p012645 - Reducing IT Spending with IBM Power Systems

Pascal Taverne
Senior IT Specialist & TCO Consultant, IBM CPO Eagle Team
IBM

© Copyright IBM Corporation 2016. Technical University/Symposia materials may not be reproduced in whole or in part without the prior written permission of IBM.
 Session objectives

• Introduce the IBM CPO Eagle Team and IT Economics Concepts to

help with competitive plays.

• Learn how IBM POWER8 helps businesses consolidating their

workloads on IBM Power Systems to reduce IT costs associated with
the proliferation of x86 servers to support infrastructure and business
critical workloads.

• The presentation will share financial and technical data from actual
business cases developed for clients to quantify their IT costs of doing
business on x86 and Power Systems.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 1

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Session summary

• Eagle IT Economics Overview

• IBM Power Systems do have lots of technical advantages ....
• Core performance
• I/O & memory-bandwidth
• Enterprise Pools - Shared Procesor Pool (SPP)
• Mobile Ressources
• Capacity on Demand
• Virtualization and Consolidation
• Infrastructure Complexity
• Availability of Infrastructure (RAS, Security)
• Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 2

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 IBM Eagle Team - IT Economics Practice

• Who we are
• Part of the IBM Competitive Project Office (CPO)
• Specialized in examining economic differences between platforms in client environments
• Focused on identifying areas for efficiencies and cost reductions
• Provide no-charge studies

• Benefits of engaging the Eagle Team

• Worldwide experience from successfully helping hundreds of clients since 2007
• … most likely we have evaluated a similar scenario before
• Leverage research and benchmarks from the various CPO teams
• We use client figures (not our own)
• … through a transparent model
• … with agreed-to assumptions
• … and iterate as required
• Provide a business case from which a client can make a financially based IT decision

© Copyright IBM Corporation 2016. Technical University/Symposia materials 3

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO – Get the Complete Picture!

TCA Hardware $ $ $ $ $ Planning

Software $ $ $ $ $ Upgrades
People $ $ $ $ $ Migration
Network $ $ $ $ $ Growth
Storage $ $ $ $ $ Parallel Costs
Facilities $ $ $ $ $ Net Present Value

QoS – Availability, Reliability, Security and Scalability

Total Cost of Ownership is much more than Total Cost of Acquisition!

© Copyright IBM Corporation 2016. Technical University/Symposia materials 4

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 5

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
Performance Difference Between POWER8 and x86
Per core performance (2-socket systems) for last 3 generations of Power vs x86

IBM Power processors continue to increase the core performance each generation
and decrease the need for licenses of per-core software

x86 Power
4000

Per Core Performance

86% Increase
3000
16% Decrease
2000 POWER8
POWER7+ 24c / 2s
POWER7 16c / 2s 3.5 GHz
1000 16c / 2s 4.2 GHz
3.55 GHz
0
POWER7 POWER7+ POWER8

Power

© Copyright IBM Corporation 2016. Technical University/Symposia materials 6

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Performance comparison – POWER8 vs. x86 E7
IBM POWER8 core performance is 2x the x86 Xeon E7-8890 v3 core performance

x86 IBM POWER8 vs. x86

Core Performance
“Haswell” POWER E870
Ratio
Intel Xeon POWER8 P8 Util: 100%
E7-8890 v3 @ 4.1 GHz x86 Util: 100%

# Cores 144 80 Benchmark Utilization

SAP 2-Tier 58,626 79,750 2.4

SPECint_rate2006 5,630 6,320 2.0

SPECfp_rate2006 2,050 5,130

(72 cores)
2.2

• Results are based on best published per core results on Xeon E7-8890 v3 processor
• SAP results are based on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application. Results valid as of October 3,
2014. . IBM Power Enterprise System E870 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 8
processors / 80 cores / 640 threads, POWER8; 4.19GHz, 2048 GB memory, 79,750 SD benchmark users running AIX® 7.1 and DB2® 10.5, Certification #: 2014034 Result valid as of
October 3, 2014. Source: http://www.sap.com/benchmark. Fujitsu PRIMEQUEST 2800E2, 8 Processors / 144 Cores / 288 Threads, Intel Xeon Processor E7-8890 v3, 2.5 Ghz, 1024 GB
memory; 58,626 SD benchmark users, running Windows Server 2012 Standard Edition and DB2 10; Certification # 2015012. Source: http://www.sap.com/benchmark .
• SPECcpu2006 results are submitted as of 5/15/2015. For more information go to http://www.specbench.org/cpu2006/results/.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 7

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 8

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 How is POWER8 optimized for big workloads?

These design decisions result in

best performance for all types of
workloads such as:
Analytics, Big Data, Java, ERP,
OLTP, HPC, …

*: POWER8 compared to Haswell EX

© Copyright IBM Corporation 2016. Technical University/Symposia materials 9

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Oracle on Power vs x86 TCO Tool – Step 1 of 6

Core Performance SMT8

Servers Perfromance
# Core
Core Ratio

POWER8 HE w/ PowerVM 2 x1

POWER8 SO w/ PowerVM 2 x1

POWER7 w/ PowerVM 1,07 x1,86

x86 "Haswell" w /OracleVM 1 x2

x86 "Haswell" w/ VMware 1 x2

© Copyright IBM Corporation 2016. Technical University/Symposia materials 10

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 11

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Shared Processor Pool Example

• Practical case with Shared Processor Pool (SPP)

• Non production environment : test, development, qualification, ...
• Asses the average workload and peaks and define a shared
Processor pool with the corresponding capacity to reduce the Oracle
licensing
CPU = 2.6 / VP = 4

CPU = 1.7 / VP = 4
= 16 processor licenses
CPU = 3.1 / VP = 4

CPU = 2.8 / VP = 4

The CPU monitoring shows a maximum cumulated workload for 6.3

èDefine a Shared Processor Pool with 7 CPUs

CPU = 2.6 / VP = 4

CPU = 1.7 / VP = 4

CPU = 3.1 / VP = 4 = 7 processor licenses

CPU = 2.8 / VP = 4

Virtual Shared Processor Pool with 7 processors

© Copyright IBM Corporation 2016. Technical University/Symposia materials 12

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Shared Processor Pool – micropartitioning and SW licensing

• Sum the ENTITLED CAPACITY of each capped LPAR in the pool

• Sum the ONLINE VIRTUAL CPUs for each uncapped LPAR in the pool
• Compare the total of capped and uncapped CPUs vs the Shared Pool
size. Use the lowest number for licensing
• LPAR 1 = capped, entitled capacity =1.6
• LPAR 2 = uncapped, Virtual CPUs =2
• Total = LPAR 1 (1.6) + LPAR2 (2) = 3.6 = round up to 4
• Pool size = 6
• Number of cores to license = 4 (lower than pool size of 6 cores)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 13

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 14

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 CoD pricing is economical and flexible

• CoD processors and memory priced to make it economically easier to

physically have “extra/standby” resources in place for easy permanent
or temporary activation

Power Capacity
on Demand
Systems
Physical E880
Price has two parts – E870
physical and activation 795 E850
Activation 780
770

No x86 server has Capacity on Demand

Specific ratio of Activation price to Physical resource price can vary depending on the model and specific
processor or memory feature. In general however, the majority of the total price is in the activation.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 15

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power8 Enterprise Pools

• Power Enterprise Pools

• Typical challenges:
• Planned Maintenance
• Rebalancing workloads
• Failover Clusters
• Passive Disaster Recovery
• Power8 Migration
• With EP, you move processor and memory activations within a defined pool of systems.

