Anda di halaman 1dari 15

(707) 595-3607

International Technology Group


609 Pacific Avenue, Suite 102
Santa Cruz, California 95060-4406
Telephone: 831-427-9260
Email: Contact@ITGforInfo.com
Website: ITGforInfo.com



Management Report
June 2014
Comparing IBM Power Systems to
Oracle Exadata Database Machine
Cost/Benefit Case for Transactional Applications
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine i
Table of Contents
INTRODUCTION 1

COMPARING COST EFFECTIVENESS 1
Summary of Results 1
Cost Variations 2
Key Differentiators 3
TECHNOLOGY DIFFERENCES 6
Core Design: Oracle Exadata Database Machine 6
Core Design: Power Systems 7
System Architecture 7
Disk Arrays 9
Flash Storage 10
Availability Optimization 11
Consolidation and Virtualization 11
CONCLUSIONS 12

List of Figures
1. Average Three-year Costs for Use of IBM Power Systems and
Oracle Exadata Database Machine for Transactional Solutions 1
2. Average Three-year Costs for Use of IBM Power Systems and
Oracle Exadata Database Machine by Numbers of Active Users 2
3. Average Three-year Costs for Use of IBM Power Systems and
Oracle Exadata Database Platform by Oracle Exadata Rack Size 3
4. Principal Differences Between Oracle Exadata Database Machine and
IBM Power Systems with Disk Arrays 5
5. Core Oracle Exadata Database Machine Design 6
6. Combined Power Systems, PowerVM and AIX Architecture 8
7. PowerVM Storage Pool Example 9


International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 1
Introduction
Demand for transaction processing solutions continues to grow. Although topics such as analytics, clouds and Big
Data tend to dominate headlines, transactional workloads are expanding in most organizations. There is a steady
momentum of new deployments and legacy replacements in enterprise resource planning (ERP), transactional e-
commerce and a wide range of other business-critical systems.
Most transaction processing systems continue to be deployed on UNIX or x86 servers, although Oracle has also
targeted its Exadata Database Machine at this space. Oracles commitment to this platform, its appeal to many
organizations with major investments in Oracle and Oracle-based software, and the companys aggressive
marketing have made Exadata an increasingly visible contender.
Exadata can clearly run transactional applications. Any server can. The real question is whether it can do so in a
cost-effective manner.
That question is the focus of this report. It compares three-year costs for use of Exadata systems and the
industrys dominant UNIX server platform, IBM Power Systems, for comparable transaction processing solutions.
It draws upon the experiences of organizations that have deployed such solutions on both platforms.
Comparisons are based on a set of 25 profile installations employing Oracle E-Business Suite (EBS), JD Edwards
EnterpriseOne, PeopleSoft and Retail solutions. Installations were in manufacturing (12), agribusiness,
distribution, engineering and construction, health care, media, retail and IT services companies, and government
agencies. Organizations ranged from 150 to more than 40,000 employees.
Profiles were constructed from a broader survey of 127 organizations in the same industries, with generally
similar business profiles and numbers of employees, which had deployed one or more of these solutions on Oracle
Exadata or IBM Power Systems.
Comparing Cost Effectiveness
Summary of Results
Overall results of cost comparisons are summarized in figure 1. Even allowing for aggressive Oracle discounting,
costs for use of Power Systems for all solutions averaged 31 percent less than for Exadata systems.
Figure 1: Average Three-year Costs for Use of IBM Power Systems and Oracle Exadata Database Machine for
Transactional Solutions
Costs include system acquisition and maintenance, licenses and support for systems as well as Oracle enterprise
software and energy costs. Systems software includes operating systems and (for Power Systems) PowerVM
virtualization. Enterprise software includes Oracle Real Application Clusters (RAC) and other tools that must be
separately licensed for use with Exadata. Power Systems are configured with the same enterprise software.
Oracle Exadata Database
Machine
IBM Power Systems & disk
arrays
System acquisition & support Systems software licenses & support
Enterprise software licenses & support Energy
988.3
1,436.1
$ thousands
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 2
Comparisons are between latest-generation Exadata X4-2 eighth-, quarter- and half-rack High Performance
models; and IBM POWER7+-based systems with between 4 and 64 cores, the AIX operating system and
PowerVM virtualization. Power Systems are supported by IBM Storwize, XIV Storage System or System Storage
DS8870 disk arrays, configured as tiered systems with solid state drives (SSDs) and hard disk drives (HDDs).
Comparisons are for production systems only. Most organizations also employed separate platforms for test,
development, quality assurance and other non-production functions, and as failover systems. All calculations
except those for energy consumption are based on discounted prices as reported by users.
As the Exadata X4 series is comparatively new (it was introduced in December 2013), many users employed
older Exadata models. Configurations were updated based on International Technology Group (ITG) estimates of
comparative transactional performance. Older Power Systems and disk arrays were also updated to latest-
generation models based on comparative transactional performance and used capacity respectively.
Personnel and installation costs are not included. As the same Oracle software suites were deployed on both
platforms, personnel costs for database administrators (DBAs) and application specialists were similar, and there
was no obvious difference in full time equivalent (FTE) staffing for server and storage administration tasks.
Although claims have been made that Exadata systems can be started up more rapidly than conventional servers
and disk arrays, actual times for both varied widely. The speed at which both were brought into production
depended on a number of factors, including the extent to which systems were preconfigured, tested and tuned for
application- and user-specific requirements.
Where this was the case for Power servers and IBM disk arrays, start-up times and costs in comparable
installations were generally similar to those for Exadata.
Cost Variations
Although results were generally consistent for all types of solutions, there were a number of variations between
system sizes and applications.
As figure 2 illustrates, the largest disparities were in installations with fewer than 1,000 active users. In those with
over 1,000, use of more expensive IBM high-end Power 770 and 780 servers and DS8870 disk arrays resulted in
closer three-year costs.
Figure 2: Average Three-year Costs for Use of IBM Power Systems and Oracle Exadata Database Machine by
Numbers of Active Users
2,636.3
1,230.5
1,452.2
601.7
2,217.1
892.8
722.5
220.4
1,000+
500-999
250-499
<250
N
u
m
b
e
r

