Cost/Benefit Case for IBM PureData System for Analytics

8/13/2019 Cost/Benefit Case for IBM PureData System for Analytics

http://slidepdf.com/reader/full/costbenefit-case-for-ibm-puredata-system-for-analytics 1/21

!"#"$% '()*

+!,!-.+.,/ 012.3

International Technology Group 609 Pacific Avenue, Suite 102 Santa Cruz, California 95060-4406 Telephone: 831-427-9260 Email: [email protected] Website: ITGforInfo.com

Cost/Benefit Case for

IBM PureData System for AnalyticsComparing Costs and Time to Value with

Oracle Exadata Database Machine X3



Copyright © 2013 by the International Technology Group. All rights reserved. Material, in whole or part, contained in this document may not be reproduced or

distributed by any means or in any form, including original, without the prior written permission of the International Technology Group (ITG). Information has beenobtained from sources assumed to be reliable and reflects conclusions at the time. This document was developed with International Business Machines Corporation

(IBM) funding. Although the document may utilize publicly available material from various sources, including IBM, it does not necessarily reflect the positions ofsuch sources on the issues addressed in this document. Material contained and conclusions presented in this document are subject to change without notice. All

warranties as to the accuracy, completeness or adequacy of such material are disclaimed. There shall be no liability for errors, omissions or inadequacies in thematerial contained in this document or for interpretations thereof. Trademarks included in this document are the property of their respective owners.

TABLE OF CONTENTS EXECUTIVE SUMMARY 1

Differentiating Appliances 1 Comparing Costs 1 Costs of Ownership 3 Conclusions 5

USER VIEW 7 Introduction 7 Demographics 7 FTE Staffing 8

Oracle Exadata Database Machine 8 IBM PureData System for Analytics 8

Deployment Experiences 9 Deployment Times 9 External Assistance 9

TECHNOLOGY VIEW 10 Oracle Exadata Database Machine 10

Overview 10

Hardware Base 11 IBM PureData System for Analytics 12

Overview 12 Architecture and Technology 13

DETAILED DATA 15 Installations 15 Cost Calculations 15 Cost Breakdowns 16

List of Figures

1. Three-year Costs of Ownership for IBM PureData System for Analytics N2001 and

Oracle Exadata Database Machine X3 Assuming Transfers of Oracle Licenses from

Replaced Systems – Averages for All Installations 1 2. Three-year Lost Opportunity Costs for Use of IBM PureData System for Analytics N2001 and

Oracle Exadata Database Machine X3 – All Installations 2

3. Oracle Exadata Enterprise Software Suite 3

4. Reported Overall Deployment Times for IBM PureData System for Analytics andOracle Exadata DataBase Machines 4

5. Industry Distribution of Survey Population 7

6. Principal Oracle Exadata-specific Technologies 10

7. Oracle Exadata X3 High Performance Models 12

8. Principal IBM PureData System for Analytics N2001 Filter Engines 13

9. IBM PureData System for Analytics S-Blade Structure 14

10. Current IBM PureData System for Analytics N2001 Product Line 14 11. Installations Summary 15

12. Three-year Costs of Ownership Breakdowns for use of IBM PureData System for Analytics N2001 and Oracle Exadata Database Machine X3 – Assumes Transfers of OracleLicenses from Replaced Systems 17

13. Three-year Costs of Ownership Breakdowns for use of IBM PureData System for AnalyticsN2001 and Oracle Exadata Database Machine X3 – Assumes All New Oracle Paid Licenses 18



International Technology Group 1

EXECUTIVE SUMMARY

Differentiating Appliances

In a world of big data and real-time information, challenges come from multiple directions. Organizations

must deal not only with fast growing data volumes, demands for more sophisticated interpretations of

data, and more rapid delivery of results. The general trend is toward use of analytical appliances.

Experience has shown that appliances offer better-optimized performance and faster deployment than

conventionally configured databases, servers and disk arrays. But there are differences in architecture andtechnology between platforms that can result in significantly different cost/benefit equations.

Nowhere is this more the case than for the two platforms that are the subject of this report: IBM PureDataSystem for Analytics, powered by Netezza technology, and Oracle Exadata Database Machine.

IBM PureData System for Analytics has, for more than a decade, employed a unique system design. This

avoids the complexities of SQL data warehouse architecture, enables exceptionally rapid applicationdevelopment and deployment, and executes analytical workloads in a simple and effective manner.

In comparison, Oracle Exadata systems are designed to accelerate performance for conventional Oracledata warehouse environments. Although the Exadata design incorporates new hardware and softwaretechnologies, it retains Oracle Database and Real Application Clusters (RAC) architecture. It is complex

and inefficient in the way it uses system resources.

The choice is not simply between different hardware platforms. It involves more basic decisions about the

nature of organizational informational architectures, and about trade-offs between compatibility with

existing software and skills, speed of new application deployment and costs. In a Big Data world,traditional assumptions may no longer apply.

Comparing Costs

There are significant variances between IBM PureData System for Analytics andExadata systems in two areas. The most obvious is in costs of ownership.

In four representative installations in digital media, financial services, retail andtelecommunications companies analyzed for this report, three-year costs for use of latest-

generation IBM PureData System for Analytics N2001 systems ranged from 34 to 54 percent less, and averaged 47 percent less than for Oracle Exadata X3 equivalents.