A B C A B C

Before After

© Copyright IBM Corporation 2016. Technical University/Symposia materials 16

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Leveraging Enterprise Pools
Mobile processor and memory activations may be re-allocated to any
system within a defined pool

20% buffer – activated on

request – configuration limit.
Exxx
BUFFER Capacity on demand:
activated on request – pay as
COD
Price you go.
Point MOBILE Activated capacity which can
STATIC move throughout the
Critical and less critical CPU RAM enterprise pool.
workloads Capacity activated and
• Production & Staging persistent to the server.
• Quality Assurance and Rules:
Development - Min 8 Active cores
- 50% of configured RAM is
active
- 25% of active RAM is
STATIC.
Traditional Capacity Model EP Capacity Model
Buffer Buffer Buffer Buffer
COD COD COD COD Mobile capacity
Price recovered from
Point failed server in room
STATIC from STATIC from MOBILE MOBILE
a.
Pre-allocated and
activated resources
Room B Room A License core from Room B from Room A

to cover HA. savingPrice

STATIC Point
MOBILE MOBILE MOBILE
Protected -

EP
Critical CRITICAL CRITICAL CRITIC
CRITICAL CRITIC
CRITICAL
Workload Non Non AL
Non AL
Non
CRITICAL CRITICAL CRITICAL CRITICAL
ROOM A ROOM B ROOM A ROOM B

Enterprise Pool with Mobile processor and memory can provide huge saving mainly
for HA & DR environment

© Copyright IBM Corporation 2016. Technical University/Symposia materials 17

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 18

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power enterprise pools with mobile and elastic CoD deliver both
flexibility and economic efficiency

• Multi-system infrastructure providing a highly available

and flexible IT environment
E880
E870

• Clients purchase the capacity needed and allocate & 780

795

rebalance virtual processor and memory resources 770

• Live Partition Mobility enables clients to migrate running

workloads between systems to enable continuous • Planned maintenance
availability • Rebalance capacity
• Failover clusters
• Server Migrations
Active - Active Planned Maintenance Resource Balancing Periodic Demand Peaks

32
System being Activate Inactive
64 64 96 64 64
Maintained 64c for
Inactive Inactive Inactive Inactive Inactive
Maint.

32c
Active

64c 64c 64c Active 64c 64c

Active Active 64c 32c Active Active
Active Active

No Intel x86 server has this quality of service

© Copyright IBM Corporation 2016. Technical University/Symposia materials 19

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Oracle on Power vs x86 TCO Tool – Step 2 of 6

Core Performance SMT8 Consolidation

Virtual
Servers Perfromance Estim. Cumul
# Core Shared Proc
Core Ratio # Core
Pool Impact
Roundup of
POWER8 HE w/ PowerVM 2 x1 x1 x1
Sum
Roundup of
POWER8 SO w/ PowerVM 2 x1 x1 x1
Sum
Roundup of
POWER7 w/ PowerVM 1,07 x1,86 x1 x1,86
Sum
Roundup of
x86 "Haswell" w /OracleVM 1 x2 x1 x2
Sum
Sum of
x86 "Haswell" w/ VMware 1 x2 x1.2 x2,4
Roundup

Roundup of Sums versus the Sum of Roundups. Vendors such as Oracle force customers to
roundup licensing for each partition of Oracle Database before they add the licenses together
of all the partitions within a server. These roundups can add to a tremendous overhead of
“lost” licensing. By using IBM Power’s Shared Processor Pools, only the total pool has to be
rounded up and then licensed within a server

© Copyright IBM Corporation 2016. Technical University/Symposia materials 20

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 21

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Statistical Multiplexing Example: The More Workloads you can
Consolidate, the Smaller is the Variance of the Sum

One Workload in one LPAR 16 Workloads in 16 LPARs 144 Workload in 144 LPARs
Server utilization = 17% Server Server utilization = 70%
Capacity Server utilization = 44%
Require
d (60)
Server
6x Peak To Average Capacity
Required Server
(364) 1.42x Peak To Average Capacity
2.25x Peak To Average Required
Average Average (2057)
Demand Average
Demand
(10) Demand
(160)
(1440)

Individual workloads Sum of both workloads

Variance in
Variance in demand
demand

n Workload demands peak at different times

n Peaks & valleys in demand tend to cancel each other out
n Statistical multiplexing results in less variance of total demand

© Copyright IBM Corporation 2016. Technical University/Symposia materials 22

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Performance comparison – POWER8 vs. x86 E7
IBM POWER8 core performance is 2x the x86 Xeon E7-8890 v3 core performance

• ALL data is PUBLISHED or SUBMITTED

x86 IBM POWER8 vs. x86
“Haswell” POWER E870 Core Performance Ratio
Intel Xeon POWER8 P8 Util: 100% P8 Util: 80% P8 Util: 80%
E7-8890 v3 @ 4.1 GHz x86 Util: 100% x86 Util: 40% x86 Util: 20%
Benchmark Utilization with Utilization without
# Cores 144 80 Utilization virtualized x86 virtualized x86

SAP 2-Tier 58,626 79,750 2.4 4.8 9.7

SPECint_rate2006 5,630 6,320 2.0 4.0 8.0

SPECfp_rate2006 2,050 5,130

(72 cores)
2.2 4.5 9.1

• Results are based on best published per core results on Xeon E7-8890 v3 processor
• SAP results are based on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application. Results valid as of October 3, 2014. . IBM Power
Enterprise System E870 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 8 processors / 80 cores / 640 threads, POWER8;
4.19GHz, 2048 GB memory, 79,750 SD benchmark users running AIX® 7.1 and DB2® 10.5, Certification #: 2014034 Result valid as of October 3, 2014. Source: http://www.sap.com/benchmark. Fujitsu
PRIMEQUEST 2800E2, 8 Processors / 144 Cores / 288 Threads, Intel Xeon Processor E7-8890 v3, 2.5 Ghz, 1024 GB memory; 58,626 SD benchmark users, running Windows Server 2012 Standard Edition
and DB2 10; Certification # 2015012. Source: http://www.sap.com/benchmark .
• SPECcpu2006 results are submitted as of 5/15/2015. For more information go to http://www.specbench.org/cpu2006/results/.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 23

