Hadoop-DS: Which SQL-on-Hadoop Rules the Herd

© 2014 IBM Corporation
Information Management
Evaluating SQL-on-Hadoop
Performance and Compatibility
IBM Big SQL Hadoop-DS Benchmark
Last revised: Oct 26, 2014

© 2014 IBM Corporation2
Agenda
About Big SQL
The TPC-DS™ Benchmark
The Hadoop-DS Benchmark
Big SQL performance
 30TB Hadoop-DS result with IBM Big SQL
 10TB Hadoop-DS comparison with Cloudera Impala™ and
Hortonworks® Hive
Conclusions
Additional detail
TPC Benchmark, TPC-DS, and QphDS are trademarks of Transaction Processing Performance Council
Cloudera, the Cloudera logo, Cloudera Impala are trademarks of Cloudera.
Hortonworks is a trademark of Hortonworks inc.
Other company, product, or service names may be trademarks or service marks of others.

The case for SQL on Hadoop
SQL has become ubiquitous in today’s data center
Customers have large existing investments
 Skills, commercial & in-house applications
70% of Big Data initiatives involve transactional data1
 Transactional big data well suited to SQL
Standardization & compatibility are essential
 Customers modernizing warehouse environments cannot afford
separate SQL dialects and tools for different data sources
1. 70% of 465 survey respondents cite transactional data as a primary target for big data initiatives - Gartner research note “Survey Analysis - Big
Data Adoption in 2013 Shows Substance Behind the Hype“ Sept 12 2013 Analyst(s): Lisa Kart, Nick Heudecker, Frank Buytendijk

IBM InfoSphere BigInsights - Big SQL
Big SQL = Big Investment Protection
Rich ANSI SQL support
Outstanding performance
Native Hadoop data sources
Federation: multiple data
sources
Extensive analytic functions
Security built-in
Native Hadoop Data Sources
CSV SEQ Parquet RC
AVRO ORC JSON Custom
Optimized SQL MPP Run-time
Big SQL
SQL based
Application
IBM invented SQL and has over thirty
years of experience engineering
advanced SQL query engines

IBM InfoSphere BigInsights - Big SQL
Application Portability &
Integration
 Native Hadoop Data
 Comprehensive file formats
Performance
 Powerful query re-writer
 Cost-based optimizer
 Sophisticated memory mgmt.
Federation
 Single SQL statement
 Multiple data sources
 DB2, Oracle, Teradata & more
Enterprise Features
 Security & Auditing
 Self-tuning & management
 Comprehensive monitoring
Rich SQL Support
 ANSI Compliant
 IBM PL SQL Compatible
 Extensive Analytic Functions
Big SQL = Big Investment Protection

About the TPC-DS Benchmark – www.tpc.org
 Models a hypothetical retail operation
 Realistic multi-domain data warehouse environment
 Retail sales, web, catalog data, inventory, demographics & promotions
 Models several aspects of business operations
 Queries, concurrency, data loading, data maintenance
 Designed for relational data warehouse product offerings
 Four broad types of queries:
 Reporting queries, Ad-hoc queries, Iterative OLAP queries, Data mining
queries – 99 queries in all
 Designed for multiple scale factors
 100GB, 300GB, 1TB, 3TB, 10TB, 30TB and 100TB
 Designed for multi-user concurrency
 Minimum of 4 concurrent users running all 99 queries
 No vendor has ever published a formal TPC-DS benchmark

Beware of Cherry Pickers & Benchmarketing!
TPC-DS has strict requirements
 All 99 queries need to be run in their entirety
 Each query is unique and tests a different
facet of the environment
 Answer set correctness must be proven
 Result must be audited
As a result, it is not valid to:
 Select individual queries
 Change queries outside of prescribed
guidelines (“minor query modifications”)
 Alter the database schema
 Configure the system on a per-query basis
 Alter the system between single-user and
multi-user tests

About the Hadoop-DS Benchmark
 Created by IBM
 The Big Data Decision Support Benchmark (Hadoop-DS) is inspired
by, and is highly compliant with TPC-DS
 Fully complies with the TPC-DS schema requirement
 Uses all 99 queries
 Meets the multi-user requirement
 Has been audited by an approved TPC-DS auditor but as a non-TPC
benchmark
 There are deviations from TPC-DS
 No data maintenance operations, referential integrity enforcement, or ACID
property validation as these are not feasible with HDFS
 Additional statistics used (advanced Big SQL capability)
 Different metric (to avoid confusion with TPC-DS)
 No price and price/performance measures included
 Not an official TPC benchmark result