Figure 1 illustrates these results.

Figure 1: Three-year Costs of Ownership for IBM PureData System for Analytics N2001 and OracleExadata Database Machine X3 Assuming Transfers of Oracle Licenses from ReplacedSystems – Averages for All Installations

456789 .:6;6%6 <6%6=6$9+67>?@9 A*

20+ B"59<6%6 CD$%9E FG5!@68DH7$ ,'(()

!7I"?$?HG@ +6?@%9@6@79 J $"KKG5% <9K8GDE9@% B95$G@@98 367?8?H9$

*LM)*N*

OL)P)N)

Q %>G"$6@;$

3-year costs

averaged 47%

less than for

Oracle

Exadata




Calculations include acquisition and deployment costs as well as maintenance and support, personnel

costs for database, system and storage administration and facilities costs. Acquisition, maintenance andsupport costs are based on discounted vendor prices reported by users.

Oracle Exadata calculations allow for transfers of Oracle enterprise software licenses from older systemsreplaced by Exadata. If all new Oracle licenses had been acquired, three-year costs of ownership for useof IBM PureData Systems for Analytics would have averaged 63 percent less.

Another type of comparison also comes into play. In fast-moving analytics markets, time to value hassignificant cost implications.

Analytical applications may yield significant bottom-line gains, often in a

matter of weeks to months. The corollary is that delays in bringing suchapplications into production may represent significant lost revenue and profit

opportunities. Competitive position may also be eroded if, in the meantime,

others exploit such opportunities.

These effects were apparent in the same four installations employed for cost of ownership comparisons.

In these cases, lost opportunity costs for deployment of IBM PureData System for Analytics weresignificantly lower than for Oracle Exadata systems. Figure 2 illustrates disparities.

Figure 2: Three-year Lost Opportunity Costs for Use of IBM PureData System for Analytics N2001

and Oracle Exadata Database Machine X3 – All Installations

Lost opportunity costs for use of IBM PureData System for Analytics ranged from 59 to 87 percent less,

and averaged 68 percent less than for Oracle Exadata systems.

These costs are for initial applications only. In practice, organizations would continue to deploy new

applications. The cumulative impact of faster deployment over multi-year periods would be a great deallarger. Disparities in lost opportunity costs would increase by wide margins.

These and other results presented in this report are based on input from 42 companies employing IBMPureData System for Analytics or Oracle Exadata systems in comparable roles.

M(RN(

P'*N'

)*PNM

STNP

( )(( '(( *(( P(( S(( T(( O(( M(( R(( )(((

<2-2/!U +.<2! V4+B!,W

1./!2U V4+B!,W

Q %>G"$6@;$

SLR))NS

'L'OPN(

)LMSONR

RP(N'

( )L((( 'L((( *L((( PL((( SL((( TL(((

32,!,V2!U C.1X2V.C V4+B!,W

/.U.V4++Y,2V!/24,C V4+B!,W

20+ B"59<6%6 CD$%9E FG5 !@68DH7$ ,'(() 456789 .:6;6%6 <6%6=6$9 +67>?@9 A*

Q %>G"$6@;$

Lost opportunity costs

averaged 68% less than

for Oracle Exadata




Further information on profiles, methodology and assumptions employed for calculations, along with cost

breakdowns for installations and platforms may be found in the Detailed Data section of this report.

Costs of Ownership

Significant differences in costs of ownership emerged in the following areas:

• Acquisition, maintenance and support costs. For readers who have followed industry debates onthis subject, it will come as no surprise that Oracle Exadata systems are expensive.

Acquisition, maintenance and support costs for Oracle Exadata hardware, and operating system

and Storage Server software (which is list priced at $10,000 per disk drive) are substantial.Organizations must also employ an extensive suite of Oracle enterprise software that includes the

products shown in figure 3.

B5G;"7% YNCN U?$% B5?79

K95 K5G79$$G5

<6%6=6$9 ))# G5 )'7 .@%95K5?$9 .;?%?G@ "#$%&''

1968 !KK8?76%?G@ V8"$%95$ "()%'''

B65%?%?G@?@# "**%&''

!;Z6@79; VGEK59$$?G@ "**%&''

<?6#@G$%?7$ B67[ "&%'''

/"@?@# B67[ "&%'''

<6%6=6$9 U?F97D789 +6@6#9E9@% B67[ "*(%'''

/4/!U Q))SLS((

Figure 3: Oracle Exadata Enterprise Software Suite

This suite adds a further $115,500 per processor (core) in list price license costs for OracleExadata systems, plus support at 22 percent of license costs per year.

IBM PureData System for Analytics hardware, along with software providing functionality

equivalent to the Oracle Exadata stack, are offered as part of a single-price package. There is a

single annual support fee for all components.

In the cost of ownership comparisons presented in this report, combined three-year acquisition,

maintenance and support costs for IBM PureData Systems for Analytics averaged 37 percent less

than for Oracle Exadata systems.

This calculation assumes transfers of Oracle enterprise software licenses from replaced systems.

If all new licenses had been acquired, three-year acquisition, maintenance and support costs for

IBM PureData Systems for Analytics would have averaged 61 percent less.

• Personnel costs. Among organizations surveyed for this report, full time equivalent (FTE)

staffing for database administrators (DBAs) was reported to be significantly less for use of IBMPureData Systems for Analytics than for Oracle Exadata systems.