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 POWER8 Performance Utilization Guarantee

System Performance Utilization Scale Out System Performance Utilization Enterprise

IBM has designed these POWER8 systems to operate at IBM has designed these POWER8 systems to operate at
industry-leading levels of efficiency, guaranteeing the industry-leading levels of efficiency, guaranteeing the system
system will perform as warranted while at a sustained will perform as warranted while at a sustained 80% utilization
65% utilization – a rate higher than common competitive for E870 and E880 and at a sustained 70% utilization for E850
platform utilization levels – rates higher than common competitive platform utilization
levels

System Performance Utilization Guarantee

When a Client acquires a POWER8 Scale Out Server and the Client runs eligible workloads,
IBM guarantees the system will perform as warranted with a System Utilization Rate of up to
65%. Should the Client not be able to achieve 65% system utilization rate, assuming there is
sufficient work to drive the machine to 65% utilization, IBM will assist with the attainment
of 65% system utilization rate, at no additional cost.

System Performance Utilization Guarantee

When a Client acquires a POWER8 E850, E870 or E880 Enterprise Server and the Client
runs eligible workloads, IBM guarantees the system will perform as warranted with a System
Utilization Rate of up to 80% (70% for E850). Should the Client not be able to achieve 80%
(or 70% for E850) system utilization rate, assuming there is sufficient work to drive the
machine to 80% utilization, IBM will assist with the attainment of 80% (or 70% for E850)
system utilization rate, at no additional cost.
The IBM POWER8 one or two-socket server (the “POWER8 Server”) must be purchased from IBM or an authorized IBM Business Partner prior to December 31, 2015, provided that the combined transaction price (not
including taxes and fees) of the POWER8 Server purchased with other IBM hardware, IBM software, IBM maintenance, and IBM services must be US $150,000 or greater. Eligible workloads are AIX, Linux, or IBM i workloads,
or any combination of the three, on PowerVM. Guarantee Period - 120 calendar days following the Date of Installation of the Eligible Machine. System Performance Guarantee Requirements - Client implements each of the
requirements set forth in the “POWER8 Performance Guarantee Requirements. Common x86 utilization levels assessed by 3rd party analysis.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 24

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Example:
IBM i and AIX Consolidation on IBM Power Systems
(POWER8)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 25

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study - Executive Overview

• The objective of this TCO Study is to demonstrate the cost savings

that can be achieved by consolidating Customer POWER 5/6 Systems
to the latest POWER8 IBM Power Systems, over a 5-year period.

• Customer Business Issue:

• Customer is running the core banking application on previous generation of IBM i (POWER6). They
are also running various POWER5 and POWER6 Systems with AIX. To improve the performance
and the scalability of their global IT solution but also to reduce the global IT cost (HW/SW
maintenance, power consumption, DP space, …) they would like to consolidate on POWER8
• Study Objectives:
• 5-year TCO Cost Comparison of their existing environment with new POWER8 solution
• Cases Evaluated:
• Case 1 – Existing POWER5/6 IBM Power Systems
• Case 2 – IBM Power Systems E870 – Enterprise Server (POWER8 4.02GHz)
• The findings:
• IBM Power System E870 offers the lowest 5 year TCO with EUR 3.5M substantial cost savings and
a potential EUR 0.87M saving in Annual Run Rate or Operating Cost compared to the existing
environment.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 26

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Comparisons

Case 1 – Baseline Case 2 – New POWER8 Servers

Existing Servers IBM Power Systems E870

Various POWER5/6 IBM Power Systems 2 x POWER8 3.42GHz (4ch/32co)
505 (x2), 510 (x3), 550 (x6), 570 (x2), 595 (x2)
15 servers 2 servers
140 cores 41 cores (22 for Prod / 19 for DR)
AIX, IBM i AIX, IBM i
PowerVM, PowerHA PowerVM, PowerHA
Oracle, WBI, DataStage Oracle, WBI, DataStage
WAS, MQ, Portal WAS, MQ, Portal
LPAR23 LPAR23
Other Non IBM Application (except Oracle) are out of Scope Other Non IBM Application (except Oracle) are out of
of the study comparison Scope of the study comparison

© Copyright IBM Corporation 2016. Technical University/Symposia materials 27

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Approach

• Study analysed Customer’s POWER5/6 Workloads

• 15x Various POWER5/6 IBM Power Systems with up to 58 LPARS

• Consolidation (IBM GHz Analysis Method) Calculations

• Calculations include:
• The Relative Performance (RPE2) of each server / processor
• A Chip Performance Factor based on the processor generation
• A Networking (Co-Location Effect) Factor
• A Utilization Factor (based on actual data from the nmon for AIX and Tango/04 for IBM i Analysis tools)

• TCO (5-Years) Cost Calculations

• Hardware, Software, People
• Networking Costs, Hosting, Energy Consumption (Consumption and Cooling)

• OS / MiddleWare / Oracle Software Stack

• AIX (5.3, 6.1, 7.1) , IBM i (7.1). Note that AIX 5.3 is not supported on POWER8 Systems LPAR
• PowerVM (Standard Edition) , PowerHA
• Oracle Database Enterprise Edition , IBM Websphere Business Integration for Financial Networks Base Server, DataStage
• Websphere ND, WebSphere MQ, Websphere Portal
• Tivoli Storage Manager

© Copyright IBM Corporation 2016. Technical University/Symposia materials 28

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Customer’s POWER5/6 Servers

Type Model Name Processor

9115 505 505 Power5+
9115 505 505 Power5+
9110 51A 510 Power5+
9110 51A 510 Power5+
9110 51A 510 Power5+
8204 E8A 550 Power6
8204 E8A 550 Power6
8204 E8A 550 Power6
9133 55A 550 Power5+
9133 55A 550 Power5+
9133 55A 550 Power5+
9406 MMA 570 Power6
9117 MMA 570 Power6
9119 FHA 595 Power6
9119 FHA 595 Power6