o
f

A
c
t
i
v
e

U
s
e
r
s

IBM Power Systems & disk arrays
Oracle Exadata Database Machine
$ thousands
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 3
Three-year costs for use of Power Systems averaged 63 percent less for installations with fewer than 250 users, 50
percent less for 250 to 499 users, 27 percent less for 500 to 999 users and 16 percent less for 1,000+ users.
(Numbers of active users is an approximate measurement of comparative performance. Although interactive
workloads are generally similar, this metric does not allow for batch processing. Numbers of active users is,
however, widely employed as a configuration sizing parameter for ERP solutions, and provides at least a general
standard for comparison.)
A breakdown of comparative costs by size of Oracle Exadata platform shows a similar picture. As figure 3
illustrates, the largest disparities were where Power Systems acted as alternatives to eighth- and quarter-rack
Exadata models. Higher comparative costs for Power Systems equivalent to half-rack models again reflect the use
of IBM Power 770 and 780 servers and DS8870 disk arrays in this bracket.
Figure 3: Average Three-year Costs for Use of IBM Power Systems and Oracle Exadata Database Platform by
Oracle Exadata Rack Size
Three-year costs for Power Systems employed as alternatives to Exadata eighth-rack models averaged 41 percent
less, while costs for quarter- and half-rack equivalents averaged 46 percent and 13 percent respectively.
Three-year Power Systems costs for deployment of Oracle E-Business Suite averaged 42 percent less than for use
of Exadata, while JD Edwards EnterpriseOne, PeopleSoft and Retail solutions averaged 27 percent, 21 percent
and 30 percent less respectively.
Key Differentiators
Lower system costs for IBM platforms than for Exadata reflect a number of factors. These include differences in
two key areas:
1. Technology. Power Systems employ a general-purpose UNIX server architecture. They incorporate a
variety of IBM innovations in cache, I/O, main memory, system-level performance optimization and other
areas. PowerVM virtualization, AIX system management, and Power Systems and AIX availability
features are recognized industry leaders.
In comparison, Exadata is a specialized appliance. It is promoted by Oracle for data warehousing,
transaction processing and database consolidation. The core design and most key features are, however,
optimized for high-volume, scan-intensive analytical workloads requiring high levels of read I/O
performance and throughput.
The characteristics of transactional workloads are, however, significantly different. Transactions involve
frequent reads and writes to a few rows of data at a time, rather than sequential scans of large volumes.
Databases are also typically smaller than for analytical systems, and I/Os involve smaller blocks of data.
2,861.0
1,517.8
844.2
2,495.7
826.5
495.7
Half
Quarter
Eighth
O
r
a
c
l
e