Among Oracle Exadata users, numbers of FTE DBAs ranged from one to more than three, andaveraged over 1.5. Among IBM PureData System for Analytics users, administrative overheadranged from 20 hours per year to – in the case of a company operating more than 30 systems –

two FTEs. The average was less than 0.5 FTE.




while moving beyond legacy commitments, rather than reinforcing them. The potential benefits in IT

savings, as well as in bottom-line business advantage, suggest this is an option that should be reviewed.




USER VIEW

Introduction

This section provides additional detail on users surveyed for this report and on their input on FTE staffingand deployment times. The following section, Technology View, provides further analysis of IBM

PureData System for Analytics and Oracle Exadata systems.

The final section, Detailed Data, outlines installations, and details configurations as well as FTE DBAstaffing levels for these. Methodologies and values employed for cost calculations are also presented.

Demographics

IBM PureData System for Analytics as well as Oracle Exadata users ranged from companies with fewer

than 300 employees to Fortune 100 and Financial Times Global 100 corporations. Industry distribution

was as shown in figure 5.

2@;"$%5D /9897GE<?#?%68

+9;?6

3?@6@7?68

C95Z?79$

19%6?8 4%>95 /G%68

20+ B"59<6%6 CD$%9E FG5 !@68D%?7$ , & ) ) # ')

456789 .:6;6%6 <6%6=6$9 +67>?@9 & # # ) & ')

Figure 5: Industry Distribution of Survey Population

Among 21 Exadata users, 13 had moved existing Oracle-based data warehouses and/or analyticsapplications to Oracle Exadata systems. The amount of work required set up and test Oracle Exadata

systems, and the amount of DBA retraining required were less than for first-time deployments.

Eight users reported new application deployments, although at least one of these involved modificationsto existing Oracle-based system. In others, Oracle Business Intelligence Enterprise Edition (OBIEE) or

other Oracle-based solutions were deployed for the first time. All users employed Oracle databaseselsewhere within organizations.

IBM PureData System for Analytics users often reported the same performance bottlenecks as those that

chose to deploy Oracle Exadata systems. New applications were typically deployed rather than existing

applications migrated.

Two organizations reported that they have moved to this platform after experiencing performance

difficulties with SQL Server-based data warehouses.

Users of both platforms reported significant performance improvements – sometimes of hundreds of

times – in moving to these. Comparisons were, however, typically with legacy systems deployed on

comparatively old hardware.




FTE Staffing

Oracle Exadata Database Machine

Oracle Exadata users reported between one and more than three FTE DBAs supporting this platform.

Numbers of FTE system and storage administrators were estimated as half to two. Most organizations had

formed Exadata teams with multiple skill sets.

Several organizations estimated that Oracle Exadata administration workloads involved around 60 percent

DBA, and 40 percent system and storage administration tasks. This appeared, however, to be a general

industry rule of thumb rather than based on experience.

The amount of training that had to be undertaken varied depending on the existing skill levels of

individuals assigned to Oracle Exadata administration. The shortest DBA retraining period for individuals

with Oracle database and RAC skills was said to be two weeks and the longest three months, with mostorganizations reporting four to six weeks.

Several respondents noted that in-house training was typically necessary, as skilled Exadata

administrators were difficult to find…hard to hire. Oracle or third-party consultants were also employedon a full- or part-time basis in at least seven organizations.

IBM PureData System for Analytics

Among 21 IBM PureData System for Analytics users, 18 reported that they employed less than one FTE

administrator. One organization reported a single FTE supporting multiple systems. Two reported twoFTEs supporting more than 20 and more than 30 systems respectively.

Among organizations reporting less than one FTE, 12 (67 percent) estimated than the actual number was

less than 0.5. Administration overhead was said to represent a fraction of one person’s time once a

week…two hours a week…a couple of hours a week…a few hours a month…less than an hour a day (to

administer five systems)…maybe six hours every three months…20 hours a year .

In most cases, IBM PureData System for Analytics administrators had previous experience with other business intelligence systems and/or databases, or as system administrators and engineers. Only fourorganizations reported hiring a PureData System for Analytics specialist.

A number of reasons were cited for low staffing levels. The most common was that (end) users interface

directly to the system.

It was also noted that we don’t have to build indexes…users write directly to the system, they don’t need

to go through a DBA…we work with complete data sets instead of having everything aggregated and

summarized first…we don’t have to use data models…(end) users do most of the work . They don’t have to

go through a developer unless there’s something out of the ordinary. Performance tuning requirements

were said to be virtually non-existent .

Training requirements were said to be minimal. One organization reported that its system had been up

and running for six months before any training was required – and that was for an (system) upgrade.




Deployment Experiences

Deployment Times

As a general principle, migration of systems is a simpler and faster than

deployment of new applications.

There was an obvious variance between Oracle Exadata and IBM PureData

System for Analytics users in this area. Most Oracle Exadata deploymentsinvolved migration of existing systems, while this was the case only for three

IBM PureData Systems for Analytics users – in these cases, existingMicroStrategy or SAS users had moved applications to this platform.

Even where organizations had migrated existing systems to Oracle Exadata,however, responses indicated that IBM PureData Systems for Analytics

appliances were brought into production faster.