© Copyright IBM Corporation 2016. Technical University/Symposia materials 29

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Customer’s POWER5/6 LPARS

Server Server Type - Prod Site Role / Mem cores / Server Server Type - DR Site Role / Mem cores /
Location (GB) LPAR Location (GB) LPAR
Name Name
System p5 505Q Express (1U) Power5+ 1.65GHz 36MB System p5 505Q Express (1U) Power5+ 1.65GHz 36MB
LPAR1 LPAR26
(2ch/4co) Prod 16 4.0 (2ch/4co) DR 16 4.0
System p5 510Q Express (2U) Power5+ 1.65GHz 36MB
LPAR2 LPAR27
(2ch/4co) Test 16 4.0 Power 550 Express (4U) Power6 4.2GHz 32MB (2ch/4co) DR 15 1.0
System p5 510Q Express (2U) Power5+ 1.65GHz 36MB
LPAR3 LPAR28
(2ch/4co) Prod 6 2.0 Power 550 Express (4U) Power6 4.2GHz 32MB (2ch/4co) DR 15 3.0
System p5 510Q Express (2U) Power5+ 1.65GHz 36MB LPAR29 System i 570 (8U) Power6 4.7GHz 32MB (4ch/8co) DR 46 1.5
LPAR4
(2ch/4co) Prod 6 2.0
LPAR30 System i 570 (8U) Power6 4.7GHz 32MB (4ch/8co) DR 25 1.0
System p5 550 Express (4U) Power5+ 1.65GHz 36MB
LPAR10 LPAR31 System i 570 (8U) Power6 4.7GHz 32MB (4ch/8co) DR 10 1.0
(2ch/4co) Prod 16 4.0
System p5 550 Express (4U) Power5+ 1.65GHz 36MB LPAR32 System i 570 (8U) Power6 4.7GHz 32MB (4ch/8co) DR 3 0.5
LPAR11
(2ch/4co) Prod 16 4.0 LPAR33 System i 570 (8U) Power6 4.7GHz 32MB (4ch/8co) DR 54 3.0
System p5 550 Express (4U) Power5+ 1.65GHz 36MB
LPAR5 LPAR48 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 55 5.6
(2ch/4co) Prod 12 4.0
LPAR49 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 2 0.6
LPAR8
Power 550 Express (4U) Power6 4.2GHz 32MB (2ch/4co) Prod 15 1.0 LPAR50 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 6 0.3
LPAR9 LPAR51 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 6 1.2
Power 550 Express (4U) Power6 4.2GHz 32MB (2ch/4co) Prod 15 3.0
LPAR52 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 9 1.8
LPAR6
Power 550 Express (4U) Power6 4.2GHz 32MB (2ch/4co) Prod 15 1.0 LPAR53 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 30 2.8

LPAR7 LPAR54 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 2 0.5

Power 550 Express (4U) Power6 4.2GHz 32MB (2ch/4co) Prod 15 3.0
LPAR55 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 0 0.0
LPAR12 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 128 7.5
LPAR56 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 4 0.2
LPAR13 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 69 4.0
LPAR57 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 200 18.0
LPAR14 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 12 1.5
LPAR58 Power 595 Power6 4.2GHz 32MB (20ch/40co) DR 10 1.0
LPAR15 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 6 0.8
LPAR34 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 12.5 2.0
LPAR16 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 34 2.0
LPAR35 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 16 1.0
LPAR17 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 29 2.0
LPAR36 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 12.5 2.5
LPAR18 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 12 1.5
LPAR37 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 49 1.0
LPAR19 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 46 3.0
LPAR38 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 1.0
LPAR20 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 5 0.5
LPAR39 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 1.0
LPAR21 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 32 1.0
LPAR40 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 14 1.0
LPAR22 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 4 0.6 LPAR41 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 14 1.0
LPAR23 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 16 1.0 LPAR42 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 0.4
LPAR24 Power 595 Power6 4.2GHz 32MB (32ch/64co) Prod 243 20.0 LPAR43 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 0.4
LPAR25 Power 595 Power6 4.2GHz 32MB (32ch/64co) Test 23 1.5 LPAR44 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 0.4
LPAR45 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 0.4
LPAR46 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 0.4
LPAR47 Power 570 (16U) Power6 3.5GHz 32MB (8ch/16co) DR 10 0.4

© Copyright IBM Corporation 2016. Technical University/Symposia materials 30

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Existing Infrastructure

p505 p550 p550 p505 p550

LPAR26 LPAR27
LPAR1 LPAR5 LPAR10
Prod DR
LPAR28
p510 p550 p550 p570
LPAR2 LPAR6 LPAR11
LPAR29 p570

LPAR7 LPAR34
p510 LPAR30

LPAR35
LPAR3
p550 LPAR31
LPAR36

P510 LPAR8 LPAR32

LPAR37

LPAR4 LPAR9
LPAR33 LPAR38

p595 LPAR39
p595
LPAR40
LPAR12 LPAR19

LPAR48 LPAR54 LPAR41

LPAR13 LPAR20

LPAR49 LPAR55 LPAR42

LPAR14 LPAR21
LPAR43
LPAR50 LPAR56
LPAR15 LPAR22
LPAR44
LPAR51 LPAR57
LPAR16 LPAR23 LPAR45
LPAR52 LPAR58
LPAR17 LPAR24 LPAR46

LPAR53
LPAR18 LPAR25 LPAR47
IBM i AIX

© Copyright IBM Corporation 2016. Technical University/Symposia materials 31

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 New Infrastructure
POWER8 E870
POWER8 E870
LPAR48 LPAR54
Prod DR
LPAR1 LPAR5
LPAR49 LPAR55

LPAR2 LPAR6 LPAR50 LPAR56

LPAR51 LPAR34
LPAR3 LPAR7

LPAR52 LPAR35
LPAR8
LPAR4
LPAR53 LPAR36

LPAR9
LPAR10 LPAR29 LPAR37

LPAR11 LPAR30 LPAR38

LPAR12
LPAR31 LPAR39

LPAR19
LPAR13 LPAR32 LPAR40

LPAR20 LPAR26 LPAR41

LPAR14

LPAR27 LPAR42
LPAR21
LPAR15
LPAR28 LPAR43
LPAR22
LPAR16 LPAR57 LPAR44

LPAR23 LPAR45
LPAR58
LPAR17

LPAR24 LPAR33 LPAR46

LPAR18
LPAR25 LPAR47
IBM i AIX

© Copyright IBM Corporation 2016. Technical University/Symposia materials 32

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Results

53% Saving
EUR 3,574,679

EUR 6,714,912 EUR 3,140,233

© Copyright IBM Corporation 2016. Technical University/Symposia materials 33

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Results

6,714,912

3,140,233

New POWER8 IBM Power Systems deliver substantial cost savings (EUR 3.5M) compared to the
installed various POWER5/6 IBM Power Systems over a 5-year TCO period

© Copyright IBM Corporation 2016. Technical University/Symposia materials 34

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Results

EUR
0.87M

IBM Power Systems offer a potential (EUR 0.87M) saving in Annual Run Rate or Operating Cost