What are the key Hadoop-DS metrics?
Primary metrics:
 Qph Hadoop-DS@SF (Single User)
• Single User Queries-per-hour at a particular scaling factor)
 Qph Hadoop-DS@SF (Multi User)
• Multi User Queries-per-hour at a particular scaling factor)
Two distinct measures
 Power run – refers to a single stream of queries running in sequence
 Throughput run – refers to a multiple streams of queries executing
concurrently. A minimum of four concurrent streams is required

What did IBM Test?
30TB Hadoop-DS benchmark with Big SQL
 Executed and audited in as compliant a manner as possible
• Demonstrate the robustness of Big SQL at scale
• Demonstrate Big SQL’s ability to run all queries
• Demonstrate Big SQL’s multi-user concurrency capability
 Letter of attestation from the auditing firm and accompanying
benchmark report.
10TB subset Hadoop-DS benchmark with 3 vendors
 Compare the Big SQL, Cloudera Impala and Hortonworks Hive
 Use the subset of queries all three vendors are able to execute
 Use an identical 17 node cluster for each vendor
 Auditor reviewed method, procedures and measurement results
Two main benchmarks were executed

Benchmark Environment
Management Node
One x3650 M4 BD
Two E5-2680 v2 2.8GHz 10-core
128GB RAM, 1866MHz
2TB 3.5” HDD
Dual-port 10GbE
RHEL 6.4
EXT4/HDFS/Parquet/ORC
Data Nodes
Seventeen x3650 M4 BD
Two E5-2680 2.8GHz 10-core
128GB RAM, 1866MHz
Ten 2TB 3.5” HDD
Four 120GB 3.5” SSD
Dual-port 10GbE
RHEL 6.4
EXT4/HDFS/Parquet/ORC
Three identical clusters deployed, one for each distribution
Note: Big SQL and Impala used
Parquet file formats. Hive used
ORC

Big SQL 3.0
Working directly
from template
Compliant query
modifications
Impala 1.4.1
Working directly
from template
Compliant query
modifications
Non-compliant query
re-write
Not working or no
re-write
IBM Big SQL runs 100% of the queries
 IBM Big SQL runs all 99 queries, 12
with allowable minor modifications
 Impala runs only 52 queries – 35 out-
of-the-box and 17 with allowable
minor modifications
 Hive runs 58 queries – 32 out of the
box, and 26 with allowable minor
modifications
Hive 0.13
Working directly
from template
Compliant query
modifications
Non-compliant query
re-write
Not working or no
re-write

Query compliance by SQL-on-Hadoop offering
IBM is the only vendor
able to run all 99
Hadoop-DS queries with
minor modifications
allowable under TPC-DS
benchmark rules
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Big SQL 3.0 Impala 1.4.1 Hive 0.13
Not working or no re-write
Non-compliant query re-write
Compliant query modifications
Working directly from template
Queries

Hadoop DS – Query Compliance Detail
Small-scale Test (1 GB) 10TB scale Test
Number of queries Big SQL 3.0 Impala 1.4.1 Hive 0.13 Big SQL 3.0 Impala 1.4.1 Hive 0.13
Original query
unchanged
87 35 32 87 31 27
Minor query
modifications
12 17 26 12 11 29
Major query re-write 0 36 32 0 30 13
Percentage of
queries that run 100% 89% 91% 100% 73% 70%
Not working or no
re-write found
0 11 9 0 27 30

IBM Big SQL – Runs 100% of the queries
Key points
 In competitive environments,
many queries needed to be re-
written, some significantly
 Owing to various restrictions,
some queries could not be re-
written or failed at run-time
 Re-writing queries in a
benchmark scenario where
results are known is one thing –
doing this against real databases
in production is another
Competitive environments require significant effort at scale
Results for 10TB scale shown here

Hadoop-DS – Performance results
Elapsed time (s) Qph-HDS@10TB Big SQL Advantage
# Queries Power Throughput Power Throughput Power Throughput
IBM Big SQL 3.0 46 2,908 6,945 5,694 9,537
Impala 1.4.1 46 10,536 14,920 1,571 4,439 3.6 2.1
Hive 0.13 46 15,949 59,550 1,038 1,112 5.4 8.5
All 99 queries @ 10TB
IBM Big SQL 3.0 99 32,361 88,764 1,101 2,409
Impala 1.4.1 99 Not Possible
Hive 0.13 99 Not Possible
All 99 queries @ 30TB
IBM Big SQL 3.0 99 104,445 187,993 1,023 2,274
Impala 1.4.1 99 Not Possible
Hive 0.13 99 Not Possible

