Big Data Performance Delivered : Oracle NoSQL DataBase on Cisco UCS

© 2010 Cisco and/or its affiliates. All rights reserved. 1

Raghunath Nambiar, Distinguished Engineer, Cisco

Ashok Joshi, Senior Director, Oracle

Big Data Performance Delivered: Oracle NoSQL Database on Cisco UCS


• Cisco Oracle Partnership and Cisco UCS Momentum

• Cisco UCS Reference Configurations for Oracle NoSQL Database

• Oracle NoSQL Database: Technology and Use Cases

• Big Data Processing: An Analogy


20 years of partnership

Initially in networking, now extended to UCS

Foundational customers to each other

Joint engineering engagements

Certifications

Cisco Validated Design (CVD)

Benchmarks

Growing Number of Joint Customers



• In Q4FY12 Data Center Revenue increased 90% year over year and 42% quarter over quarter

• In Q4FY12, UCS bookings increased 58%, year over year

• As of Q4FY12 Cisco UCS achieved an annualized run rate of over 1.6 billion dollars

• As of August 2012, there are over 15,800 unique UCS customers

• More than half of all Fortune 500 customers have invested in UCS

• 347 customers have booked over one million in UCS product and 813 have booked over $500,000

• Over 2600 Channel Partners are actively selling UCS and over 1200 UCS specialized partners in the channel worldwide

• 65 orld Record Performance Benchmarks to date


Maintained #2 in N. America

(21.9%) and #2 in the US

(22.2%)1

x86 Blade servers are growing

over twice as fast as the overall

x86 computing market2

Source: 1 IDC Worldwide Quarterly Server Tracker, Q2 2012, August, 2012, Revenue Share 2 IDC Q1 CY12 Server Forecaster, Based on x86 Blade Revenue

Worldwide

North America

Maintained #3 worldwide in x86

Blades with 15.2%, just behind

IBM’s 15.4%1

UCS momentum is fueled by

game-changing innovation;