© Copyright IBM Corporation 2016. Technical University/Symposia materials 35

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study - Executive Overview

The objective of this TCO Study is to demonstrate the cost savings that could be
achieved by refreshing the actual IBM Power System 795 (POWER7) by a new IBM
Power System E880 (POWER8).
Study Objectives:
• 5-year TCO Cost Comparison
Cases Evaluated:
• Case 1 –Existing IBM Power 795 (POWER7) with Oracle DB
• Case 2 – Technical Refresh to of the existing p795 to a new IBM Power E880
(POWER8) with Oracle DB
• Case 3 – Technical Refresh to of the existing p795 to a new IBM Power E880
(POWER8) with IBM DB2
The findings:
• IBM POWER8 refresh with DB2 requires 66% fewer cores than the existing
POWER7 server – this translates into lower hardware maintenance cost and software
costs. Infrastructure cost (DP room, Electricity & cooling) are also lower
• IBM POWER8 E880 with DB2 offers the lowest 5 year TCO - $8.3M vs $12.2M for
POWER8 E880 with Oracle vs $34M for POWER7 795 with Oracle
• IBM POWER8 E880 with DB2 offers the lowest annual operating rate: $1.2M vs
$2.7M for POWER8 E880 with Oracle vs $7.6M for POWER7 795

© Copyright IBM Corporation 2016. Technical University/Symposia materials 36

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Comparisons

Case 1 – Baseline Case 2 – New Case 3 – New

Existing Server IBM Power Systems E880 IBM Power Systems E880
POWER 795 @ 4.0 Ghz POWER E880 @ 4.35 Ghz POWER E880 @ 4.35 Ghz
Active Cores: 96 of 128 Active Cores: 64 of 96 Active Cores: 64 of 96
Active Memory: 1536 GB Active Memory: 2048 GB Active Memory: 2048 GB
2 servers (1 for Prod + 1 for DR) 2 servers (1 for Prod + 1 for DR) 2 servers (1 for Prod + 1 for DR)
192 cores (96 x2) 72 Cores (64 + 8) 72 Cores (64 + 8)
IBM AIX IBM AIX IBM AIX
Oracle DB Oracle DB IBM DB2
Oracle Workload Sizing Oracle Workload Sizing DB2 Workload Sizing
162 (81 + 81) cores 102 (51 + 51) cores 42 (41 + 1) cores

© Copyright IBM Corporation 2016. Technical University/Symposia materials 37

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 POWER7 Servers/Lpars

Prod DC DR DC

840F627 / p595 840F617 / p595

FRAM_1 FRAM_2 FRAM_1 FRAM_2

FRAM2-APP- HANDSET-DB- FRAM2-APP- HANDSET-DB-

SVR SVR SVR SVR

LGR-PROD- LGR-PROD-
LAB LAB
DB-2 DB-2

NIM PRAC NIM PRAC

PROD-ORA- PROD-ORA- PROD-ORA- PROD-ORA-

DB-SVR-6 DB-SVR-7 DB-SVR-6 DB-SVR-7

PROD-ORA- PROD-ORA- PROD-ORA- PROD-ORA-

DB-SVR-8 DB-SVR-9 DB-SVR-8 DB-SVR-9

PROD-ORA- PROD-ORA- PROD-ORA- PROD-ORA-

DB-SVR4 DB-SVR5 DB-SVR4 DB-SVR5

PROD- PROD-
PROD1 104 PROD1 104
ORADV-SVR3 ORADV-SVR3

SURE-DB- SURE-DB-
PROD2 112 PROD2 112
SVR SVR

WEB WEB
VIOS VIOS
SERVICES SERVICES

WVALE- ZSMART-BI- WVALE- ZSMART-BI-

ORADB-SVR1 COGNOS ORADB-SVR1 COGNOS

ZSMART-SAN ZSMART-SAN

© Copyright IBM Corporation 2016. Technical University/Symposia materials 38

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power Enterprise Pool – Regular Mode

Pool Totals
192 Cores

Prod DR Activations:
96-core E880 96-core E880 72 Acitve
4.35 GHz 4.35 GHz
Activations: - 16 static
Activations:
64 Acitve 8 Acitve - 56 mobile
- 8 static - 8 static 120 “dark”
- 56 mobile - 0 mobile
32 “dark” 88 “dark”

32

Dark Dark
Mobile 88 Mobile
56
Static Static

8 8

© Copyright IBM Corporation 2016. Technical University/Symposia materials 39

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power Enterprise Pool - Disaster Mode

Pool Totals
96 Cores

Prod DR Activations:
96-core E880 96-core E880 64 Acitve
4.35 GHz 4.35 GHz
Activations: - 8 static
Activations:
64 Acitve 64 Acitve - 56 mobile
- 8 static - 8 static 32 “dark”
- 56 mobile - 56 mobile
32 “dark” 32 “dark”

32

Dark Dark
Mobile Mobile
56
Static Move 56 Mobiles
Static

© Copyright IBM Corporation 2016. Technical University/Symposia materials 40

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power Enterprise Pool – 20% Growth

Pool Totals
192 Cores

Prod DR Activations:
96-core E880 96-core E880 78 Acitve
4.35 GHz 4.35 GHz
Activations: Activations: - 16 static
70 Acitve 64 Acitve - 62 mobile
- 8 static - 8 static 114 “dark”
- 62 mobile - 0 mobile
26 “dark” 88 “dark”

26
Move 6 Mobiles

Dark Dark
Mobile Mobile
62 88
Static Static

8 8

© Copyright IBM Corporation 2016. Technical University/Symposia materials 41

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power Enterprise Pool – More on Activations

• Each server must have some static processor core activations.

• All systems must be ordered with at least 50% of the memory active. A max of
75% of the memory activations can be mobile.