E
x
a
d
a
t
a

R
a
c
k

S
i
z
e

IBM Power Systems & disk arrays
Oracle Exadata Database Machine
$ thousands $ thousands
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 4
Although Oracle has boosted Exadata transactional performance with its Write Back Flash Cache feature,
this is an overlay on an architecture designed for a different purpose. Most distinctive Exadata features
enable levels of I/Os per second (IOPS) and bandwidth that exceed by wide margins the requirements
of all but the largest transactional users.
As a transaction processing system, Exadata capabilities are not significantly better than those of general-
purpose servers. Its costs, however, are a great deal higher. This is also the case for database
consolidation applications.
A further difference is that Exadata does not employ virtualization. It supports multiple application and
workloads through conventional timesharing techniques. Management mechanisms are database-centric
and, in comparison with those of Power Systems, rudimentary.
In comparison, PowerVM for Power Systems is one of the industrys most sophisticated server
virtualization architectures. Management mechanisms are tightly integrated, and are among the most
granular and effective in existence. These and other technology differences are discussed in more detail in
the following section.
2. Packaging and pricing. Although Exadata can in principle be configured with 20 or more racks, in
practice most sales have been of eighth- and quarter-rack models. In the new X4 series, these are
configured with 24 and 48 Intel database server cores respectively. Larger half- and full-rack systems are
configured with 96 and 192.
In comparison, Power Systems offer more granular increments. Calculations presented in this report, for
example, include use of 4-, 8-, 12-, 16-, 24-, 32-, 48- and 64-core models.
Lack of configuration granularity affects Exadata costs not only for the system itself, but also for the
enterprise software suite including Oracle Database, RAC and other tools which are licensed based on
numbers of database server cores. Oracles move from four-to-six core database processors has tended to
increase costs. The entry-level X4-2 eighth-rack model, for example, has 24 cores. Its predecessor had 16.
The effects of higher Power Systems granularity are particularly apparent in smaller installations, where
many users surveyed for this report employed four- to eight-core models. Higher granularity also,
however, contributed to substantial cost differences in larger deployments.
In terms of storage, use of disk arrays with Power Systems also offers greater configuration flexibility.
Storage tiers and media may be varied to meet the needs of individual users. Flash memory, in the form of
SSDs, can also be employed in Power Systems and in IBM as well as third-party arrays attached to these.
There are also significant variations between Exadata systems, and Power Systems and disk arrays in third-party
support, availability optimization and other areas. Principal differences are summarized in figure 4.
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 5