Cisco is quickly passing

established players1

UCS After Only

Three Years

X86 S

erv

er

Bla

de M

ark

et

Share

, Q

2C

Y12

1

0% 10% 20% 30% 40% 50%

Hitachi Ltd

Oracle

Fujitsu

NEC

Dell

Cisco

IBM

HP

UCS #3 with 15.2%

0% 10% 20% 30% 40% 50%

Oracle

Dell

IBM

Cisco

HP

UCS #2 with 21.9%



Smart Solutions

Vblock

FlexPod

VXI

RISC

Migration

Enterprise Apps

Databases Business Analytics and Big Data

Virtual Desktop

Healthcare Financial Services

Manufacturing Retail Vertical Solution Focus

Applications

Management

Operating System

and Hypervisor

Compute

Network

Virtualization

Operating Systems

Applications

Information

VSPEX

http://en.wikipedia.org/wiki/File:Suse_logo.svg


Best CPU Performance

Best Virtualization Performance

Best Cloud Computing

Performance

Best Enterprise Application

Performance

Best Enterprise Middleware

Performance

Best HPC Performance

VMmark 1.x

2-socket B200 M1

VMmark 1.x

Blade Server B440

M1

VMmark 2.1

Overall C460 M2

VMmark 2.1

4-socket C460 M2

SPECfp_rate_base2

006 2-socket

C260 M2

SPECint_rate_base

2006 X86 2-socket

B200 M1

SPECompLbase200

1

2-socket B200 M2

SPECompLbase200

1

2-socket B200 M2

SPECjbb2005

2-socket C260 M2

SPECjbb2005

X86 2-socket B200

M2

Oracle E-Business

Suite Medium Model

Payroll Batch B200

M2

Oracle E-Business

Suite Medium Model

Order-to-Cash B200

M2

VMmark 1.x

Overall C460 M1

VMmark 1.x

2-socket B200 M1

VMmark 2.0

Overall B200 M2

VMmark 2.1

2-socket Blade B200

M2

SPECint_rate_base

2006 X86 2-socket

B200 M2

SPECfp_rate_base2

006 X86 2-socket

B200 M1

SPECompMbase20

01

2-socket B200 M2

SPECompMbase20

01

2-socket B200 M2

SPECjbb2005

X86 2-socket B230

M1

SPECjAppServer20

04

1-node 2-socket

C250 M2

Oracle E-Business

Suite Medium Model

Payroll Batch B200

M2

Oracle E-Business

Suite Ex-large Model

Payroll Batch B200

M2

VMmark 1.x

Overall C460 M1

SPECint_rate2006

X86 4-socket C460

M2

SPECint_rate_base

2006 X86 2-socket

B200 M2

SPECompLbase200

1

2-socket B230 M2

LS-Dyna

4-socket C460 M1

SPECjbb2005

2-socket B230 M2

SPECjAppServer20

04

2-node B230 M1

Oracle E-Business

Suite Xtra Large

Model Payroll B200

M3

SPECjEnterprise201

0 Overall B440 M1

VMmark 1.x

2 –socket Blade

B230 M1

VMmark 1.x

2-socket B250 M2

VMmark 2.1

Two–node 4-socket

C460 M2

SPECint_rate_base

2006 2-socket

C260 M2

SPECfp_rate_base2

006 X86 2-socket

B200 M2

SPECompMbase20

01

2-socket B230 M2

LinPack

2-socket B200 M2

SPECjbb2005

2-socket B230 M2

SPECjbb2005

X86 4-socket C460

M1

Oracle E-Business

Suite Large Model

Order-to-Cash B200

M3

Oracle E-Business

Suite Xtra Large

Model Payroll Batch

B230 M2

SPECint_rate_base

2006 X86 4-socket

C460 M1

SPECfp_rate_base2

006 X86 4-socket

C460 M1

SPECompMbase200

1

4-socket C460 M2

SPECompMbase200

1

4-socket C460 M1

SPECjbb2005

4-socket B440 M2

SPECjbb2005

X86 2-socket B230

M1

SPECjEnteprise201

0

2-node B440 M2

TPC-C

Oracle DB 11g &

OEL

C250 M2

SPECompMbase200

1

4-socket C460 M1

SPECompMbase200

1

4-socket C460 M2

SPECjbb2005

X86 2-socket C220

M3

SPECfp_rate_base2

006 X86 2-socket

C220 M3

SPECint_rate_base

2006 X86 2-socket

C220 M3

TPC-H 1000GB

Microsoft SQL

Server

C460 M2

TPC-H 100GB

VectorWise

C250 M2

SPECompMbase200

1

2-socket C240 M3

SPECfp_base2006

X86 2-socket C220

M3

TPC-H 300GB

VectorWise

C250 M2

SPECompLbase2001

2-socket C220 M3

Cisco UCS Benchmarks that held world record performance records as of date of publication

A History of World Record Performance on Industry Standard Benchmarks

Best HPC Performance

HPC


23 World Record Oracle-based Benchmarks !

Oracle Application Standard

Benchmark 12.1.3

Payroll Batch – Extra Large kit

Order to Cash Batch – Large kit

B200 M3

Siebel 8.1.1.4

10,000 Users

World Record Benchmarks Performance Proof Points

Oracle Application Standard

Benchmark 12.1.2

Payroll Batch – Extra Large kit

Order to Cash Batch – Large kit

B200 M3

JD Edwards 9.1

Oracle VM

2,000 Users

PeopleSoft 9.0

North American Payroll

255,319 Payments/Hour

YCSB Mix Wordkload

1,244,550 Mix OPS/Sec

TPC-C Two Socket

1,609,186.39 tpmC, $0.47/tpmC

C240 M3


1,024,380 1,053,100

1,503,544

1,609,186

$0.65 $0.58

$0.53 $0.47

$0.00

$0.20

$0.40

$0.60

$0.80

$1.00

$1.20

$1.40

0

200,000

400,000

600,000

800,000

1,000,000

1,200,000

1,400,000

1,600,000

1,800,000

HP ProLiant DL380 G7 Cisco UCS C250 M2 IBM Flex System x240 Cisco UCS C240 M3

tpmC

$/tpmC

TP

C-C

Tra

nsa

ctio

ns-P

er-

Min

ute

(tp

mC

)

TP

C-C

Price

-Pe

rfo

rma

nce (

$/t

pm

C)

* As of September27,2012. Visit www.tpc.org for up-to-date information

2 x Intel Xeon E5-2690

2.90GHz 768GB Memory

7% faster, 11% less

expensive than IBM

using identical Intel

processors and memory

capacity

New

Romley EP

2 x Intel Xeon X5690

3.46 GHz 384GB Memory

3% faster, 11% less

expensive than HP

using identical Intel

processors and memory

capacity

Westmere EP

© 2010 Cisco and/or its affiliates. All rights reserved. 12 Source: http://blog.programmableweb.com/2011/05/25/who-belongs-to-the-api-billionaires-club/

Twitter ~ 150API calls/sec

• Workload: YCSB Mixed workload

(5% updates/95% reads)