Static activations Power E870/E880 POWER7+ 770 / 780 Power 795

Static core minimum 8 4 25% or 24 (which ever larger)
Static memory minimum 25% 25% 25%

• Mobile activations ordered against a specific E870, E880, 770, 780, 795
• But “kept” or “inventoried” in the pool’s HMC, not on that specific
server
• A server is unaware of mobile activations unless the HMC assigns the
activation to it
• Static activations kept on the specific server

static mobile

© Copyright IBM Corporation 2016. Technical University/Symposia materials 42

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Results
35,000$
5+Year)TCO)Comparison)
30,000$

25,000$
Thousands)

20,000$
64 % Saving 75 % Saving

15,000$

10,000$

5,000$

0$
AsIs$ $E880(Oracle$$ $E880(DB2$$
Migra5on$ ($$$$ ($$$$ 704,750$
DR$(All)$ 10,566,911$ 3,869,348$ 1,078,923$
Energy$ 310,204$ 114,646$ 104,619$
Hos5ng$ 15,490$ 8,095$ 8,095$
Networking$ 12,500$ 12,500$ 12,500$
People$ 1,800,000$ 1,800,000$ 1,800,000$
SoLware$ 10,002,164$ 3,794,192$ 2,348,124$
Sys$SoLware$ 4,228,056$ 784,680$ 674,856$
Hardware$ 7,065,997$ 1,773,862$ 1,614,677$

5-year TCO USD 34 001 323 USD 12 157 324 USD 8 346 545

© Copyright IBM Corporation 2016. Technical University/Symposia materials 43

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Results

Accumulated)TCO)Cost)Comparison)
35000"
USD 34 001 323
30000"
Thousands)

25000"

20000"

15000"
USD 12 157 324
10000"
USD 8 346 545
5000"

0"
Year"1" Year"2" Year"3" Year"4" Year"5"

"AsIs"" "E8801Oracle"" "E8801DB2""

IBM POWER8 System with DB2 delivers substantial cost savings (USD 25.6M) compared to the
POWER7 795 System over a 5-year TCO period

© Copyright IBM Corporation 2016. Technical University/Symposia materials 44

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 TCO Study Results

Annual)Run)Rate)TCO)Cost)Comparison)
!10,000.00!!
!9,000.00!!
!8,000.00!!
Thousands)

!7,000.00!!
!6,000.00!! USD 6.4M
!5,000.00!!
!4,000.00!!
!3,000.00!!
!2,000.00!!
!1,000.00!!
!"!!!!
Year!1! Year!2! Year!3! Year!4! Year!5!

!AsIs!! !E880"Oracle!! !E880"DB2!!

IBM POWER8 System with DB2 offers a potential (USD 6.4M) average saving in Annual Run Rate or
Operating Cost (4.9M Compare to POWER8 with Oracle)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 45

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Oracle on Power vs x86 TCO Tool – Step 3 of 6

Core Performance
Consolidation Virtualization
SMT8
Virtual
Servers Estim. Cumul Typical CPU Cumul
Perfromance # Shared
# utilization with # Core
Core Ratio Core Proc Pool
Core virtualization
Impact
POWER8 HE w/ Roundup 80% (IBM
2 x1 x1 x1 x1 x1
PowerVM of Sum Commit.)
POWER8 SO w/ Roundup 65% (IBM
2 x1 x1 x1 x1,23 x1,23
PowerVM of Sum Commit.)
POWER7 w/ Roundup 60% (IBM
1,07 x1,86 x1 x1,86 x1,23 x2,29
PowerVM of Sum Commit.)
x86 "Haswell" w Roundup
1 x2 x1 x2 40% x2 x4
/OracleVM of Sum
x86 "Haswell" w/ Sum of
1 x2 x1.2 x2,4 40% x2 x4,8
VMware Roundup

Try to ensure that each Processor CORE is running at close to 100% utilization as
possible
A good average is close to 80-90%
Large processor pools provide better protection against unexpected workloads

© Copyright IBM Corporation 2016. Technical University/Symposia materials 46

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 47

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Infrastructure Complexity Increase Staff costs

• Enterprise complexity
• Complexity increases with scale, staff turnover, Enterprise Availablity
clustering, changes, etc
• When complexity is no longer manageable,

Availablity
availability decreases
• Operational challenges increase with volume
(e.g. PD/PSI, security, performance management)
Complexity

• Strategies for reducing complexity

• Deploy common tooling and enterprise wide automation
• Standardize - utilize reference architectures
• Use larger virtualized servers

• Appropriate strategies may vary with the size of the environment

© Copyright IBM Corporation 2016. Technical University/Symposia materials 48

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Labor Cost

Labor Cost Example

Example:
• UCS C240 M4 (2U) Xeon E5-2670v03 12-Core 2.3GHz (2ch/24co)
• Power System E870 40 Cores (7U)
Based on Step 4 with only perfromance ratio :

Core Performance Consolidation

Virtualization
SMT8 (estimated)
TCA Server w / Oracle
Cumul Virtual Shared Cumul
Licence Core CPU utilization Estim. #
# Core # Core Proc Pool
Ratio with virtualization Core
Impact
80% (IBM Roundup of
Power8 w/ PowerVM 2,1 x1 x1 x1 x1 x1
Commit.) Sum
x86 "Haswell" w/ Roundup of
1 x2.1 40% x2 x4.1 x1 x4.1
OracleVM Sum
x86 "Haswell" w/ Sum of
1 x2.1 40% x2 x4.1 x1.2 x5
VMware Roundup

6 x 86 Xeon E5-2670v3 24 cores can be consolidated on 1 x E870 40 cores

Labor Model For Servers
• Annual Cost @ ~ 120.000 € per year (customer Hours per
Hours per
provided - fully burdened, blended rate) # of Physical Total # of HW or SW
Distributed Server - Labor Year for FTE Ratio %
Servers VMs Tasks (per
• FTE Hours per year @ ~ 1.670 hours Tasks
year)
• 52 weeks x 38 hrs / week = 1976 hours, minus
x86 24 72 36 2592
• 190 hours for vacation (5 weeks) Software Labor
Power 4 72 36 2592
• 120 hours for sick leave / paid holidays / training
x86 24 0 32 768
& administration Hardware Labor
Power 4 0 32 128
• Best Fit data indicates
x86 24 72 3 360
• Software Tasks are 36 hours per software image per Subtotal
year Power 4 72 2 720
• Assume this applies to all software images Available Hours per FTE 1 670
• Hardware Tasks are 32 hours per physical server # of FTEs x86 2,01
per year # of FTEs Power 1,63
19%
• Assume this applies to Intel and Unix Servers

© Copyright IBM Corporation 2016. Technical University/Symposia materials 49

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Oracle on Power vs x86 TCO Tool – Step 4 of 6

FTE Cost
TCA Server w / Oracle Licence
FTE
# FTE
Ratio

Power8 w/ PowerVM 0,8 x1

x86 "Haswell" w /OracleVM 1 X1,2
x86 "Haswell" w/ VMware 1 X1,2

© Copyright IBM Corporation 2016. Technical University/Symposia materials 50

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 51

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Power Availability of Infrastructure to avoid loss in revenue

• Reliability, Availability and Serviceability (RAS)

• IBM PowerHA® SystemMirror® software
• 99.997% application uptime
• 7x24 Enterprise customer operations Hardware redundancy
• Automated recovery
• Special error-handling technologies Reliable components
• Failure detection
• Monitoring
• IBM develops, tests, integrates the entire stack for RAS
• I/O drawers / memory management unit
• Processors and all chips in CEC
• Hypervisor (PowerVM) and VIOS
• Device drivers, PCI adapters
• Operating system (AIX, System i, Linux)
• Middleware and Clustering software

