Professional Documents
Culture Documents
David Rhoderick
IBM SWG
April 2011
Agenda
A
wo ccru
rkl in
oa g b
d c en
Cost per unit of work
on efi
M so ts
ai
nf lid of
ra ati
m
e on
~ 200 MIPS
Processor
Processor
= Processor
Processor
Processor
Bank of China **
IBM System z9 and DB2 12,000
TCS BaNCS HP maximum benchmark 10,716
9,445*** Transactions/second 9,445
380 Million Accounts 8,983
IBM benchmark for customer 8,000 8,024
7,443
6,622
5,723
4,665
4,360
State Bank of India* 4,000
3,120
HP Itanium Superdome 2,603
TCS BaNCS 1,589
10,716 Transactions/second
0
500 Million Accounts
Largest banking benchmark 0 10,000 20,000 30,000 40,000 50,000
performance claimed by HP
MIPS
* SOURCE: Clement Report; http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA1-4027ENW.pdf Feb 2010
** SOURCE:http://www.enterprisenetworksandservers.com/monthly/art.php?2976 Source: InfoSizing FNS BANCS Scalability on IBM System z Report Date: September 20, 2006
*** Standard benchmark configuration reached 8024 tps, a modified prototype reached 9445 tps
(31,675 MIPS)
448 processors
Note: Both platforms scaled to the same performance rating (10,716 tps)
(40,313 MIPS)
996 processors
Note: Both platforms scaled to the same performance rating (10,716 tps)
years $20.0
SUN M8000 solution
Cost
$15.0
Distributed TCO is Linux on system z
2.9x 25 : 1
Rule of Three:
The cost of deploying a new application will usually be less on a
mainframe if:
1. It is an incremental workload on an existing mainframe
2. It can make use of a specialty processor
3. Disaster recovery is required
zEnterprise Economics April 2011 11
What % Of The Average IT Bill Does The
Mainframe Comprise?
Mainframe Cost as a % of IT Cost
Lesson Learned
Many applications access VSAM data on z
Some CICS logic moved down to WebLogic (1,000 MIPS),
Some CICS logic moved to DB2 store procedure
Inefficiency of data access from distributed servers actually increased
MIPS
DB2
webSlogic DB2
webSlogic CICS
VSAM VSAM
Open
Open Read
Read Close
Read Open
Read Read
Close
zEnterprise Economics April 2011 17
Government Agency
Data Expansion
Customer concerns
Mainframe too expensive
Wanted to move applications from mainframe to Bull (p5 based)
servers
Lesson Learned
Most data in IMS and DB2 on z
Unfeasible to move IMS on z to Oracle on UNIX
Database expansion from IMS hierarchical to Oracle RDB
2x-3x expansion of database
additional 2x-3x CPU for data processing
Scalability limitation of Oracle RAC
Need to partition large database
Round-robin fail-over arrangement of Oracle RAC servers would crippled
performance
$8,000,000
Distributed
Approached by Oracle to move $6,000,000 Mainframe
$2,000,000
$0
Customer concerns
TCO
Lesson Learned
Overlapped distributed server deployment resulted in over 20%
overhead
Distributed servers are typically leased in 36-month cycle
6 or more months are needed to prepare and deploy new servers,
which require a 6-month overlap for each 30-month period, or a 20%
overhead
Installing and removing distributed servers requires significant resource
as well 15 FTE
Customer concerns
Need to deploy a sales incentive application
mainframe too expensive
Lesson Learned
Client does not use VWLC Pricing, existing white space
capacity can support the new application, only $0.8M of
application tools will be needed
In comparison, the distributed solution would cost over
$18M
Customer concerns
The customer needed to add new distributed servers for
Oracle applications
The local utility company prohibits adding more cables in
the metropolitan area
Lesson Learned
Consolidate 56 HP servers into 4 IFL avoid the power
constraint
Fewer cores also reduce software license cost
Processing comparisons
Language expansion (CICS/COBOL path lengths are highly
optimized)
Conversion factor (MIPS to RPE) worsens as I/O rates increase
Oracle RAC inefficiencies compared to DB2
PowerVM on PS701
15 workloads
8 core Blade
per POWER7 blade
Workloads $14,325 per workload
IBM WebSphere ND
Monitoring software
On 4 core Older Intel
PowerVM on PS701
2 workloads
8 core Blade
per POWER7 blade
Workloads $107,437 per workload
more parallel
threads
IBM WebSphere ND
Monitoring software
On 8 core Nehalem
servers
Online banking workloads, each
23 workloads
driving 460transactions per per 32-way z/VM
z/VM on zEnterprise CEC
second with light I/O 32 IFLs
IBM WebSphere ND
Monitoring software
On 4 core Older Intel
Online banking workloads, each
driving 22 transactions per second
270 workloads per 32- z/VM on zEnterprise CEC
with light I/O
way z/VM 32 IFLs
25 heavy
True Value: Cost Per Workload
240 heavy workloads
I/O
Workloads
235 light
Cost Of Hardware Is Misleading
workloads
Note: 3yr TCA. CPO benchmarks. Equal mix of WAS ND and DB2 workloads. List prices.
zEnterprise Economics April 2011 34
Agenda
Platforms Optimized For Different Consistent Structured Management
Workloads
Growth Business
Commercial Claims Processing
Two Existing Commercial Claims Processing Systems
Homegrown CICS/DB2 Application on Mainframe
ISV 3rd-party Package running on HPUX
Which
platform is the low-cost option for future growth in
commercial claims?