Disparities were striking. The fastest reported IBM PureData System for Analytics deployment, forexample, involved availability of reporting applications and data to 500+ end users within four days, and

full production operations supporting 3,000+ users in less than three weeks.

The fastest Oracle Exadata deployment involved initial availability in two weeks, and full production

operations for 200 reporting users in approximately one month. In this case, however, applications anddata had been migrated before installation of the production system. The next fastest deployment time

was reported to be two months.

At the other end of the spectrum, the longest reported deployment cycles were 3 months and 24 months

for IBM PureData System for Analytics and Oracle Exadata systems respectively. These cycles were for

exceptionally large projects.

External Assistance

All Oracle Exadata users that responded on this subject reported use of Oracle Advanced Customer

Support Services and/or Oracle business partners to handle startup, and three reported assistance by

Oracle Consulting.

Engagements ranged from two weeks to more than eight months, and typically involved installation and

setup, performance tuning and migration-related work. Oracle services and/or business partner personnel

were typically involved throughout the deployment process.

Among IBM PureData System for Analytics users, 15 reported that they employed IBM or business

partner assistance, and six that in-house staff handled deployment. IBM or business partner personnel

were involved for between four days and three months.

The fastest reported

IBM PureDataSystem for Analytics

deployment provided

reporting data to

500+ users within

four days




HCC is an Oracle implementation of columnar technology that has become widely adopted for high-

volume analytics applications. It offers significantly higher levels of raw throughput and compressionthan row-based data structures.

As the term hybrid suggests, HCC combines column- and row-based data structures. This approach

retains compatibility with Oracle Database’s row-based design.

Although HCC can in principle achieve extremely high levels of compression – Oracle has claimed ratesof more than 10 times for data warehouses and 15 to 50 times for archival (infrequently accessed) data –

most users experience lower overall levels. HCC operates on table segments rather than logs, indexes andother data structures, and tends to be most effective when (1) large tables are processed sequentially, and

(2) when data is static rather than frequently modified.

Oracle Exadata software also offloads a number of secondary functions from

database servers, including data mining model scoring and incremental

database backups.

Although the core design is geared to analytical applications, two features

support other Oracle Exadata roles. I/O Resource Manager is designed to

facilitate database server consolidation for all types of Oracle-basedapplication. Flash Cache Write-Back (FCWB) accelerates processing of read

operations that are characteristic of transactional applications.

A central characteristic of the Oracle Exadata design should be highlighted. Hybridization means that: (1)

the overall environment is complex. Administrators must deal with partitioned Oracle databases, RACclusters and Exadata-specific hardware and software features. (2) Use of system resources is inefficient.High levels of system overhead are generated.

One example is HCC, which converts data to columns for compression purposes, then converts them backinto rows for processing.

As a result, the considerable processing power offered by Oracle Exadata systems does not translatedirectly into application-level performance. A greenfield design – such as that adopted by Netezza –

would have been more efficient.

Hardware Base

Oracle Exadata systems are offered in two model groups, which may be configured with varying numbers

of database and storage servers. The latter are equipped with flash memory and High Performance (HP)or High Capacity (HC) hard disk drives (HDDs).

Current systems, which were introduced in October 2012, include X3-2 models offered in eighth, quarter,half and full rack configurations (the eighth rack configuration is a repackage of the quarter-rack modelwith half of the database servers, flash cards and HDDs disabled); and X3-8 models offered in standard

configurations of one to eight racks. X3-2 models may not be upgraded to X3-8 equivalents.

In principle, Oracle supports X3-8 systems in configurations of up to 18 racks, and larger systems are

possible. There appear, however, to be few if any production systems of this size.

Oracle Exadata X3-2 database servers are equipped with Intel E5 (Sandy Bridge) eight-core processors,while X3-8 models continue to employ the same Intel E7 10-core processors as the previous generation of

X2-8 systems. Exadata Storage Servers for both model groups employ Intel E5 six-core processors, flashcards and SAS (Serial Attached SCSI) 15K (HP models) or Near-line (NL) SAS (HC models) HDDs.

Oracle Exadata

hybridization means

that: (1) the overall

environment is complex

and (2) use of system

resources is inefficient




Current high performance models are summarized in figure 7. This table shows total numbers of

processors and cores for each configuration; e.g., 4/32 x E5 2.9 GHz means four Intel E5 processors witheight cores each. Usable capacities for SAS drives are as stated by Oracle.

41!VU. .A!<!/! <!/!0!C. +!V`2,. A*^'

