POWER8 Claims

advertisement
Power 競爭分析:
要做就做最好的 (Best of The Best)
May 23, 2014
POWER8 – 新亮點
POWER8 – “看到”ARM的開放, 占領Tablet, Phone市場
“促成” OpenPOWER, 伺服器市場的選擇.
POWER8 – 領先業界的4大特色
(what you expected)
More Cores
Industry
Best Practice
More Threads
Industry Leading
12 processor cores per socket (50% more than before) that
deliver better per core performance
SMT8 – 8 dynamic threads per core, supporting SMT1, 2, 4, &
8 modes dynamically across VMs
What this means
What this means
Enjoy better scale up performance, and more throughput per
scale out server node.
More Cache
Industry Leading
At 100MB, 3X the on-chip cache as POWER7 – plus 128MB of
new off-chip cache as well
What this means
Memory-intensive applications (like database) will perform
better as memory latency is reduced
You choose – Deploy VM’s in the optimal SMT mode based on
application needs.
More Bandwidth
Industry Leading
2.3X our prior gen to memory, and 2.4X our prior gen to I/O.
What this means
Data-hungry applications (like big data & analytics) will respond
twice as fast and scale more efficiently.
POWER8 – 獨有的4大創新功能
CAPI
Industry
Innovation
Transactional Memory
Innovation
Extended
Open interface allows PCIe3 devices to participate in
operations at memory speed without risk.
Borrowed from the mainframe, this technology speeds up
memory writes by reducing contention.
What this means
What this means
Gain orders of magnitude application performance with PCI
card technology w/o hiring specialized skills
A feature that improved OLTP database performance by 45%
on System z is now available on Power.
Native PCIe
Innovation
On Power
PowerKVM
Innovation
On Power
Integrating PCIe Gen 3 into the processor boosts performance
by eliminating logic overhead.
KVM, the open-source virtualization solution, can be used to
manage Linux-only systems.
What this means
What this means
I/O intensive data applications will run faster due to high
bandwidth, low latency communications.
Data centers can now standardize their clouds with a single
open-source virtualization technology.
POWER8 delivers insights 82x faster (超級快: 傳統資料庫Turbo加速引擎)
Running Cognos BI reports and analytics on POWER8 with DB2 with BLU Acceleration versus Ivy Bridge with a
traditional database
• DB2 with BLU Acceleration on POWER8 for Cognos BI is ‘Fast on Fast on Fast!’
• Real world Cognos BI Telco workload
– 60 concurrent users running 70%/25%/5% split of simple,
intermediate and complex reports
• First processer designed for Big Data with massive parallelism
and bandwidth for real-time results
– 18x more throughput for simple reports
– 40x more throughput for intermediate reports
– 747x better throughput for complex reports
Dynamic
Query
Dynamic
Cubes
Compatible
Query
Based on IBM internal tests as of April 17, 2014 comparing IBM DB2 with BLU Acceleration on Power with a comparably tuned competitor row store database server on x86 executing a materially identical 2.6TB BI workload in a controlled laboratory environment. Test measured 60
concurrent user report throughput executing identical Cognos report workloads. Competitor configuration: HP DL380p, 24 cores, 256GB RAM, Competitor row-store database, SuSE Linux 11SP3 (Database) and HP DL380p, 16 cores, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3
(Cognos). IBM configuration: IBM S824, 24 cores, 256GB RAM, DB2 10.5, AIX 7.1 TL2 (Database) and IBM S822L, 16 of 20 cores activated, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3 (Cognos). Results may not be typical and will vary based on actual workload, configuration,
applications, queries and other variables in a production environment.
POWER8 thrives on the complex (實證: 資料越大, 越複雜 => P8輕鬆完成)
Running Cognos BI reports and analytics on POWER8 with DB2 with BLU Acceleration versus Ivy Bridge with a
traditional database
Reports
per Hour
System
Simple
Query
Intermediate
Query
Complex
Query
POWER
POWER8
42750
7408
202
X86
Ivy Bridge
Speed up
2267
185
0.27
18x
40x
747x
Dynamic
Query
Dynamic
Cubes
Compatible
Query
Larger and more complex queries have more speedup
Based on IBM internal tests comparing IBM DB2 with BLU Acceleration system with a comparably tuned competitor configuration executing a materially identical 2.6TB operational analytics workload in a controlled laboratory environment. Test measured
60 concurrent user report throughput executing identical Cognos report workloads. Report per hour (RPH) metric calculated for each category of reports as total completed reports/hours to completion of all reports in the category. Competitor configuration:
HP DL380p, 24 cores, 256GB RAM, Traditional Database, SuSE Linux 11SP3 (Database) and HP DL380p, 16 cores, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3 (Cognos). IBM configuration: IBM S824, 24 cores, 256GB RAM, DB2 10.5, AIX 7.1
TL2 (Database) and IBM S824, 16 cores, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3 (Cognos). Results may not be typical and will vary based on actual workload, configuration, applications, queries and other variables in a production environment.