Oracle Exadata Database Machine IBM Power Systems & Disk Arrays
Core design Specialized hardware & software appliance.
Designed for high-volume, scan-intensive
analytical workloads requiring high levels of
read I/O performance & throughput.
General-purpose UNIX server & disk array
designs. Can be configured & optimized for
multiple types of application.
Leader in industry benchmarks measuring
transactional & data warehouse performance.
Application &
workload sharing
Does not support use of virtual machines.
Application & workload sharing is in timeshared
mode using Oracle Database Resource
Manager & I/O Resource Manager.
Acts as database server only. Additional
application servers may be required.
Advanced server virtualization architecture.
Supports multiple types of hardware- &
software-based partitions, extensive resource
sharing & integrated mixed workload
management.
May act as database & application server.
Virtual machines allow both to be hosted on the
same physical servers.
Also supported by IBM SAN Volume Controller
(SVC) solution for cross-platform, multivendor
storage virtualization.
Configurability Fixed ratio configurations of processor units,
memory, flash devices, & high-performance
SAS & Nearline SAS (NL-SAS) drives in
system units.
Offered in eighth-, quarter-, half- & full-rack
models. Additional Storage Expansion Racks
may be employed.
Server & disk array units separately
configurable with high levels of granularity.
Servers support IBM & third-party arrays with
wide range of SSDs & HDDs.
Availability
Optimization
Standard hardware reliability, availability &
serviceability (RAS) features; RAID 10 with
double or triple disk mirroring.
Supports Oracle Database, Data Guard, Active
Data Guard & Maximum Availability
Architecture.
Advanced hardware RAS & operating system-
level availability optimization features. Servers
& disk arrays support multiple RAID levels.
Supports same Oracle availability solutions as
Exadata. IBM & third-party solutions for high
availability (HA) clustering, failover & recovery
may also be employed.
Database & operating
system support
Oracle Database 11g R2 & 12c.
Oracle Linux & Solaris.
Oracle supported Databases: 8i, 9i, 10g, 11g &
12c. IBM DB2 & Informix, Sybase, others.
IBM AIX & IBM i, RHEL & SLES.
Figure 4: Principal Differences Between Oracle Exadata Database Machine and
IBM Power Systems with Disk Arrays
Power Systems and disk arrays offer not only a more cost-effective, but also a more flexible alternative to Exadata
for transaction processing. Systems may be configured in a more granular manner, using non-IBM hardware and
software. Servers may host non-Oracle software. Application and database serving may be handled by same
physical platform. Disk arrays may be shared by multiple servers and applications.
Power Systems also support a broader range of Oracle database versions than Exadata users are not required to
migrate to Database 11g R2 or 12c and operating systems, including industry standard Red Hat Enterprise
Linux (RHEL) and SUSE Linux Enterprise Server (SLES) distributions.
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 6
Technology Differences
Core Design: Oracle Exadata Database Machine
Oracle Exadata originated as a response to high-performance analytical appliances offered by Netezza and others,
which became popular in the early and mid-1990s. The system was introduced in 2008.
The core Exadata design combines a conventional Oracle database and RAC architecture with a separate
subsystem that offloads I/O processing. Oracle databases and RAC are implemented in Database Servers, and I/O
processing in Storage Servers. Software overlays integrate these components, and an InfiniBand fabric enables
high-speed communications between them.
Figure 5 illustrates this design.
Figure 5: Core Oracle Exadata Database Machine Design
The original design objective was to maintain Oracle database and RAC compatibility while addressing a key
constraint of conventional Oracle database architecture limited I/O throughput and offering appliance-
competitive performance for scan-intensive analytical workloads.
The principal Oracle Exadata technologies, including Hybrid Columnar Compression (HCC), Smart Scan, Storage
Indexes and Smart Flash Cache are designed to minimize traffic over the InfiniBand fabric.
HCC is an Oracle implementation of columnar technology that has become widely adopted for high-volume
analytics applications. Columnar technology offers significantly higher levels of raw throughput and compression
than row-based data structures when scanning large data sets. It is, however, a great deal less efficient in dealing
with specific records.
Columnar data structures have rarely been employed for transaction processing. Oracle proposes that, in an ERP
environment, HCC may be employed to accelerate reporting processes.
Database Servers
Oracle Database 11g/12c RAC Cluster
Intel E5 6-core (X4-2) or E7 10-core (X3-8)
Storage Servers
Intel E5 6-core
NL-SAS HP SAS Flash
InfiniBand
Hybrid Columnar Compression
Smart Scan/Storage Indexes
Smart Flash Cache
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 7