• Result: 1,244,550 OPS/Sec

• 0.88 ms Read Latency

• 4.47 ms Update Latency

2 x Intel Xeon E5-2690 Processors

128 GB Memory

Cisco UCS VIC 1225

CPA Enhanced with Flash

Cisco Big Data Common Platform (CPA) is a highly scalable architecture

designed to meet variety of scale-out application demands

Using Yahoo! Cloud Serving Benchmark (YCSB)



Big Data is Big Market & Big Business

$50 Billion Market by 2017


• Major role in internet revolution. Global leader in networking

• Complete Big Data platform portfolio with UCS

• Areas of focus: Massive scale-out solutions including NoSQL and Hadoop

• Exclusive partnership: Cisco UCS + Oracle NoSQL Database. Joint lab. Massive scale-testing

• Differentiation: Seamless management integration and data integration capability with UCS Oracle infrastructure


UCS 6200 Series

Fabric Internments

Nexus 2232

Fabric Extenders

UCS Manager

UCS 240 M3 Servers

LAN, SAN, Management

Building Blocks Cisco Big Data Common Platform (CPA) is a highly scalable architecture

designed to meet variety of scale-out application demands


6200 Series FIs

and

2232 Series FEX

16 x C240 M3 Servers

256 Processor Cores

4 TB Memory (Up to 12

TB Supported)

15K SAS : 113 TB

10K SAS: : 338 TB

7.2K SATA : 384 TB

Flash : 20TB

Reference Configurations

160 Servers in a single

UCS Management domain

10K Servers with

UCS Central


Extendable to multi-data center

implementations for disaster recovery

and business continuity

Big Data

NoSQL Database

Availability

Backup

Snapshot

Unified management

Seamless data integration

Integrated

Solutions:

FlexPod

Vblock

Oracle Database

and Enterprise

Applications

Cisco UCS B-Series

Blade Servers

Oracle NoSQL

Database

Cisco UCS C-Series

Rack Mount Servers

Internal IO

bandwidth and

capacity

Single magt interface

Global view of all components

status inc failure power

One click bios/fw upgrades

http://mapr.com/

http://paraccel.com/


• Operational Simplification:

Simplified &policy-based

management to manage the

cluster

• Modular Solution: With

modular framework,

infrastructure and expansion

modules simplify deployment

• Risk-reduction: Pre-

validation, sizing and

performance optimization

reduces integration and

deployment risk

• Lower TCO: With reduced

managed switch nodes in

scale-out solutions

Business Benefits

• Scalability: Modular building block, scalable up to 2560 processor cores and 7.2 PB in single management domain

• Performance: Best in class performance of compute and network for massively scale-out applications

• Management & Monitoring: Unified management across cluster (up to 10000 nodes)

Architectural Benefits

UCS Fabric Interconnects provide

Common Management Plane for

Scale-out

UCS C-Series Rack servers with

Internal Storage provide the

Compute Nodes

Components


SmartPlay Solution Bundles

Tested, validated and optimized for cost and rapid deployments




Make Better Decisions Using Big Data

ANALYZE

DECIDE ACQUIRE

ORGANIZE


Oracle Exadata

Oracle Advanced Analytics

Oracle Big Data Appliance

Big Data

Acquisition

NoSQL DB Driver

Application

HDFS, Hadoop, CDH

Map Reduce

ORCH - Stats Map Reduce

Hive - Activities Map Reduce

Pig - Sessionize

Real Time

Access

Batch

Processing

Big Data

Connectors

ORE, OEP,

Endeca

OLH, ODC,

ODI, External

Tables

Oracle Exalytics

Endeca Information Discovery

Oracle Business Intelligence EE


• Categories of big data solutions:

Use new data in order to “enrich” value of existing databases and repositories

Invent new applications with big data

• Many technologies: NoSQL stores, distributed file systems, relational database systems… Simple key-value operations, map-reduce, SQL, Analytics, graph traversal, semantic analysis, spatial relationships, custom code…

Endpoint facing” or end-user facing systems,

storage,

transformation,

analytics and Business intelligence


• Simple Key-Value Data Model

• Sharded, Horizontally Scalable

• Highly Available

• Elastic Configuration

• Simple administration

• Transparent load balancing

• Commercial grade software and support

Features

Scalable, Highly Available, Key-Value Database

Application

Storage Nodes Datacenter B

Storage Nodes Datacenter A

Application

NoSQL DB Driver

Application

NoSQL DB Driver

Application


• Scalability Dynamic data sharding (shared nothing) and distribution

Optimized data access via intelligent driver

• High availability One or more replicas

Resilient to failures within a shard

No single point of contention/control in entire system

Disaster recovery through location of replicas

• Transparent load balancing Reads from master or replicas

Driver is network topology & latency aware

Scalable, Highly Available, Optimized


• High performance and predictable latency

• Transactions and indexed access

• Flexible schema – Diverse, changing data sets

Key-value store; major key/minor key paradigm for modeling data

• Horizontal scalability and Availability

• Simplified management

• Based on mature and proven technology

Berkeley DB Java Edition High availability


• Cost

TBs to PBs of simple/unstructured data

Lower $ per operation (hardware and RDBMS license fees)

