+-----------+----------------+-------+----------------+---------+--------+

Change-Id: I45ed2dab835efeb64bb74891cb43065894892682
Reviewed-on: http://gerrit.cloudera.org:8080/14381
Reviewed-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
Tested-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
3 files changed
tree: fc7bd0ce30c306ade830ffe19e62aef9613592ef
  1. be/
  2. bin/
  3. cmake_modules/
  4. common/
  5. docker/
  6. docs/
  7. ext-data-source/
  8. fe/
  9. impala-parent/
  10. infra/
  11. lib/
  12. query-event-hook-api/
  13. security/
  14. shaded-deps/
  15. shell/
  16. ssh_keys/
  17. testdata/
  18. tests/
  19. www/
  20. .clang-format
  21. .clang-tidy
  22. .gitattributes
  23. .gitignore
  24. buildall.sh
  25. CMakeLists.txt
  26. EXPORT_CONTROL.md
  27. LICENSE.txt
  28. LOGS.md
  29. NOTICE.txt
  30. README.md
  31. setup.cfg
README.md

Welcome to Impala

Lightning-fast, distributed SQL queries for petabytes of data stored in Apache Hadoop clusters.

Impala is a modern, massively-distributed, massively-parallel, C++ query engine that lets you analyze, transform and combine data from a variety of data sources:

  • Best of breed performance and scalability.
  • Support for data stored in HDFS, Apache HBase and Amazon S3.
  • Wide analytic SQL support, including window functions and subqueries.
  • On-the-fly code generation using LLVM to generate CPU-efficient code tailored specifically to each individual query.
  • Support for the most commonly-used Hadoop file formats, including the Apache Parquet project.
  • Apache-licensed, 100% open source.

More about Impala

To learn more about Impala as a business user, or to try Impala live or in a VM, please visit the Impala homepage.

If you are interested in contributing to Impala as a developer, or learning more about Impala's internals and architecture, visit the Impala wiki.

Supported Platforms

Impala only supports Linux at the moment.

Export Control Notice

This distribution uses cryptographic software and may be subject to export controls. Please refer to EXPORT_CONTROL.md for more information.

Build Instructions

See bin/bootstrap_build.sh.

Detailed Build Notes

Impala can be built with pre-built components or components downloaded from S3. The components needed to build Impala are Apache Hadoop, Hive, HBase, and Sentry. If you need to manually override the locations or versions of these components, you can do so through the environment variables and scripts listed below.

Scripts and directories
LocationPurpose
bin/impala-config.shThis script must be sourced to setup all environment variables properly to allow other scripts to work
bin/impala-config-local.shA script can be created in this location to set local overrides for any environment variables
bin/impala-config-branch.shA version of the above that can be checked into a branch for convenience.
bin/bootstrap_build.shA helper script to bootstrap some of the build requirements.
bin/bootstrap_development.shA helper script to bootstrap a developer environment. Please read it before using.
be/build/Impala build output goes here.
be/generated-sources/Thrift and other generated source will be found here.
Build Related Variables
Environment variableDefault valueDescription
IMPALA_HOMETop level Impala directory
IMPALA_TOOLCHAIN“${IMPALA_HOME}/toolchain”Native toolchain directory (for compilers, libraries, etc.)
SKIP_TOOLCHAIN_BOOTSTRAP“false”Skips downloading the toolchain any python dependencies if “true”
CDH_BUILD_NUMBERIdentifier to indicate the CDH build number
CDH_COMPONENTS_HOME“${IMPALA_HOME}/toolchain/cdh_components-${CDH_BUILD_NUMBER}”Location of the CDH components within the toolchain.
CDH_MAJOR_VERSION“5”Identifier used to uniqueify paths for potentially incompatible component builds.
IMPALA_CONFIG_SOURCED“1”Set by ${IMPALA_HOME}/bin/impala-config.sh (internal use)
JAVA_HOME“/usr/lib/jvm/${JAVA_VERSION}”Used to locate Java
JAVA_VERSION“java-7-oracle-amd64”Can override to set a local Java version.
JAVA“${JAVA_HOME}/bin/java”Java binary location.
CLASSPATHSee bin/set-classpath.sh for details.
PYTHONPATHWill be changed to include: “${IMPALA_HOME}/shell/gen-py” “${IMPALA_HOME}/testdata” “${THRIFT_HOME}/python/lib/python2.7/site-packages” “${HIVE_HOME}/lib/py” “${IMPALA_HOME}/shell/ext-py/prettytable-0.7.1/dist/prettytable-0.7.1” "${IMPALA_HOME}/shell/ext-py/sasl-0.1.1/dist/sasl-0.1.1-py2.7-linux-x "${IMPALA_HOME}/shell/ext-py/sqlparse-0.1.19/dist/sqlparse-0.1.19-py2
Source Directories for Impala
Environment variableDefault valueDescription
IMPALA_BE_DIR“${IMPALA_HOME}/be”Backend directory. Build output is also stored here.
IMPALA_FE_DIR“${IMPALA_HOME}/fe”Frontend directory
IMPALA_COMMON_DIR“${IMPALA_HOME}/common”Common code (thrift, function registry)
Various Compilation Settings
Environment variableDefault valueDescription
IMPALA_BUILD_THREADS“8” or set to number of processors by default.Used for make -j and distcc -j settings.
IMPALA_MAKE_FLAGS""Any extra settings to pass to make. Also used when copying udfs / udas into HDFS.
USE_SYSTEM_GCC“0”If set to any other value, directs cmake to not set GCC_ROOT, CMAKE_C_COMPILER, CMAKE_CXX_COMPILER, as well as setting TOOLCHAIN_LINK_FLAGS
IMPALA_CXX_COMPILER“default”Used by cmake (cmake_modules/toolchain and clang_toolchain.cmake) to select gcc / clang
USE_GOLD_LINKER“true”Directs backend cmake to use gold.
IS_OSX“false”(Experimental) currently only used to disable Kudu.
Dependencies
Environment variableDefault valueDescription
HADOOP_HOME“${CDH_COMPONENTS_HOME}/hadoop-${IMPALA_HADOOP_VERSION}/”Used to locate Hadoop
HADOOP_INCLUDE_DIR“${HADOOP_HOME}/include”For ‘hdfs.h’
HADOOP_LIB_DIR“${HADOOP_HOME}/lib”For ‘libhdfs.a’ or ‘libhdfs.so’
HIVE_HOME“${CDH_COMPONENTS_HOME}/{hive-${IMPALA_HIVE_VERSION}/”
HBASE_HOME“${CDH_COMPONENTS_HOME}/hbase-${IMPALA_HBASE_VERSION}/”
SENTRY_HOME“${CDH_COMPONENTS_HOME}/sentry-${IMPALA_SENTRY_VERSION}/”Used to setup test data
THRIFT_HOME“${IMPALA_TOOLCHAIN}/thrift-${IMPALA_THRIFT_VERSION}”