A further characteristic of the Oracle Exadata design should be highlighted. Hybridization results in an inefficient
design. High levels of overhead are generated. As a result, the considerable processing power offered by Oracle
Exadata systems does not translate directly into application-level performance.
Oracle moved to boost Exadata transactional performance with the introduction of Write Back Flash Cache in
October 2012. This feature accelerates processing of write operations that are characteristic of transactional
workloads. Writes as well as reads may now be automatically written to Smart Flash Cache in the original
Exadata design, writes were written to HDDs.
Oracle claims that use of Write Back Flash Cache can accelerate write I/Os per second by up to 20 times on
Exadata systems. User experiences have been more varied. Two organizations that contributed to this report
experienced 2 to 3 times, and 5 to 10 times acceleration. Actual gains are application-dependent. Significant
performance improvements may occur with exceptionally write-intensive workloads (e.g., large batch runs).
Write Back Flash Cache is an overlay on an architecture designed for high-performance query processing. Even
when it is employed, Exadata continues to handle transactional workloads in an inefficient manner. Large
segments of Exadata functionality and cost provide, at best, limited value in most transaction processing
environments.
Core Design: Power Systems
System Architecture
Power Systems originated in 1990 as a general-purpose UNIX server design, although it has been enhanced a
great deal since that time. Key components including Reduced Instruction Set Computing (RISC) processors,
partitioning, multithreading and others have become progressively more sophisticated.
System-level performance potential has been optimized at all levels of design and implementation. Key
capabilities include exceptionally effective compiler- and operating system-level performance acceleration,
including chip symmetric multithreading; low levels of symmetric multiprocessing (SMP) overhead; and
extensive system-level integration and optimization of performance-related features.
Power has evolved from a CPU-intensive design to one designed to deliver what IBM describes as balanced
performance. The focus has shifted toward more efficiently handling mixed, fluctuating workloads such as those
that characterize most modern-day transactional systems.
Part of this transition has involved close integration of PowerVM with AIX system and workload management.
The implications are important. In a mixed workload environment, the amount of work that a server can perform
over time depends not only on processor, memory and I/O throughput, but also on the mechanisms that allocate
and reallocate system resources as demands change.
Mixed workload management capabilities have been reinforced by new features in POWER7-generation systems.
These include the following:
Intelligent Threading allows workloads to be executed using one, two or four threads per core. The
system can automatically determine which to use for optimum performance, or system administrators may
select the number of threads employed.
Intelligent Cache allows systems to dynamically vary cache utilization as workload characteristics
change. The system may automatically determine the appropriate level of cache for specific workloads.
Continuous performance optimization is provided for both features.
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 8
One result is that, over time, Power Systems may realize significantly higher levels of capacity utilization than
less well-optimized platforms. Experiences with ERP systems have, for example, shown that Power Systems may,
over periods of months to years, execute workload volumes that are up to 40 percent larger are than indicated by
point-in-time measurements of performance.
Key components of overall Power Systems, PowerVM and AIX architecture are illustrated in figure 6. This
architecture allow users to manipulate a wider range of variables including subsystems, threads, processors,
cache, main memory and I/O, multiple types of partition, multiple threads and dedicated and pooled processors
with higher levels of granularity and flexibility than any competitive platform.
Figure 6: Combined Power Systems, PowerVM and AIX Architecture
This presentation shows use of hardware-based Logical Partitions (LPARs) and software-based micro-partitions,
which run inside LPARs. These may be configured in increments as small as 1/100
th
of a core and 1/100
th
of a
core respectively.
In addition, Virtual I/O Servers (VIOSs) allow instances running in multiple LPARs to share a common pool of
I/O devices. VIOSs may be duplexed for redundancy.
An extension, Shared Storage Pools, allows VIOSs to form a cluster across one or more Power Systems. External
storage capacity may be dynamically allocated and reallocated between systems as illustrated in figure 7. Thin
provisioning i.e., the ability to allocate space dynamically as data is written to disk, rather than in preset
volumes is supported.
!"#$%&' ")* +,#!,# !"#$%&' ")* +,#!,#
Physical Processors
Shared
Processor Pool