© Copyright IBM Corporation 2016. Technical University/Symposia materials 52

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 ITIC 2014 Survey –
POWER Has Highest Availability for 5th Year in a row

Planned Downtime - Hours Per month

Note: This survey included a
Power AIX 0.09 5.4 minutes/month variety of different system
environments (stand-alone
and clustered) over multiple
Intel RHEL 0.18 10.8 minutes/month platform generations

HP UX 0.39 At Least 2X Better

Availability than
Intel SUSE 0.44 26.4 minutes/month x86 Systems

Intel Ubuntu 0.55

Windows Server 2012 0.76 45.60 minutes/month

Oracle SPARC Solaris1.26 1.26

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Source:
2014 ITIC Server Reliability Survey: http://itic-corp.com

© Copyright IBM Corporation 2016. Technical University/Symposia materials 53

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Downtime costs

x86 x86 Power

Calculations
Windows Linux AIX
Average hours per month 0,76 0,18 0,09
Annual downtime hours per year 9,12 2,16 1,08
Average revenue loss per downtime hour (Insurance) $1 202 404 $1 202 404 $1 202 404
Total Annual Downtime Costs (annual revenue loss) $10 965 924 $2 597 193 $1 298 596

© Copyright IBM Corporation 2016. Technical University/Symposia materials 54

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Oracle on Power vs x86 TCO Tool – Step 5 of 6

Availability Cost
Server
Downtime
Downtime
Ratio

Power8 w/ AIX 0,11 X1

x86 "Haswell" w / Linux 0,5 X2
x86 "Haswell" w/ Window 1 X8,4

© Copyright IBM Corporation 2016. Technical University/Symposia materials 55

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Agenda

1. Eagle IT Economics Overview

2. IBM Power Systems do have lots of technical advantages ....
1. Core performance
2. I/O & memory-bandwidth
3. Enterprise Pools - Shared Procesor Pool (SPP)
4. Mobile Ressources
5. Capacity on Demand
3. Virtualization and Consolidation
4. Infrastructure Complexity
5. Availability of Infrastructure (RAS, Security)
6. Other Cost Considerations (Power & Cooling, Floor Space, Network)

© Copyright IBM Corporation 2016. Technical University/Symposia materials 56

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Other Cost Considerations when Choosing a Platform

• Infrastructure Costs Compared example

Data Centre
# of Max Power Total Power
Distributed Server - Cooling Yearly Cost Power &

*
Servers Physical (KWh) (KWh)
Power Vs x86 Efficiency (0,16cts/KWh) Cooling Ratio
Servers Theoritical Theoritical
(100%) Monthly Cost of Annual Cost
x86 24 0,90 21,50 21,50 60 320 € Frame Service Total Frame
Frame SqM Frame Space Per of Frame
Power + Cooling 45% Space Multiplier SqM
Power 4 2,95 11,81 11,81 33 133 € SqM Space

# of Total needed
0,62 1,5 1,55 600 € 11 160 €
Distributed Server - #U per Yearly Cost Floor Space
Servers Physical Total # U Rack (Max
Power Vs x86 Server (7200€/SqM)* Ratio
Servers 36U of 42U)
x86 24 2 48 2 22 320 €
Floor Space
Power 4 7 28 1 11 160 €
50%

# of WAN Ports
Distributed Server - SAN Ports per Average Cost
Servers Physical per Server/ Yearly Cost Network Ratio
Power Vs x86 Server/ Switch per Port
Servers Switch
x86 24 2 2 500 € 48 000 €
Network
Power 4 4 4 500 € 16 000 €
67%

Example:
• UCS C240 M4 (2U) Xeon E5-2670v3 12-Core 2.3GHz (2ch/24co)
• Power System E870 40 Cores (7U)
Based on Step 4 with only performance ratio :
Core Performance
Virtualization Consolidation (estimated)
SMT8
TCA Server w / Oracle
Cumul Virtual Shared Cumul
License Core CPU utilization with Estim. #
# Core # Core Proc Pool
Ratio virtualization Core
Impact
Roundup of
Power8 w/ PowerVM 2,1 x1 80% (IBM Commit.) x1 x1 x1 x1
Sum
x86 "Haswell" w/ Roundup of
1 x2.1 40% x2 x4.1 x1 x4.1
OracleVM Sum
Sum of
x86 "Haswell" w/ VMware 1 x2.1 40% x2 x4.1 x1.2 x5
Roundup

6 x 86 Xeon E5-2670v3 24 cores can be consolidated on 1 x E870 40 cores

© Copyright IBM Corporation 2016. Technical University/Symposia materials 57

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Oracle on Power vs x86 TCO Tool – Step 6 of 6

Power &
Floor Space Network
Cooling
Server
Network
Power Ratio Floor Ratio
Ratio

Power8 X1 X1 X1
x86 "Haswell" X1,8 X2 X3

© Copyright IBM Corporation 2016. Technical University/Symposia materials 58

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 IT Economics Studies
Use a business case to make a technically and financially based IT decision
Cloud Assessment
Perform a Health Check to find the right private, public or hybrid cloud solution
Examine workload size and activity, SLA and provisioning requirements, and instance costs

Workload Placement Assessment

Consolidate, offload, and place new workloads on alternative platforms
Exploit and compare platform attributes to optimize workload performance and costs

Business Value Assessment (BVA)

Understand solution attributes and how business requirements are mapped
Quantify financial benefits and compare to alternatives to determine the most compelling case

Mobile Assessment
Mitigate high-volume, low-value mobile transaction costs
Evaluate the effects of throughput, response time and other KPIs in mobile topologies

Analytics Assessment
Determine the most cost-effective infrastructure for analytics solutions
Exploit platform attributes and efficient storage solutions for Analytics and Big Data

Chargeback Analysis
Align chargeback policies to actual IT costs
Identify and overcome chargeback policies that drive adverse IT decisions

IT Best Practice Benchmarking

Compare actual IT environment with best practices in the IT industry
Improve forecast and actual spend

Available at no-charge to IBM Clients and Business Partners

© Copyright IBM Corporation 2016. Technical University/Symposia materials 59
IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Steps for an Eagle study

1. Request a study - Contact the IBM Eagle Team at [email protected]

2. Initial Call - A senior Eagle consultant will contact you to
• Understand the client situation
• Identify where a study would best help your account
• Help you identify the executive sponsor
• Join a call with you and the executive sponsor to gain their commitment
• Determine the workshop date and meeting logistics with you and the client

