Ougn2013 high speed, in-memory big data analysis with oracle exalytics

High Speed, Big Data Analysis using Exalytics
Mark Rittman, Technical Director, Rittman Mead
OUGN 2013 Conference, April 2013, Oslo

T : +44 (0) 8446 697 995 or (888) 631 1410 (USA) E : enquiries@rittmanmead.com W: www.rittmanmead.com

Mark Rittman

• Mark Rittman, Co-Founder of Rittman Mead
• Oracle ACE Director, specialising in Oracle BI&DW
• 14 Years Experience with Oracle Technology
• Regular columnist for Oracle Magazine
• Author of two Oracle Press Oracle BI books
‣ Oracle Business Intelligence Developers Guide
‣ Oracle Exalytics Revealed
• Writer for Rittman Mead Blog :
http://www.rittmanmead.com/blog
• Email : mark.rittman@rittmanmead.com
• Twitter : @markrittman


About Rittman Mead

• Oracle BI and DW platinum partner
• World leading specialist partner for technical excellence, solutions delivery and
innovation in Oracle BI
• Approximately 30 consultants worldwide
• All expert in Oracle BI and DW
• UK based
• Offices in US, Europe (Belgium) and India
• Skills in broad range of supporting Oracle tools:
‣ OBIEE
‣ OBIA
‣ ODIEE
‣ Essbase, Oracle OLAP
‣ GoldenGate
‣ Exadata


What is Oracle Exalytics?

• Hardware Element
‣ Sun Fire X4470 M2 server
‣ 1TB RAM, 40 Cores, 3.6TB HDD
• Software Element
‣ OBIEE 11.1.1.6 with Exalytics Enhancements
‣ Oracle Essbase 11.1.2 with Exalytics Enhancements
‣ Oracle TimesTen 11.2.2.2 for Exalytics
‣ New in v1.1 : Support for Oracle Endeca
Information Discovery, Golden Gate, ODI
‣ Runs on 64-bit Oracle Linux
(Exalogic distribution)
• OBIEE and Essbase are licensed as
Oracle BI Foundation
• Exalytics features can only be used in
conjunction with Exalytics hardware


Exalytics as the Exa-Machine for BI

• Runs the BI layer on a high-performance, multi-core, 1TB server
• In-memory cache used to accelerate the BI part of the stack
• If Exadata addresses 80% of the query performance,
Exalytics addresses the remaining 20% Oracle BI

‣ Consistent response times for queries
‣ In-memory caching of aggregates
‣ 40 cores for high concurrency In-Memory DB/Cache
‣ Re-engineered BI and OLAP software
that assumes 40 cores and 1TB RAM

ERP/Apps DW


Enables High-Density Analysis of Big Data

• BI tier is rarely the bottleneck, but it can be if very dense visualizations are used
‣ Sparklines, grid of charts etc
• Exalytics’ 40 cores and 1TB RAM make higher density presentation viable
‣ Single query sent to the database
‣ Exalytics breaks data up to create microcharts
• Also helps support high numbers of concurrent users (100+)


Also Supports Essbase, and Endeca Information Discovery

• In-Memory Essbase for planning, budgeting and sales analysis-style OLAP applications
• Endeca Information Discovery for search/analytic applications against diverse data

Oracle
Exalytics
In-Memory
Machine


Oracle’s Wider Strategy for Business Analytics

• Connect to all of your data, from all your sources,
• Subject it to the full range of possible inquiry
• Package solutions for known problems and fixed sources, and
• Deploy to PCs and mobile devices, on premise or in the cloud

Any Data, Full Range of Integrated On Premise,
Any Source Analytics Analytic Apps On Cloud,
On Mobile


Big Data : New Datasets of Higher Variety, Volume and Velocity

• Traditional BI datasets have been relatively small, and well structured
• Financial data and other metrics, with attributes and hierarchies to slice-and-dice it
• Big data is all about collecting and analyzing data sets of wider scope
‣ Volume - TBs of data collected from sensors, transactions and other low-granularity
data sources
‣ Variety - unstructured, semi-structured as well as structured sources
‣ Velocity - data arriving in real-time, and analyzed in real-time


Hadoop and MapReduce

• Apache Hadoop is one of the most well-known Big Data technologies
• Family of open-source products used to store, and analyze distributed datasets
• Hadoop is the enabling framework, automatically parallelises and co-ordinates jobs
• MapReduce is the programming framework
for filtering, sorting and aggregating data
‣ Map : filter data and pass on to reducers
‣ Reduce : sort, group and return results
• MapReduce jobs can be written in any
language (Java etc), but it is complicated