Virtual processors
Dedicated
Processors




Physical processors
Shared
Processor Pool



Virtual processors
AIX 7
System & Workload Management
Intelligent Threads Intelligent Cache
POWERVM
"#$%
"#$%
"#$%
"#$%
LPAR
Micro-partitions

LPAR
Micro-partitions

Virtual LAN

International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 9
Figure 7: PowerVM Storage Pool Example
Another feature, Active Memory Expansion, enables system-managed compression and decompression of data in
memory. Compression rates of up to 50 percent are supported; i.e., useable main memory may be up to double
physical memory.
Disk Arrays
Power Systems are typically configured with internal and/or external controller-based disk arrays. This approach
allows organizations to share external storage resources between multiple servers, and offers a number of other
potential benefits. These include the following:
RAID technology allows Power Systems to provide enterprise-class availability while enabling more
efficient use of disk resources. IBM offers a no-charge online Disk Magic service enabling customers to
determine optimum RAID configurations to meet their performance targets.
Exadata systems employ double mirroring (one active to one redundant) or triple mirroring (one active to
two redundant), depending on model.
Established IBM and third-party tools may be leveraged for such functions as snapshot copying, and real-
time replication for disaster recovery. IBM tools include FlashCopy Manager, as well as Metro Mirror
and Global Mirror for synchronous and asynchronous real-time replication respectively.
IBM Real-time Compression may be employed to reduce storage capacity by up to 80 percent, with
corresponding reductions in storage system, software and administration costs. Time and bandwidth for
such processes as backup and replication may also be significantly reduced.
Real-time Compression is a standard feature of IBM Storwize V7000 midrange arrays, and is supported
for a wide range of other IBM and non-IBM systems. For Storwize V7000, IBM guarantees customer
compression savings of 50 percent or greater. The company will provide hardware and software
authorizations to match any shortfall at no additional cost.
IBM SAN Volume Controller (SVC), the companys strategic cross-platform storage virtualization solution, also
supports Power Systems and most IBM and third-party disk arrays employed with these. SVC enables
organizations to improve capacity utilization; reduce storage costs; and implement consistent replication,
compression and tiering policies across large bases of multivendor servers and disk arrays.
Power
System
VIOS
Power
System
VIOS
Power
System
VIOS
Storage Pool
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 10
Flash Storage
Oracle has placed a strong marketing emphasis on its use of flash memory in the form of Flash Cache in the
companys latest Exadata models.
For example, in launching the new X4-2 and upgraded X3-8 in December 2013, Oracle doubled flash cache
capacity for all models by substituting 800 gigabyte (GB) flash for the 400 GB devices employed in earlier X-3
models. A new Flash Cache Compression function said to offer 2:1 compression was also introduced.
A full-rack Exadata system can now in principle be configured with up to 44 terabyte (TB) of physical or 88 TB
of compressed Flash Cache capacity. This calculation assumes a 2:1 compression level. As there is still little user
production experience with this capability, it is unclear whether this would occur in practice.
The announcement also stated this capacity is sufficient to hold the vast majority of OLTP databases entirely in
flash memory. Exadata performance is said to run to hundreds of thousands to millions of IOPS. However, it does
not necessarily translate into competitive differentiation against Power Systems.
Two qualifications should be made. First, Exadata Flash Cache capacities and IOPS levels far exceed the
requirements of most transactional users. To date, such requirements have been characteristic only of very large
user organizations and IT services companies.
Second, flash memory may also be employed to accelerate performance on Power Systems, and on hybrid and all-
flash arrays attached to these. In October 2013, for example, IBM introduced new 387 GB and 775 GB flash
drives for Power Systems.
IBM has reengineered the TMS product line with new performance and reliability, availability and serviceability
(RAS) features. The new IBM FlashSystem line currently includes the 840, introduced in December 2013, which
is designed for high-end (Tier 1) storage applications, the midrange 820 and the entry-level 720.
All have been deployed by Oracle Database and applications customers. Major improvements in online as well as
batch transaction processing performance have been reported.
FlashSystems may be used with SVC, enabling integration with broader virtualized storage environments and
supporting IBM Real-time Compression. A further benefit is that, through SVC, a wide range of mature IBM and
non-IBM tools may be employed for high-volume snapshot copying, remote replication, storage management and
other functions. Few all-flash arrays provide even a fraction of such enterprise-class capabilities.
Power Systems also support most new types of all-flash array offered by a growing number of vendors, including
IBM itself, for exceptionally I/O-intensive workloads. The current IBM FlashSystem line employs technology
from Texas Memory Systems (TMS), which IBM acquired in 2012.
The latest-generation FlashSystem 840, introduced in January 2014, supports up to 41 TB or 49 TB of usable
flash capacity per frame in RAID 5 and RAID 0 configurations respectively. IBM Real Time Compression
enables compression levels of up to 5:1; i.e., effective capacity per frame may exceed 200 TB. Multi-frame
configurations may scale into the multiple-petabyte range.
FlashSystem 840 models may be equipped with IBM SVC, enabling integration into virtualized enterprise
environments. The company has also focused on providing enterprise-class availability, recoverability and
security. These capabilities will become increasingly critical as all-flash arrays move into the IT mainstream.
In April 2013, IBM announced plans to invest $1 billion through 2015 in development of and support for flash
storage technologies. This total does not include TMS acquisition costs.
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 11
Availability Optimization
For more than a decade, Power Systems have been generally recognized to deliver higher levels of uptime
including avoidance of unplanned and planned outages than any competitive platform. This has been the case
for standalone as well as clustered configurations.
Industry surveys have consistently shown that Power Systems and AIX deliver higher levels of availability than
competitive platforms.
A recent survey of 550 organizations worldwide comparing availability for 14 server platforms found, for
example, that IBMs AIX 7.1 on Power Systems averaged around 10 minutes downtime per server per year, and
recorded the least amount of overall downtime for the best reliability among survey respondents.
*