• Scalability – Scale out, don’t scale up

• Flexible schema – Diverse, changing data sets

• Performance

High rate of data capture

High volume of simple queries

Eliminate ORM overhead

• Availability

Low cost highly available, distributed data store


Oracle NoSQL Database Use Cases

Use Cases SIMPLE QUERIES DYNAMIC SCHEMA HIGH VOLUME REAL-TIME DATA ACCESS

High-throughput data capture

Customer profile management

Click-through data processing

Sensor & statistics data capture

Social networks

Personalization

Authentication & Content management Mobile application backend infrastructure

NoSQL Database is the “customer-facing” repository


• - provides PaaS platform for deploying applications over the cloud

Oracle NoSQL Database is exposed as a service through their cloud infrastructure

• - Oracle Platinum Partner, built an online betting platform for their customer (Passoker) using Oracle NoSQL Database

Success Stories


• Challenges

Provide a platform for real-time sports data feeds

Establish a technology to receive XML files and rapidly process them for relevance

Reduce development time and simplify day-to-day administration

Scale-out the platform as demand grows and players take up gaming in greater numbers

• Benefits of Oracle NoSQL Database

Reduced development and implementation time by 75%

Simplified day-to-day administration

Simplified rapid correlation between events and gaming

Improved operational efficiency due to ongoing commercial support

Eliminated daily maintenance related to single node point-of-failure

Enabled more application development and integration, better testing, shorter time to market

Global Sports Betting application


Scenario •Credit card swipe at point of sale (PoS) •Look up customer profile (interactive) •Authorization systems determines approval •Approval sent to PoS and recorded

Key Features •Fast response time, high throughput •High availability and scalability •Huge transaction rates, simple data •Connectivity to Big Data Analytics and Discovery

Goal: Better, Faster Fraud Detection and Response

Look Up

NoSQL DB

Authorization

Server Approval

Status

Record

Transaction


Capture Patient

Monitoring Data

Alerting

System NoSQL

DB

Scenario •Patient uses multiple devices at home •Medical data periodically sent to NoSQL DB •App monitors and alerts patient state •Alerts sent to medical or emergency personnel, recorded in profile

Key Features •High performance and high availability •High throughput event capture •Huge volumes of data •Simple data, flexible data model

Goal: Better Patient Care at Lower Cost


Scenario •User clicks on a web page •Cookie used to look up user profile •Appropriate ad placed on page

Key Features •Fast response times and high throughput •High availability and scalability •Simple data, flexible data model •Connectivity to Big Data Analytics and Discovery

Goal: Optimize Internet Advertising to Increase Revenues

Ad to be

Displayed

Expert

System NoSQL

DB

Look Up User Profile


Architecture

Hadoop Cluster

Multi-Dimensional

Reporting

RDBMS

NoSQL DB Driver

Ad Server Application


• Problem

Very low latency requirements – Publishers require < 75 ms response time from the ad serving platform

Extreme data velocity – Multi-millions of requests per second

Highly available – 24/7 sites

Revenue maximization – Deliver the most relevant ad to maximize revenue

• Solution – Where to use a NoSQL Database?

Cookie store – NoSQL database used to store cookies and associated behavioral segments

Track behavioral data – Beacons utilized during browsing to store timestamp, frequency, and behavioral segments by cookie

Optimize ad delivery – Recency, frequency, and behavioral segments used to determine optimal ad to deliver to user


Oracle NoSQL Database Cisco UCS

Simple data model Radically simplified

architecture

Easy to scale Modular scalability

Predictive performance High performance

High availability Fully redundant architecture

Integration with UCS Oracle Database Infrastructure

Enterprise-class service and support


Oracle on Cisco UCS

www.cisco.com/go/oracle

Oracle NoSQL Database

http://www.oracle.com/us/products/database/nosql

Cisco UCS Ecosystem for Oracle:

Extend Support to Big Data and Oracle NoSQL Database http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/ns944/le_34301_wp.PDF



http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/ns944/le_34301_wp.PDF

http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/ns944/le_34301_wp.PDF


Thank you.

Big Data Performance Delivered : Oracle NoSQL DataBase on Cisco UCS

Technology

flexible schema

intel xeon

suite medium

suite xtra

payroll batch

changing data

socket b200

socket b200