Big Data in the Context of the Oracle Exalytics Platform

• Exalytics, through in-memory aggregates and InfiniBand connection to Exadata,
can analyze vast (structured) datasets held in relational and OLAP databases
• Endeca Information Discovery can analyze unstructured and semi-structured sources
• InfiniBand connector to Big Data Applicance + Hadoop connector in OBIEE supports
analysis via Map/Reduce
• Oracle R distribution + Oracle Enterprise R supports SAS-style statistical analysis
of large data sets, as part of Oracle Advanced Analytics Option
• OBIEE can access Hadoop datasource through another
Apache technology called Hive


New in OBIEE 11.1.1.7 : Hadoop Connectivity through Hive

• MapReduce jobs are typically written in Java, but Hive can make this simpler
• Hive is a query environment over Hadoop/MapReduce to support SQL-like queries
• Hive server accepts HiveQL queries via HiveODBC or HiveJDBC, automatically
creates MapReduce jobs against data previously loaded into the Hive HDFS tables
• Approach used by ODI and OBIEE
to gain access to Hadoop data
• Allows Hadoop data to be accessed just like
any other data source (sort of...)


An example Hive Query Session: Connect and Display Table List

[oracle@bigdatalite ~]$ hive
Hive history file=/tmp/oracle/hive_job_log_oracle_201304170403_1991392312.txt

hive> show tables;
OK
dwh_customer
dwh_customer_tmp
Hive Server lists out all “tables”
i_dwh_customer
that have been defined within the
ratings
Hive
src_customer
environment
src_sales_person
weblog
weblog_preprocessed
weblog_sessionized
Time taken: 2.925 seconds


An example Hive Query Session: Display Table Row Count

hive> select count(*) from src_customer;
Request count(*) from table

Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 1
In order to change the average load for a reducer (in bytes):
Hive server generates
set hive.exec.reducers.bytes.per.reducer=
In order to limit the maximum number of reducers: MapReduce job to “map” table
set hive.exec.reducers.max= key/value pairs, and then reduce
In order to set a constant number of reducers: the results to table count
set mapred.reduce.tasks=
Starting Job = job_201303171815_0003, Tracking URL =
http://localhost.localdomain:50030/jobdetails.jsp?jobid=job_201303171815_0003
Kill Command = /usr/lib/hadoop-0.20/bin/
hadoop job -Dmapred.job.tracker=localhost.localdomain:8021 -kill job_201303171815_0003

2013-04-17 04:06:59,867 Stage-1 map = 0%, reduce = 0%
2013-04-17 04:07:03,926 Stage-1 map = 100%, reduce = 0%
2013-04-17 04:07:14,040 Stage-1 map = 100%, reduce = 33% MapReduce job automatically run
2013-04-17 04:07:15,049 Stage-1 map = 100%, reduce = 100% by Hive Server
Ended Job = job_201303171815_0003
OK

25
Time taken: 22.21 seconds Results returned to user


Importing Hadoop/Hive Metadata into RPD

• HiveODBC driver has to be installed into Windows environment, so that
BI Administration tool can connect to Hive and return table metadata
• Import as ODBC datasource, change physical DB type to Apache Hadoop afterwards
• Note that OBIEE queries cannot span >1 Hive schema (no table prefixes)

2

1

3


Set up ODBC Connection at the OBIEE Server (Linux Only)

• OBIEE 11.1.1.7+ ships with HiveODBC drivers, need to use 7.x versions though
• Configure the ODBC connection in odbc.ini, name needs to match RPD ODBC name
• BI Server should then be able to connect to the Hive server, and Hadoop/MapReduce
[ODBC Data Sources]
AnalyticsWeb=Oracle BI Server
Cluster=Oracle BI Server
SSL_Sample=Oracle BI Server
bigdatalite=Oracle 7.1 Apache Hive Wire Protocol

[bigdatalite]
Driver=/u01/app/Middleware/Oracle_BI1/common/ODBC/
Merant/7.0.1/lib/ARhive27.so
Description=Oracle 7.1 Apache Hive Wire Protocol
ArraySize=16384
Database=default
DefaultLongDataBuffLen=1024
EnableLongDataBuffLen=1024
EnableDescribeParam=0
Hostname=bigdatalite
LoginTimeout=30
MaxVarcharSize=2000
PortNumber=10000
RemoveColumnQualifiers=0
StringDescribeType=12
TransactionMode=0
UseCurrentSchema=0


Oracle Advanced Analytics : In-Database Predictive Analytics

• In-Database Predictive Analytics and Statistical Analysis
• Massively-Scalable, able to analyze huge volumes of data
• Exposed through SQL and R, enabling broad usage

Predictive Comprehensive Predictive Analytic
Analytics Platform Built-Inside the DatabaseData Mining, Text MiningStatistical Analysis (built on R)B
‣Scalable and Parallel
Tightly-Integrated with SQL
Works Inside Exadata and
Big Data Appliance

Text Mining

Statistics

Data Mining


What is R, and Oracle R Enterprise?