Multiple factors have contributed to this reputation. The system includes some of the industrys most advanced
RAS microelectronics. The AIX operating system incorporates a wide range of standard UNIX and IBM-
developed features to minimize both the frequency and duration of outages.
A number of availability features are mainframe-derived. According to IBM, the companys Power and System z
(mainframe) design teams jointly developed the availability optimization features of POWER7-generation
systems. Mainframes enjoy the highest levels of uptime of any major platform.
Consolidation and Virtualization
Since its introduction, the Exadata platform has been promoted by Oracle for database server consolidation, and
many Oracle users have adopted it in this role. Power Systems and AIX have also been widely deployed to
consolidate Oracle as well as non-Oracle applications for more than a decade.
A key difference has been that Exadata systems are designed for database serving, and consolidation exercises
have typically required additional farms of x86-based application servers. PowerVM virtualization allows
database and application serving to be handled on the same physical platform.
Exadata does not employ virtualization in the normal sense of this term; i.e., the ability to host multiple database
and/or application instances in hardware or software-defined partitions on the same physical server. Neither
Oracle VM nor third-party virtualization solutions are supported on this platform. It functions as a timeshared
system. In comparison, PowerVM represents the industrys most sophisticated server virtualization architecture.
On Exadata systems, allocation of system resources is handled by the Database Resource Manager, a standard
feature of Oracle Database 11g and 12c, along with the Exadata-specific I/O Resource Manager (IORM). IORM
manages access to I/O resources provided by Storage Servers based on user-defined application priorities and
service-level targets.
Although these capabilities have proved popular among organizations employing Exadata systems for server
consolidation, they do not represent a significant competitive differentiator. Database Resource Manager also runs
on AIX, while allocation of I/O resources to meet service-level targets is a standard UNIX system management
function. AIX provides comparable capabilities.
It remains to be seen how much market traction the Exadata consolidation approach will gain. The Resource
Manager has not been widely employed for server consolidation on other platforms, and most organizations
continue to employ UNIX- or x86-based virtualization tools for Oracle consolidation projects.
Oracle also appears to be moving toward the new Multitenant feature in Database 12c as its main consolidation
focus. Although there is still little experience with Database 12c it became generally available only in June 2013
potential use of pluggable databases has appealed to many Oracle users. Database 12c, including Multitenant
capability, is fully supported on AIX.

!
*ITIC 2013 Global Server Hardware and Server OS Reliability Survey, February 2013
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 12
Conclusions
High-performance appliances such as Oracle Exadata have been deployed in analytical applications, and new
technologies such as columnar data processing have provided value for these. But the cost/benefit equation for
transaction processing solutions is different.
General-purpose UNIX server and disk array architectures remain competitive for transactional applications and
workloads. This is particularly the case for IBM Power Systems, which remain by wide margins the dominant
platform in this space. In performance and cost-effectiveness, as well as in availability and other areas, Power
Systems remain industry leaders.
Most transactional users are familiar with general-purpose UNIX server architectures. They enable integration of
technologies such as flash memory and storage tiering in a comparatively simple and non-disruptive manner, and
offer greater configuration flexibility. They also allow organizations to support non-Oracle as well as Oracle
software, and to employ non-IBM hardware and software as part of broader multivendor environments.
In comparison, Exadata is a more complex and proprietary platform requiring less familiar skill sets. It is unclear
what value it might provide to transactional users to offset these characteristics. In terms of capability and cost for
transaction processing, Oracle Exadata does not enjoy an edge over Power Systems. Platform selections should be
made accordingly.
International Technology Group June 2014
Comparing IBM Power Systems to Oracle Exadata Database Machine 13
International Technology Group
ITG sharpens your awareness of whats 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 on investment
(ROI) processes and methodologies. In 2004, the firm received a Decade of Education Award from 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.
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 to respond 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.
Clients include a cross section of IT end users in the private and public sectors representing multinational
corporations, 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).




























Copyright 2014 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 been obtained from sources assumed to be reliable and reflects conclusions at the time. Although the document
may utilize publicly available material from various sources, it does not necessarily reflect the positions of such 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 the material contained in this document or for interpretations thereof. Trademarks included in this document are the property
of their respective owners.
ORW03036 -USEN-00

Anda mungkin juga menyukai