.2-`/` 1!Vd eY!1/.1 1!Vd `!U3 1!Vd 3YUU 1!Vd

' <6%6=6$9 C95Z95$

• (K*, I H& (D\ ]-^

(&, ]P ZR[

* C%G56#9 C95Z95$

• ,K), I H& (D' ]-^

• , I #'' ]P 46<=@

(D# _P 46<=@ 5<5@>

T*D( _P 7=</6>U

• *` I ,'' ]P *&a MRM

*'D` _P 21=F

T#D& _P 7=</6>U

' <6%6=6$9 C95Z95$

• #K)( I H& (D\ ]-^

&*( ]P ZR[

* C%G56#9 C95Z95$

• ,K), I H& (D' ]-^

• *( I #'' ]P 46<=@

#D` _P 46<=@ 5<5@>

T(D# _P 7=</6>U

• ), I ,'' ]P *&a MRM

(*D, _P 21=F

T\D& _P 7=</6>U

P <6%6=6$9 C95Z95$

• `K,# I H& (D\ ]-^

* _P ZR[

O C%G56#9 C95Z95$

• *#K`# I H& (D' ]-^

• (` I #'' ]P 46<=@

**D( _P 46<=@ 5<5@>

T)D$) _P 7=</6>U

• `# I ,'' ]P *&a MRM

&'D# _P 21=F

T((D& _P 7=</6>U

M <6%6=6$9 C95Z95$

• *,K*(` I H& (D\ ]-^

( _P ZR[

)P C%G56#9 C95Z95$

• (`K*,` I H& (D' ]-^

• &, I #'' ]P 46<=@

((D# _P 46<=@ 5<5@>

T$D#$ _P 7=</6>U

• *,` I ,'' ]P *&a MRM

*''D` _P 21=F

T#& _P 7=</6>U

41!VU. .A!<!/! <!/!0!C. +!V`2,. A*^M

) 1!Vd ' 1!VdC P 1!VdC M 1!VdC

' <6%6=6$9 C95Z95$

• *,K*,' I H$ (D# ]-^

# _P ZR[

)P C%G56#9 C95Z95$

• (`K*,` I H& (D' ]-^

• &, I #'' ]P 46<=@

((D# _P 46<=@ 5<5@>

T$D& _P 7=</6>U

• *,` I ,'' ]P *&a MRM

*''D` _P 21=F

T#& _P 7=</6>U

P <6%6=6$9 C95Z95$

• )(K)(' I H$ (D# ]-^

` _P ZR[

'M C%G56#9 C95Z95$

• &,K)), I H& (D' ]-^

• **( I #'' ]P 46<=@

##D` _P 46<=@ 5<5@>

T*#D\ _P 7=</6>U

• )), I ,'' ]P *&a MRM

('*D, _P 21=F

T\' _P 7=</6>U

M <6%6=6$9 C95Z95$

• ,#K,#' I H$ (D# ]-^

*, _P ZR[

ST C%G56#9 C95Z95$

• **(K,$( I H& (D' ]-^

• ((# I #'' ]P 46<=@

`\D, _P 46<=@ 5<5@>

T(\D\ _P 7=</6>U

• ,$( I ,'' ]P *&a MRM

#')D( _P 21=F

T*`' _P 7=</6>U

)T <6%6=6$9 C95Z95$

• *(`K*(`' I H$ (D# ]-^

)( _P ZR[

))' C%G56#9 C95Z95$

• ((#K*%)## I H& (D' ]-^

• ##` I #'' ]P 46<=@

*$\D( _P 46<=@ 5<5@>

T&\D$ _P 7=</6>U

• *%)## I ,'' ]P *&a MRM

`',D# _P 21=F

T),' _P 7=</6>U

Figure 7: Oracle Exadata X3 High Performance Models

In Oracle Exadata X3-2 eighth and quarter rack models, flash memory devices as well as HDDs employsimple (1:1) mirroring. Larger X3-2 and X3-8 models employ triple (2:1) mirroring for redundancy.

Oracle Exadata X3-2 and X3-8 models may be configured with up to 18 Storage Expansion Racks with 4,

9 or 18 additional Storage Servers each.

IBM PureData System for Analytics

Overview

IBM PureData System for Analytics is based on the Netezza Performance Server (NPS) architecture.

Netezza, which introduced its first NPS product in 2002, pioneered the data warehouse appliance market

during the 2000s. IBM acquired the company in 2010.

The core NPS design employed a unique combination of MPP, filtering, streaming and compression

technologies to deliver industry-leading query performance. Use of comparatively low-cost field-

programmable gate array (FPGA) processors and commodity disk drives offered highly competitive price/performance levels.




The simplicity of NPS architecture also contributed to its popularity. NPS systems have a longstanding

reputation for exceptionally fast deployment and low administration overhead.

The fourth generation of NPS systems, the TwinFin family, was introduced in 2009 and later rebranded asIBM PureData System for Analytics N1001. The fifth generation was introduced in January 2013.

N2001 systems employ more powerful Intel and FPGA processors, and higher-capacity SAS disks.

Performance is, according to IBM, approximately three times higher than for N1001 equivalents.

Currently, half-rack to four-rack models are offered, and it is expected that quarter-rack and larger multi-rack configurations will be added. A modified version, DB2 Analytics Accelerator for z/OS, offloads

analytical processing from IBM System z mainframes running the z/OS version of IBM’s DB2 database.

N2001 systems are designed to operate as part of the IBM Big Data Platform, an enterprise-levelarchitecture that addresses organization, management and exploitation of structured data as well as newunstructured data types.

Architecture and Technology

IBM PureData System for Analytics architecture is built around four main components:

1. MPP employs a proprietary compiler that divides query workloads into segments, namedSnippets, which are then executed in parallel by blade-based Snippet Processors (S-blades).

2. Filtering ensures that data not required for a specific query is screened out before being passed to

processor memory. According to IBM, typically 95 to 98 percent of user data is excluded in thismanner. Filtering is accomplished using the software-based engines shown in figure 8.