• R is a statistical language similar to
Base SAS, or SPSS
• Open-source, run by the R Project
(http://www.r-project.org)
• R environment is a suite of client/server
products for statistical data manipulation
and graphical analysis
• Modeling and Analysis performed
in-memory using “frames”
• Enhanced by community-contributed packages
• R distribute the open-source version of R
with Oracle Linux
• Oracle R Enterprise extends R to allow analysis
against frames stored in Oracle tables, views
and embed R scripts in database PL/SQL packages


Capabilities of R Compared to SQL (Built-In Stats Functions)

• R provides a wide variety of statistical and graphical techniques
• Linear and non-linear modeling, classical statistical tests, time-series analysis
• Classification, clustering and other capabilities
• Matrix arithmetic, with scalar, vector, matrices, list and data frame (aka table) structures
• Extensible through community-contributed packages, and interacts with C++, Java etc
• Available for Oracle Database 11gR2 through the Advanced Analytics Option
• Extends the (free) SQL statistical capabilities provided by Oracle Database
‣ Ranking, Windowing, Reporting
‣ Lag/Lead, First/Last
‣ Linear Regression, Inverse Percentile etc


Oracle R Enterprise

• Regular R is constrained by only working with in-memory datasets (frames)
• Data from tables and other database structures has to be loaded into memory
• Oracle R Enterprise (ORE) removes this constraint by allowing frames to reside in DB
• Automatically exploits database parallelism, plus Oracle scalability / resilience
• ORE provides three key areas of functionality
‣ Embedded R
‣ In-Database Statistics Engine
(R extensions for Oracle SQL)
‣ Transparency Layer
(access RBDMS-based frames)


Typical R / ORE Topology

• Analyst workstation contains Open Source R client tools
• Models created in-memory on workstation In-Memory
R Engine
• ORE provides capability to access datasets Analyst Workstation /
Laptop
stored in Oracle RBDMS, transparently (2 core,
16GB RAM)
• Database has ORE embedded within it
In-Memory
• ORE provides data for workstation, and R Engines
spawned by DB
can spawn its own R sessions for
in-database R analysis Oracle Database
Server with
• Enables lights-out R analysis, plus connectivity ORE

to Hadoop and Map/Reduce via
Oracle R Connector for Hadoop ORE Hadoop Connector

Hadoop Server
(Oracle Big
Data Appliance)


Oracle Exalytics In-Memory Machine as a Platform for Endeca & R

• Exalytics server acts as an
analysis workstation “on steroids”
• 1TB Ram + 40 cores for multiple R engines
• Infiniband connection to Exadata and ORE
• Endeca Information Discovery uses 40 CPU
cores for massively parallel indexing, analysis
• More of Endeca Server datastore in RAM
• OBIEE uses RAM and cores for TimesTen
datamart, in-memory data source federation
and Presentation Server caching, performance
• All products benefit from high-end server
features and fast connectivity to Exadata + BDA


An Example OBIEE / Endeca / R on Exalytics Scenario

• “Airline On-Time Performance and Causes of Flight Delays” dataset
• Provide by Bureau of Transportation Statistics, Research and Innovative Technology
Administration, United States Department of Transportation
• Dataset containing 123M rows of non-stop US domestic flight legs
• Source and destination airports, operator, aircraft type
• Type and duration of delay, delay reason
• Freely-available “big data” set
• What can OBIEE, R and EID tell us?
‣ OBIEE - dashboard analysis + drilling
‣ EID - discovery, and analysis of supporting
information describing delays, reasons etc
‣ ORE - deep insight into specific questions


Endeca Information Discovery Search/Analytic Dashboard

• Search capabilities of EID
help us explore messy and
unfamiliar data
• “McDonel” corrects to “McDonnell”
• Matches to all variants of
“McDonnell Douglas”
• Dashboard tells us that
Delta Airlines has suffered
Any variant of
most delays using McDonnell McDonnell
Douglas is
Douglas aircraft, regardless of retrieved.
variations of spelling


