Performance Evaluation of Cloudera Impala GA

Copyright © CELLANT Corp. All Rights Reserved. h t t p : / / w w w . c e l l a n t . j p / 1 1

Performance Evaluation ofCloudera impala 1.0

May 1, 2013CELLANT Corp. R&D Strategy Division

Yukinori SUDA@sudabon

Copyright © CELLANT Corp. All Rights Reserved. h t t p : / / w w w . c e l l a n t . j p /

v  Support for a subset of ANSI-‐‑‒92 SQLv  CREATE, ALTER, SELECT, INSERT, JOIN, and subqueries

v  Support for partitioned joins, fully distributed aggregations, and fully distributed top-‐‑‒n queries

v  Support for a variety of data formats:v  Hadoop native (Apache Avro, SequenceFile, RCFile with Snappy, GZIP, BZIP, or uncompressed)

v  text (uncompressed or LZO-‐‑‒compressed)v  Parquet (Snappy or uncompressed)

v  Support for all CDH4 64-‐‑‒bit packages:v  RHEL 6.2/5.7, Ubuntu, Debian, SLES

v  Connectivity via JDBC, ODBC, Hue GUI, or command-‐‑‒line shellv  Kerberos authentication and MR/Impala resource isolationv  etc

Cloudera Impala GA was released !!

2


Our System Environment

3

v  Install using Cloudera Manager Free Edition 4.5.2

Master Slave

11 Servers

All servers are connected with 1Gbps Ethernet through an L2 switch

ActiveNameNode

DataNodeTaskTrackerImpalad

Stand-‐‑‒byNameNode

JobTrackerstatestored

3 Servers












v CPUl Intel Core 2 Duo 2.13 GHz with Hyper Threading

v Memoryl 4GB

v Diskl 7,200 rpm SATA mechanical Hard Disk Drive * 1

v OSl Cent OS 6.2

Our “wimpy” Server Specification

4


v  Use CDH4.2.1 + Impala version 1.0v  Use hivebench in open-‐‑‒sourced benchmark tool “HiBench”

l  https://github.com/hibenchv  Modified datasets to 1/10 scale

l  Default configuration generates table with 1 billion rowsv  Modified query sentence

l  Deleted “INSERT INTO TABLE …” to evaluate read-‐‑‒only performancev  Combines a few storage format with a few compression method

l  TextFile, SequenceFile, RCFile, ParquestFilel  No compression, Gzip, Snappy

v  Comparison with job query latencyv  Average job latency over 5 measurements

Benchmark

5


•  Uservisits table–  100 million rows–  16,895 MB as TextFile–  Table Definitions

•  sourceIP string•  destURL string•  visitDate string•  adRevenue double•  userAgent string•  countryCode string•  languageCode string•  searchWord string•  duration int

•  Rankings table–  12 million rows–  744 MB as TextFile–  Table Definitions

•  pageURL string•  pageRank int•  avgDuration int

Modified Datasets

6


SELECT 　sourceIP, 　sum(adRevenue) as totalRevenue, 　avg(pageRank) FROM 　rankings_̲t RJOIN ( 　SELECT 　　sourceIP, 　　destURL, 　　adRevenue 　FROM 　　uservisits_̲t UV 　WHERE 　　(datediff(UV.visitDate, '1999-‐‑‒01-‐‑‒01')>=0 　　AND 　　datediff(UV.visitDate, '2000-‐‑‒01-‐‑‒01')<=0) 　) NUV

ON 　(R.pageURL = NUV.destURL)group by sourceIPorder by totalRevenue DESClimit 1;

Modified Query

7


Benchmark Result (Hive)cited from “Performance evaluation of Cloudera impala 0.6 beta...”

8

0 50 100 150 200 250

No Comp.

Gzip

Snappy

Gzip

Snappy

TextFile

SequenceFile

RCFile

235.843

227.883

213.616

234.289

197.894

Avg. Job Latency [sec]


Benchmark Result (Impala)

9

0 50 100 150 200 250

No Comp.

Gzip

Snappy

Gzip

Snappy

Snappy

Text

File

Sequence

File

RCFile

Parquet

File

36.61

29.736

24.024

26.083

19.586

16.2

Avg. Job Latency [sec]


v Exchange the order of JOINed Tables like belowSELECT

sourceIP, sum(adRevenue) as totalRevenue, avg(pageRank)FROM

(SELECT sourceIP, destURL, adRevenue FROM uservisits_̲ps UV WHERE (datediff(UV.visitDate, '1999-‐‑‒01-‐‑‒01')>=0 AND datediff(UV.visitDate, '2000-‐‑‒01-‐‑‒01')<=0)) NUV

JOINrankings_̲ps R

ON(R.pageURL = NUV.destURL)

group by sourceIPorder by totalRevenue DESClimit 1;

v Resultl Parquet compressed as Snappy: 34.374 sec

Additional Experiments

10


v Parquet + Snappy is the fastestv Specifically,

l ParquetFile compressed as Snappy: 16.2 secv Need to take care the order of JOINed tables

v Hope for future extensionl Support UDFl Window Functionl etc

Conclusion

11

Copyright © CELLANT Corp. All Rights Reserved. h t t p : / / w w w . c e l l a n t . j p / 12

Letʼ’s try it out on your envrionment!!Thanks!

Performance Evaluation of Cloudera Impala GA

Technology