B5Gf97% .@#?@9 W16?>0= 37? 799>5>==<0. 536789 2<?< /<=>2 39 C<0<8>?>0= =C>5141>2 19 ?@> MHOHQ_ 56<7=> 34 ?@>

MNO =?<?>8>9? />19A C035>==>2D

X?$?=?8?%D .@#?@9 W16?>0= 37? 03;= 34 2<?< ?@<? =@3762 93? /> B1=1/6> ?3 G7>0. />19A >I>57?>2 b >1?@>0 />5<7=> ?@>

0>5302= @<2 />>9 8<0F>2 2>6>?>2 /. <9 ><061>0 G7>0.% 30 />5<7=> ?@>. @<2 />>9 <22>2 ?3 ?@>

2<?</<=> <4?>0 ?@> =?<0? 34 ?@> 5700>9? G7>0.D [<19?<19= RQJY TR?38151?.% Q39=1=?>95.% J=36<?139

: Y70</161?.U 538C61<95> <? =?0><819A =C>>2=D

19$%5?7% .@#?@9 W16?>0= 37? 799>5>==<0. 03;+6>B>6 2<?< /<=>2 39 E-HZH C0>215<?> 56<7=>= 34 ?@> MNO =?<?>8>9?

/>19A C035>==>2D

Figure 8: Principal IBM PureData System for Analytics N2001 Filter Engines

3. Streaming allows data to be transferred to and from disks more rapidly – by orders of magnitude

– than conventional MPP architectures. Data is moved between disks and S-blades in

asynchronous mode (i.e., transmission delays and protocol overhead are minimal).

In N2001 systems, each S-blade may handle up to 40 simultaneous data streams from as many

disks at rates of up to 130 megabytes (MB) per second each. In a full-rack system with 240 active

disk drives, aggregate bandwidth is 240 x 130 MB/second = 31.2 gigabytes (GB)/second or, withdata compression, close to 128 GB/second. IBM cites an overall throughput level of 450 terabytes

(TB) per hour per rack.

4. Compression employs a set of algorithms that compress numeric, integer and temporal (date and

time) data written to disk during load, insert and update operations. The system automatically

chooses the best compression algorithm to use depending upon data characteristics.

Earlier Netezza NPS systems typically realized two to four times compression rates. In N2001

models, algorithm enhancements as well as more powerful FPGAs have, according to IBM,increased typical rates by two to three times.




In N2001 as in earlier systems, S-Blades combine Intel processors executing core NPS logic with FPGA-

based filtering and control engines. A separate Compression Engine decompresses data before transferringit to Intel processors. This structure is illustrated in figure 9.

Figure 9: IBM PureData System for Analytics S-Blade Structure

Current S-Blades combine IBM HX5 blade servers with dual Intel E7-2830 eight-core 2.13 GHz Linux processors and 128 GB RAM, and dual eight-core Xilinx FPGAs. Up to seven active S-Blades aresupported in a full rack.

Disk storage is provided by 600 GB 10K SAS drives in 12 enclosures per rack. A full single rack system

contains 288 drives, of which 240 are active for database operations, 14 provide swap/log space and 34act as spares. Data is striped across primary disks, and duplicated in a RAID 1 configuration.

The current N2001 product line is summarized in figure 10.

+G;98 ((S ()( ('( (P(

Y$95 ;6%6 \, _P *\( _P )`# _P $,` _P

167[$ c * ( #!7%?Z9 ;?$[ ;5?Z9$ *(' (#' #`' \,'

!7%?Z9 C^086;9$ # $ *# (`

2@%98 K5G79$$G5 7G59$ ,# **( ((# ##`

3B-! 7G59$ ,# **( ((# ##`

Figure 10: Current IBM PureData System for Analytics N2001 Product Line

C%596E?@#

;6%6

3B-!

B65$9 .@#?@9

B5Gf97% X?$?=?8?%D 19$%5?7%

.@#?@9 .@#?@9 .@#?@9

3?8%95?@#

VGEK59$$?G@ .@#?@9

VG@%5G8

.@#?@9

`AS 2@%98 .O B5G79$$G5$




In all models, dual redundant symmetric multiprocessing (SMP) hosts implement system-wide SQL

compiler, query plan, management, optimization and other functions. In N2001 models, IBM HX5 bladeservers and RHEL 6 are employed in this role.

DETAILED DATA

Installations

Cost comparisons presented in this report were based on the installations summarized in figure 11.

1./!2U

V4+B!,W

<2-2/!U +.<2!

V4+B!,W

/.U.V4++Y,2V!/24,C

V4+B!,W

32,!,V2!U C.1X2V.C

V4+B!,W

0YC2,.CC B1432U.

[76?1+5@<99>6 0>?<16>0

$' _Pd 7=>0 2<?<

L9619> 539?>9? C03B12>0

('' _Pd 7=>0 2<?<

O<92619>% 83/16> : J9?>09>?

=>0B15>= C03B12>0

)''d _P 7=>0 2<?<

Y1B>0=141>2 876?19<?139<6

?0<219A =>0B15>=

&''d _P 7=>0 2<?<

!BBU2V!/24,C

M<6>= : 19B>9?30. <9<6.=1=%

57=?38>0 />@<B130 832>619A%

C6<9919A% 8>05@<921=19A%

430>5<=?19A% 5<8C<1A9

8<9<A>8>9?% 39>+39+39>

8<0F>?19A% B<0137=

Q7=?38>0 ?<0A>?19AV 39619>

<2B>0?1=19A : 539?>9?