Discovery Environment Provides Many Ways to Explore Data

“San Fran” provides completion
suggestions across all attributes,
allowing us to discover many
representations of “San Fran” that
may be present in the data

Unlike search engines, EID
has a rich, SQL-like language for
aggregating and calculations, to
Similarly, with each step of the
populate graphs and visualizations
data exploration, all available
in the dashboard
records in current filter set are
summarised by facets - prompts
for every available attribute,
populated by only the valid values
that will lead to non-empty results
sets, allowing users to uncover
relationships and patterns in the
data using attributes that they
may not even be aware existed


Text Analytics through EID Lexical / Parsing / Enrichment Features

• Flexible key/value data model and unstructured text enrichment capabilities of EID allow
text analytics to be combined with data discovery and analytics

Most common issue for older, Whilst for newer 777, most
MD-88 aircraft is “corroded or common issue is “inoperative
cracked skin on the fuselage” emergency lights in the cabin”


Exalytics OBIEE Dashboard : “Speed of Thought” Analysis

123m rows of data, analyzed live
with detail in Oracle DB, and
aggregates in TimesTen

“Go-less” prompts and dashboard
controls for instant response to
filter changes

Interactive visuals, in the form of
maps, graphs, tables, scorecards
and KPIs


Using R to Answer In-Depth, Statistical Questions

• Are some airports more prone to delays than others?
• Are some days of the week likely to see fewer delays than others?
‣ Are these differences significant?
• How do arrival delay distributions differ for the best and worst 3 airlines compared to the
industry?
‣ Are there significant differences among airlines?
• For American Airlines, how has the distribution of delays for departures and arrivals
evolved over time?
• How do average annual arrival delays compare across select airlines?
‣ What is the underlying trend for each airline?


Preparing the Dataset for R, and Running R Queries

• Create R frames using data
from Oracle RDBMS, using
ORE transparency layer
• Create R queries to manipulate
flight delays data
• Build regression models
• Score and rank data
ontimeSubset <- subset(ONTIME_S, UNIQUECARRIER %in%
• 40 cores and 1TB RAM in Exalytics c("AA", "AS", "CO", "DL","WN","NW")) res22
<- with(ontimeSubset, tapply(ARRDELAY,
allows multiple R engines to list(UNIQUECARRIER, YEAR), mean, na.rm = TRUE))
g_range <- range(0, res22, na.rm = TRUE)
be spawned, processing larger rindex <- seq_len(nrow(res22))
cindex <- seq_len(ncol(res22))
datasets than desktop workstation par(mfrow = c(2,3))
for(i in rindex) {
could support temp <- data.frame(index = cindex, avg_delay = res22[i,])
plot(avg_delay ~ index, data = temp, col = "black",
axes = FALSE, ylim = g_range, xlab = "", ylab = "",
main = attr(res22, "dimnames")[[1]][i])
axis(1, at = cindex, labels = attr(res22, "dimnames")[[2]])
axis(2, at = 0:ceiling(g_range[2]))
abline(lm(avg_delay ~ index, data = temp), col = "green")
lines(lowess(temp$index, temp$avg_delay), col="red")}


Integrating with OBIEE and Oracle BI Publisher

• R scripts can be embedded in BI Publisher data models
• Results returned as image vectors in XML, and rendered as BI Publisher output
• R scripts can also be referenced in functions etc and included in OBIEE RPD


R Analysis Output within the OBIEE Dashboard

Regression analysis used to
predict average delay for a route,
using ORE integration within
OBIEE BI Repository

Display flight delay per airport for
top N busiest airports
with parameters that are passed to
live R engines, using R script in BIP
data model


Summary

• Analytics options available with Oracle Database and Oracle Fusion Middleware support
a wide range of analytic tools and engines
• Oracle Exalytics is an excellent platform to run these on, based on RAM and CPU #
• OBIEE, with TimesTen for Exalytics and the Summary Advisor, supports
“speed of thought” analytics using a rich, interactive dashboard
• Endeca Information Discovery provides a search / analytic interface to enable you to
discover the questions that need answering
• Oracle R Enterprise, part of the Advanced Analytics Option for Oracle Database,
enables deep analysis and insights using the R statistical language and environment
• OBIEE can now connect to Hadoop/MapReduce, through Hive
• A combined, integrated analysis toolset based on an Oracle Engineered System


Ougn2013 high speed, in-memory big data analysis with oracle exalytics

More Related Content

What's hot (20)

Similar to Ougn2013 high speed, in-memory big data analysis with oracle exalytics (20)

More from Mark Rittman (20)

Ougn2013 high speed, in-memory big data analysis with oracle exalytics