0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
18,000
Big SQL Impala Hive
Power run (single-stream) – seconds
As measured across the subset of 46 queries that Impala and Hive can both run
IBM Big SQL – Leading performance
Up to 5.4x
FASTER!!
48:28
2:55:36
4:25:49
3.6x faster than Impala, 5.4x faster than Hive
seconds

0
10000
20000
30000
40000
50000
60000
70000
Big SQL Impala Hive
Throughput run - 4 streams, average elapsed time
As measured across the subset of 46 queries that Impala and Hive can both run
IBM Big SQL – Leading performance
1:55:45
4:08:40
16:32:30
Up to 8.5x
FASTER!!
2.1x faster than Impala, 8.5x faster than Hive
seconds

30TB Hadoop-DS Results
Because other distributions could not run the 99 required
queries, it was only possible to obtain a result for Big SQL
IBM had hoped to obtain partial results @ 30TB (comparing
queries that would run across distributions)
Testing convinced us that the number of queries that
competitors could run @ 30TB was sufficiently small that a
detailed comparison would not be valid

Big SQL – Scalability and Throughput
Four concurrent query streams @30TB in 1.8x time of a single stream
0
50,000
100,000
150,000
200,000
Power Run Throughput Run
ElapsedTime(secs)
Elapsed Times for Big SQL Hadoop-DS @30TB.
Single & 4 streams. 99 queries.
99 queries
396
queries

Audited Results
 Letters of attestation are available for
both Hadoop-DS benchmarks at
10TB and 30TB scale
 InfoSizing, Transaction Processing
Performance Council Certified
Auditors verified both IBM results as
well as results on Cloudera Impala
and Hortonworks Hive
 These results are for a non-TPC
benchmark. A subset of the TPC-DS
Benchmark standard requirements
was implemented

Conclusions
Big SQL is the only SQL-on-Hadoop engine able to run a
full Hadoop-DS workload
 Complete schema
 All 99 queries
 Multi-user test
 Ran at both 10TB and 30TB data volumes
 Together this test makes for a good predictor of compatibility with
real applications
IBM Big SQL is the best performing solution by a large
margin
 ~ 3.6 times better than Cloudera Impala
 ~ 5.4 times better than Hortonworks Hive

Thank you!

Additional Slides

0 500 1,000 1,500 2,000 2,500 3,000
Big SQL
Impala
Hive .13
4 concurrent streams and 99 queries
Query throughput for Hadoop-DS @ 10TB
87
12
99 queries could not be run
Effective query throughput (Qph-HDS@10TB)

0 500 1,000 1,500 2,000 2,500
Big SQL
Impala
Hive .13
6 concurrent streams and 99 queries
Effective query throughput (Qph-HDS@30TB)
Query throughput for Hadoop-DS @ 30TB

The Common Query Set
 While Big SQL ran all queries, many of the Hadoop-DS
queries would not run on Impala or Hive
 On both platforms, some additional queries could be
made to run by re-writing the queries (something that is
not permitted in the TPC-DS benchmark specification)
 At 10TB scale, several queries failed at run-time
 This set of 46 queries are the common set that ran at 10
TB scale and could thus be compared
 The testing team deliberately included some queries
with non-compliant query modifications where the
changes were judged to be minor in order to have a
reasonable number of queries to compare
46 queries could be run on Big SQL, Impala and Hive at 10TB
Queries shown in blue are part of the common set

About the TPC-DS queries
The queries are diverse, and many are complex
Reflecting real business needs – a random sample:
 Find customers returning items more frequently than normal (q1)
 States with customers most ammenable to premium priced offers (q6)
 List key metrics for unadvertised in-store promotions by demographic (q7)
 Identify similar customers purchasing through multiple sales outlets (q10)
 Find customers shifting purchasing habits to the web (q11)
 Key measures for catalog sales fulfilled from an alternate warehouse (q16)
 Find frequently sold items and the circumstances under which repeat sales
take place (q23)
 Understand the products and retail locations where items are likely to be
return and subsequently re-purchased via the catalog (q29)
 Display customers making significant local purchases comparing to buying
potential based on dependents and vehicles owned (q34)

Benchmark Environment
X3650BD Data node #1
10 GbE switch 10 GbE private net
IBM Blue net
Mgmt net
X3650BD Master host
Three identical clusters deployed, one for each distribution

Hadoop-DS: Which SQL-on-Hadoop Rules the Herd

More Related Content

What's hot (16)

Similar to Hadoop-DS: Which SQL-on-Hadoop Rules the Herd (20)

More from IBM Analytics (20)

Recently uploaded (20)

Hadoop-DS: Which SQL-on-Hadoop Rules the Herd