2>61B>0. : 3?@>0 <CC615<?139=

430 419<951<6 =>0B15>=% 39619>

=>0B15>=%

?>6>53887915<?139=% ?0<B>6 :

3?@>0 /7=19>==>=

R9<6.=1= 34 5<66 2>?<16 0>5302

TQYZU% =351<6 8>21< : /16619A

2<?< 430 0<9A> 34 <CC615<?139=V

>DAD% 53=? : C0341?</161?.

<9<6.=1=% 7=<A> 430>5<=?19A%

5@709 0>275?139

Z><6+?18> ?0<219A <9<6.=1= 430

538C61<95> : 0>A76<?30.

<CC615<?139=V 12>9?1415<?139 34

</9308<6 C<??>09= 430 40<72

2>?>5?139% <9?1 839>.

6<792>019A >?5D

20+ BY1.<!/! CWC/.+ 341 !,!UW/2VC

e(''*S -<64 0<5F

'D& W_H YPR

=/"#'84/(& &$4/> ? 1!8)

e(''*S W766 0<5F

'D# W_H YPR

=/"#'84/(& &$4/> @ +//A)

e(''*S ( 0<5F=

'D& W_H YPR

=/"#'84/(& &$4/> @ 4'(&,)

e(''*S # 0<5F=

*D' W_H YPR

=/"#'84/(& &$4/> B 4'(&,)

41!VU. .A!<!/! <!/!0!C. +!V`2,.

X)+(S -<64 0<5F

*D# W_H=

=/"#'84/(& &$4/> C +//A)

X)+(S W766 0<5F

'D\ W_H

=/"#'84/(& &$4/> B 4'(&,)

X)+`S ( 0<5F=

*D& W_H=

=/"#'84/(& &$4/> C 4'(&,)

X)+`S ) 0<5F=

)D* W_H=

=/"#'84/(& &$4/> ;< 4'(&,)

Figure 11: Installations Summary

Oracle Exadata configurations were High Performance models equipped with 600 GB 15K SAS drives.

For these installations, configurations and FTE staffing levels for IBM PureData Systems for Analytics N2001 and Oracle Exadata X3 systems were based on user-reported data. In some cases, organizations

employed older IBM PureData Systems N1001 (TwinFin) or Oracle Exadata X2-2 or X2-8 systems.

Where this was the case, configurations were updated to latest-generation equivalents based on

comparative performance estimates by the International Technology Group.

Cost Calculations

Costs were calculated as follows:

• Oracle Exadata system costs include acquisition of Database Machines and Storage Servers;

three years of Oracle Premier Support for Systems (hardware maintenance) and Premier Supportfor Operating Systems; and licenses and three-year Software Update and Support subscriptions

for Storage Server Software.




• Oracle Exadata enterprise software costs include initial licenses and three-year Software Update

and Support subscriptions for Oracle Database 11g or 12c Enterprise Edition, RAC, Partitioning,Advanced Compression and Diagnostics, and Tuning and Database Lifecycle Management Packs.

• IBM PureData System for Analytics costs were calculated based on IBM package pricesincluding hardware as well as software components. Costs also include two years of support (thefirst year of coverage is included in initial acquisition prices).

• Personnel costs were calculated based on annual salaries of $115,633 for Oracle DatabaseMachine DBAs; $95,532 for IBM PureData System for Analytics DBAs; and $79,373 for system

and storage administrators for both platforms. Salaries were increased by 55.48 percent to allowfor bonuses, benefits and other per capita costs, and multiplied for three years.

Calculations also include appropriate training courses provided by Oracle or IBM. The duration

of these, and the number of individuals trained varied between installations.

• Deployment costs for Oracle Exadata systems for all installations, and for IBM PureData System

for Analytics for the telecommunications and financial services companies were for time spent by

external professional services staff, charged at $2,000 per person-day for both platforms.

Deployment costs for IBM PureData System for Analytics for the retail and digital media

companies calculated based on single-price service offerings from IBM business partners.

Calculations for both platforms for all installations include T&E expenses for onsite time by

external staff.

• Facilities costs are for energy consumption by appliances. Calculations are based on vendor

specifications and assume near-24/365 operations over a three-year period. A conservative

assumption for average cost per kilowatt-hour was employed.

Cost calculations do not include Oracle Customer Data and Device Retention or IBM Hard Drive

Retention Services.

All cost values are for the United States.

Cost Breakdowns

Costs of ownership breakdowns are presented in figures 12 and 13.

1./!2U

V4+B!,W

<2-2/!U +.<2!