3. On-site Workshop - Two hours

• Determine scope of study (objectives, challenges and environment)
• Define study scenarios
• Share best practices and benchmarks learned from hundreds of client studies
• Identify action owners, study time line and next steps

4. Eagle Analysis - Approximately four weeks

• Analysis and report preparation is performed by Eagle consultant off-site
• Weekly checkpoint with you and the client’s assigned focal

5. On-site Study Presentation - One to two hours

• Eagle consultant presents findings and provides recommendations
• Client receives a detailed assessment based on their customized scenarios

© Copyright IBM Corporation 2016. Technical University/Symposia materials 60

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Eagle Presence World Wide

World wide and local teams expert in Eagle methodology and on-site client
consulting
Eagle consultants staffed in 18 countries across all GEOs

IBM Eagle Team - IT Economics Practice

Craig Bender, Director
IBM Eagle Team - IT Economics Practice
Worldwide Practice Leader
Somers, NY, USA

John Gustavson, CTO

IBM Eagle Team - IT Economics Practice
Worldwide Practice Leader
Detroit, MI, USA

Americas EMEA AP
Christopher T. von Koschembahr Alfredo Micarelli J C Yao
IBM Eagle Team IBM Eagle Team IBM Eagle Team
Americas Practice Leader EMEA Practice Leader AP Practice Leader
New York, NY, USA Rome, Italy Tokyo, Japan

© Copyright IBM Corporation 2016. Technical University/Symposia materials 61

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Case Study 1: Consolidation of x86 Servers onto Power 8

Business Problem
Direction to reduce IT spend with perceived lower cost of Linux x86

Study Objective:
• Demonstrate the cost savings that can be achieved by
consolidating the client’s SAP Oracle Database
workloads to IBM Power Systems, using a 5 Year TCO
Case 1 – Existing Case 2 – New Case 3 – New (DR)
Study Servers IBM Power Systems S822 IBM Power Systems E870
IBM Flex System x240 POWER8 3.42GHz POWER8 4.02GHz
Xeon E5-2697v2 2.7GHz (2ch/20co) (4ch/32co)
(2ch/24co) 3 servers 2 servers
Options Evaluated: Flex System Chassis & 60 cores 64 (16 Static + 48 Mobile
CoD) cores
Flex System Manager
• Case 1 – Existing IBM Flex System x240 – Compute 12 servers
288 cores
Node (Intel Ivy Bridge-EP)
Database Workload Database Workload Database Workload
• Case 2 – IBM Power Systems S822 – Scale-out Server Oracle 144 cores¹ Oracle 28 cores¹ Oracle 20 cores¹

(POWER8 3.4GHz)
• Case 3 – IBM Power Systems E870 – Enterprise Server
(POWER8 4.02GHz)

76% Saving 76% Saving

Benefit:
• Savings of kr.42 M over five years using two
Enterprise Power E880 servers
• 76% lower TCO than the x86 solution
kr. 55,491,524 kr. 13,290,566 kr. 13,396,682

© Copyright IBM Corporation 2016. Technical University/Symposia materials 62

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Cast Study 1: 5 Year TCO Comparison
60,0
Accumulated TCO Cost Comparison

50,0
60
Cost (Millions) DKK

40,0
50

Cost (Millions) DKK

30,0
40

20,0
5 Year TCO 30
Kr. 42 M
10,0 76% Saving 20

0,0
Flex System x240 3x Power Systems 2x Power Systems 10
(Existing) S822 E870
Energy 1 272 921 563 279 532 971
0
Hosting 102 966 30 890 72 076 Year 1 Year 2 Year 3 Year 4 Year 5
Networking 900 950 225 237 150 158
People - - -
Flex System x240 (Existing) 3x Power Systems S822
Software 52 189 194 10 147 899 7 248 499 2x Power Systems E870
Sys Software 819 612 954 702 933 261
Hardware 205 883 1 368 560 4 459 717

kr. 55,491,524 kr. 13,290,566 kr. 13,396,682

© Copyright IBM Corporation 2016. Technical University/Symposia materials 63
IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Case Study 2: Refreshing Power 770 Servers onto Power 8

Business Problem
Server refresh required to replace older hardware

Study Objective:
The objective of this TCO Study is to demonstrate
the cost savings that can be achieved by refreshing
to IBM POWER8 technology, over a 5-year period
for a stockbroker.

Options Evaluated: 5 Year TCO

$5.96 M
• Case 1 – IBM Power (POWER7) 770 – Existing
Infrastructure Today 33% Saving
• Case 2 – IBM Power System (POWER8) S824
• Case 3 – IBM Power System (POWER8) E870

Benefit:
• The “Enterprise Class” IBM Power System
E870 offers the lowest 5 year TCO - US$ 11.8M
• Software Costs are the most dominant factor
• IBM POWER8 requires 56% fewer Oracle
workload cores than IBM POWER7

© Copyright IBM Corporation 2016. Technical University/Symposia materials 64

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Case 3: Off-Load from Power to x86 escalates costs!
Business Problem
Leverage lowest cost solution to run business applications

Client Solution
An IT outsourcing company in Switzerland had a Power Continue to run applications on Power. The increase
install base but was convinced an x86 solution would be in core count and physical servers in the x86 solution
more economical. The client issued an RFP for the lowest drove higher run rate.
priced solution to run their business applications.
Increasing SW licensing, energy and labor significantly
5 Year TCO
affected IT costs as the company continued to grow
Benefit 14.5 M USD
10% year to year.
Run rate compare between Power and x86 indicated a Savings
Existing Power Environment x86 Environment
cumulative savings of $14.5M over 5 years using the 257 cores, 5 servers
Calculated equivalent:

Power solution. 799 cores, 50 Servers

Additional costs not included in the $14.5M delta were

migration, space and storage to deploy and house the
x86 solution.

Cost in $M

© Copyright IBM Corporation 2016. Technical University/Symposia materials 65

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Session Evaluations

YOUR OPINION MATTERS!

1 2 3 4
Submit four or more session
evaluations by 5:30pm Wednesday
to be eligible for drawings!
*Winners will be notified Thursday morning. Prizes must be picked up at
registration desk, during operating hours, by the conclusion of the event.

© Copyright IBM Corporation 2016. Technical University/Symposia materials 66

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.
 Continue growing your IBM skills

ibm.com/training
provides a comprehensive
portfolio of skills and career
accelerators that are designed
to meet all your training needs.

If you can’t find the training that is right for you with our
Global Training Providers, we can help.

Contact IBM Training at [email protected]

Global Skills Initiative

© Copyright IBM Corporation 2016. Technical University/Symposia materials 67

IBM Systems Technical Events | ibm.com/training/events may not be reproduced in whole or in part without the prior written permission of
IBM.