36
zEnterprise Economics April 2011 36
Two Commercial Claims Processing Systems
Which system
costs less for
Production Servers
HP 9000 Superdome rp4440 future
HP Integrity rx6600
growth?
Total MIPS 11,302
Production Servers
HP 9000 Superdome rp4440
$0.12 per claim
HP Integrity rx6600
Unified management of
virtual machines
Manage ensemble as a
AIX on Power DataPower z/OS single workload with service
Blade
XI50z goals
messages messages
DataPower
XI50z
5,117 messages per sec
$33 per mps
Tests consists of measuring maximum throughput DataPower
of ESB while performing a variety of message
mediation workloads: pass-through, routing,
XI50z in zBX
transformation, and schema validation HS 22, 8 cores
T3-1B z196
SAP
Upgrade to $1.2M
20 workloads
new SPARC T3
3yr TCA
hardware HW+SW
20
SPARC T5440
32 core servers 23 POWER7 blades
538,120 total SAPs in zBX
640 cores total 184 cores total
zBX z196
myself? Do-It-Yourself
72 workloads
SAP
20 workloads
POWER7 blades
in zBX racks with
zManager zEnterprise
(Manage+Automate)
t
consistent structured
en
350
em
management practices drive
ov
300
pr increases in labor productivity
Im
250
x
18
200
150
100 t Small scale consolidation with
x Imp rovemen
50
3.9 ad hoc management achieves
0 lesser gains
2000/01 2007/08
Year
97% reduction
in labor time
Time
zManager monitors
virtual machine
Workload 2 (W2)
performance and
automatically adjusts
15 minutes CPU resources as
Goal
needed
110% Noadditional
No additionalCPU
CPU
resources needed
resource needed
100% 100%
10% more CPU
resources needed Unexpected 20% spike
in average demand,
Performance Manager
Unexpected increases entitlement by
20% spike in 20%
average 13.2% 13.2%
demand
Workload 1 Workload 1
11%
Average utilization
Average utilization
11%
Performance manager
reduces entitlement by
20%
11% Workload 2 Workload 2
11%
8.8%
Must overprovision CPU resource for both workloads Performance manager enables trading off resource from
by 10% to handle unexpected spike in demand lower priority workload, avoiding the need to
overprovision
$30.1K
Web Facing
Do-It-Yourself per workload
72 workloads Data per year
Processing
20 workloads
zBX Cost of
Acquisition $6,994K
Labor $ 714K
Total (3yr) $7,708K
72 Web facing hybrid
applications, 2 per POWER7 56
blades
POWER7 blades
20 SAP hybrid applications, 1
per POWER7 blade. DIY blades
over provisioned by 10%
in two zBX racks with
zManager zEnterprise $27.9K per
because no zManager
performance manager (Manage+Automate) workload per year
Cost of CICS/DB2 and SAP
components of workloads not
Reduced Reduced by
Capacity/Performance
by 41% 52%
Management
Automation to isolate and
fix issues
Reduced by
9%
zEnterprise integrated
management reduces cost
of labor per workload
The customer examples described are presented as illustrations of how those customers have used IBM products and the results they may have achieved. Actual environmental costs and performance characteristics may vary by customer.
Information concerning non-IBM products was obtained from a supplier of these products, published announcement material, or other publicly available sources and does not constitute an endorsement of such products by IBM. Sources for
non-IBM list prices and performance numbers are taken from publicly available information, including vendor announcements and vendor worldwide homepages. IBM has not tested these products and cannot confirm the accuracy of
performance, capability, or any other claims related to non-IBM products. Questions on the capability of non-IBM products should be addressed to the supplier of those products.
All statements regarding IBM future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.
Some information addresses anticipated future capabilities. Such information is not intended as a definitive statement of a commitment to specific levels of performance, function or delivery schedules with respect to any future products. Such
commitments are only made in IBM product announcements. The information is presented here to communicate IBM's current investment and development activities as a good faith effort to help with our customers' future planning.
Performance is based on measurements and projections using standard IBM benchmarks in a controlled environment. The actual throughput or performance that any user will experience will vary depending upon considerations such as the
amount of multiprogramming in the user's job stream, the I/O configuration, the storage configuration, and the workload processed. Therefore, no assurance can be given that an individual user will achieve throughput or performance
improvements equivalent to the ratios stated here.
Prices are suggested U.S. list prices and are subject to change without notice. Starting price may not include a hard drive, operating system or other features. Contact your IBM representative or Business Partner for the most current pricing in
your geography.
Photographs shown may be engineering prototypes. Changes may be incorporated in production models.
Trademarks of International Business Machines Corporation in the United States, other countries, or both can be found on the World Wide Web at http://www.ibm.com/legal/copytrade.shtml.
24% 20%
lower cost per lower cost per
hospital bed airline passenger
in the long run, the marketplace rewards those that make the optimum
use of the right computing resources in the right way as evidenced by
business performance.
-- * Dr. Howard Rubin, CEO and Founder Rubin Worldwide
* Based on an analysis of actual IT spend and business performance, comparing companies with greater than average mainframe mix vs. less than average
mainframe mix
zEnterprise Economics April 2011 65
Case Study:
zEnterprise Reduces Infrastructure Labor Hours
5000
4500
4000
28%
Yearly Labor Hours
fewer labor
3500 hours Change Mgmt
3000 Security Mgmt
2500 Asset Mgmt
2000 Capacity/Perf Mgmt
1500 Deployment Mgmt
1000
500
0
Do-It-Yourself zEnterprise