V4+B!,W

/.U.V4++Y,2V!/24,C

V4+B!,W

32,!,V2!U C.1X2V.C

V4+B!,W

20+ BY1.<!/! CWC/.+ 341 !,!UW/2VC ,'(()

!7I"?$?%?G@ $'*%(&' *%((\%&'' (%#$&%''' #%$(%&''

+6?@%9@6@79 J $"KKG5% (&(%#&' ##(%&)' `\*%''' *%$$(%*''

<9K8GDE9@% \`%&)' *,&%\`& )``%\'' ,,)%#&`B95$G@@98 ((`%`'' *`#%(#' ())%`'' #&,%&\\

367?8?%?9$ *'%`)\ *\%)&& )`%$*' $$%#(*

/4/!U aQb )L'R)LMTR 'L(P)LT)( PL('OLP)( OLMR'L(OM

41!VU. .A!<!/! <!/!0!C. +!V`2,. A*

!7I"?$?%?G@ ,#$%&'' *%,*(%*'' #%)#`%#'' ,%*&)%'''

+6?@%9@6@79 J $"KKG5% )&$%,)' \($%,\, (%&$#%(,# )%,*$%#,'

<9K8GDE9@% ('\%&'( #*\%''# *%'&$%,,( (%`&`%,&&

B95$G@@98 $)$%&&' #$%(#' $`)%'(\ *%&\`%,''

367?8?%?9$ **%`$* (*%,$` &#%*\# `*%(\(




/4/!U aQb )LRTPL(S* *LPO*LO)M MLM)OLSPR )PL*(RL((O

Figure 12: Three-year Costs of Ownership Breakdowns for use of IBM PureData System for Analytics N2001 and Oracle Exadata Database Machine X3 – Assumes Transfers ofOracle Licenses from Replaced Systems




1./!2U

V4+B!,W

<2-2/!U +.<2!

V4+B!,W

/.U.V4++Y,2V!/24,C

V4+B!,W

32,!,V2!U C.1X2V.C

V4+B!,W

20+ BY1.<!/! CWC/.+ 341 !,!UW/2VC ,'(()

!7I"?$?%?G@ $'*%(&' *%((\%&'' (%#$&%''' #%$(%&''

+6?@%9@6@79 J $"KKG5% (&(%#&' ##(%&)' `\*%''' *%$$(%*''

<9K8GDE9@% \`%&)' *,&%\`& )``%\'' ,,)%#&`

B95$G@@98 ((`%`'' *`#%(#' ())%`'' #&,%&\\

367?8?%?9$ *'%`)\ *\%)&& )`%$*' $$%#(*

/4/!U aQb )L'R)LMTR 'L(P)LT)( PL('OLP)( OLMR'L(OM

41!VU. .A!<!/! <!/!0!C. +!V`2,. A*

!7I"?$?%?G@ *%)`,%$'' (%)&*%)'' ,%&,,%''' \%`#\%'''

+6?@%9@6@79 J $"KKG5% `#&%&'( *%#*&%&,` #%')$%``' ,%'&,%`('

<9K8GDE9@% ('\%&'( #*\%''# *%'&$%,,( (%`&`%,&&

B95$G@@98 $)$%&&' #$%(#' $`)%'(\ *%&\`%,''

367?8?%?9$ **%`$* (*%,$` &#%*\# `*%(\(

/4/!U aQb *L)R)L)'S PLO((LOR( )'LPRMLOTS '(LPPPL*TO

Figure 13: Three-year Costs of Ownership Breakdowns for use of IBM PureData System for Analytics N2001 and Oracle Exadata Database Machine X3 – Assumes All New Oracle

Paid Licenses



ABOUT THE INTERNATIONAL TECHNOLOGYGROUP

ITG sharpens your awareness of what’s happening and your competitive edge

. . . this could affect your future growth and profit prospects

International Technology Group (ITG), established in 1983, is an independent research and management

consulting firm specializing in information technology (IT) investment strategy, cost/benefit metrics,infrastructure studies, deployment tactics, business alignment and financial analysis.

ITG was an early innovator and pioneer in developing total cost of ownership (TCO) and return oninvestment (ROI) processes and methodologies. In 2004, the firm received a Decade of Education Awardfrom the Information Technology Financial Management Association (ITFMA), the leading professional

association dedicated to education and advancement of financial management practices in end-user IT

organizations.

The firm has undertaken more than 120 major consulting projects, released more than 250 management

reports and white papers and more than 1,800 briefings and presentations to individual clients, user

groups, industry conferences and seminars throughout the world.

Client services are designed to provide factual data and reliable documentation to assist in the decision-making process. Information provided establishes the basis for developing tactical and strategic plans.

Important developments are analyzed and practical guidance is offered on the most effective ways torespond to changes that may impact complex IT deployment agendas.

A broad range of services is offered, furnishing clients with the information necessary to complement

their internal capabilities and resources. Customized client programs involve various combinations of thefollowing deliverables:

Status Reports In-depth studies of important issues

Management Briefs Detailed analysis of significant developments

Management Briefings Periodic interactive meetings with management

Executive Presentations Scheduled strategic presentations for decision-makers

Email Communications Timely replies to informational requests

Telephone Consultation Immediate response to informational needs

Clients include a cross section of IT end users in the private and public sectors representing multinationalcorporations, industrial companies, financial institutions, service organizations, educational institutions,

federal and state government agencies as well as IT system suppliers, software vendors and service firms.

Federal government clients have included agencies within the Department of Defense (e.g., DISA),Department of Transportation (e.g., FAA) and Department of Treasury (e.g., US Mint).

International Technology Group609 Pacific Avenue, Suite 102Santa Cruz, California 95060-4406Telephone: 831-427-9260

Email: [email protected]

Website: ITGforInfo.com ---WAL12395-USEN-00

Cost/Benefit Case for IBM PureData System for Analytics

Documents

Transcript of Cost/Benefit Case for IBM PureData System for Analytics