© 2014 International Business Machines Corporation
7
Concurrency Test: DB2 on POWER8 vs. Competitor on x86
The results are clear! (資料庫應用對比
Simple Reports
45000
8000
200
40x
747x
35000
30000
25000
20000
15000
10000
Reports per Hour
7000
Reports per Hour
Reports per Hour
Complex Reports
Intermediate Reports
18x
40000
: P8 vs Intel Ivy Bridge)
6000
5000
4000
3000
100
50
2000
5000
150
1000
0
DB2 on Other
P8
on x86
0
0
DB2 on Other on
P8
x86
DB2 on Other on
P8
x86
Based on IBM internal tests as of April 17, 2014 comparing IBM DB2 with BLU Acceleration on Power with a comparably tuned competitor row store database server on x86 executing a materially identical 2.6TB BI workload in a controlled laboratory
environment. Test measured 60 concurrent user report throughput executing identical Cognos report workloads. Report per hour (RPH) metric calculated for each category of reports as total completed reports/hours to complete all reports in the category.
Competitor configuration: HP DL380p, 24 cores, 256GB RAM, Competitor row-store database, SuSE Linux 11SP3 (Database) and HP DL380p, 16 cores, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3 (Cognos). IBM configuration: IBM S824, 24 cores,
256GB RAM, DB2 10.5, AIX 7.1 TL2 (Database) and IBM S822L, 16 of 20 cores activated, 384GB RAM, Cognos 10.2.1.1, SuSE Linux 11SP3 (Cognos). Results may not be typical and will vary based on actual workload, configuration, applications,
queries and other variables in a production environment.
© 2014 International Business Machines Corporation
8
NO trade-offs: 2x Throughput at Half the Cost of x86 (還覺得P8貴嗎?)
POWER8 Delivers Over TWICE the Throughput Relative to Ivy Bridge-EP at 47% Lower Cost for an Online
Banking Workload (銀行業線上交易處理系統: Total HW+ SW vs Price/Perfor)
Web Application
Power S824
AIX
WAS WAS
AIX
…..
AIX
WAS
WAS
AIX
AIX
8 VMs
183,800
$3.09
PowerVM
DB2
User Interactions
per second
per UI per sec
WebSphere on platform
Database off platform
2S/24 Core POWER8 (3.525 GHz)
Lower cost
HP - ProLiant DL380p Gen8
WAS
…..
RHEL RHEL
WAS
WAS
Faster
47%
Online Banking Workload v3.6
WAS
2.1x
6 VMs
RHEL RHEL
85,939
$5.84
VMware ESXi
DB2
User Interactions
per second
per UI per sec
per UI per sec
WebSphere on platform
Database off platform
2S/24 Core Ivy Bridge-EP (2.7 GHz)
Both Servers configured to achieve maximum throughput
This is an IBM internal study designed to replicate a typical IBM customer workload usage in the marketplace. The results were obtained under laboratory conditions, and not in an actual customer environment. IBM's internal workload studies are not benchmark applications, nor are they based
on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or testing conditions may produce different results and may vary based on actual configuration, applications, specific queries and other variables in a production
environment. Prices, where applicable, are based on published US list prices for both IBM and competitor, and the cost calculation compares the cost per request for the 3yr life of the machine. 3 year total cost of acquisition comparisons are based on similar expected hardware, software, service
& support offerings
還覺得P8貴嗎? 算給你看: 買一台賺兩台
銀行業線上交易處理系統 (Friendly Bank Throughput Test v3.6 Pricing)
Cost of IBM:
Cost of HP:
IBM - Power S824 (STG Estimate) = $71,391
HP - ProLiant DL380p Gen8 (2U) Xeon E5-2697
v2 2.7GHz: LinuxRH: (2ch/24co) = $17,066
PowerVM Standard (inc 3yr 24x7 SWMA) = 362 ($ per core)
* 24 (cores) = $8,688
AIX Express (inc 3 yr 24x7 SWMA) = $557 * 24 (cores) =
$13,368
P840 HW Total = $71,391 + $8,688 + $13,368 = $93,447
WAS ND = $202 (per PVU) x 70 (server PVU rating) x 24
(no of cores) x 1.4 (S&S yrs 2-3) = $475,104
2.1x
Faster
47%
Lower cost
per UI per sec
Cost of VMware EE (2 socket) = $9,546
Total HW = $17,066 + $9,546 = $26,612
WAS ND = $202 (per PVU) x 70 (server PVU
rating) x 24 (no of cores) x 1.4 (S&S yrs 2-3) =
$475,104
HW + SW cost = $26,612 + $475,104= $501,716
Total HW + SW = $ 93,447+ $475,104= $568,551
UIPS = 85939 (6 VMs)
UIPS = 183,800 (SMT 8)
Cost per UIPS = $501,716/ 85939= $5.84
Cost per UIPS = $ 568,551/ 183,800 = $3.09
183,800
$3.09
User Interactions
per second
per UI per sec
WebSphere on platform
Database off platform
85,939
$5.84
User Interactions
per second
per UI per sec
WebSphere on platform
Database off platform
Derive insights faster with Linux on POWER8
(相同App, 相同OS, 不同CPU, P8就是更快 67%, Linux on P8 vs. Linux on Intel)
67% Faster Reports and Analytics with Cognos BI on POWER8 vs. Ivy Bridge-EP
• Derive insights 67% faster
with Cognos Business Intelligence (Cognos
v10.2) running on Linux on POWER8 versus
Linux on Intel Ivy Bridge-EP
• Utilize Predictive Analytics to get the
answers before your competitors
67% faster is based on IBM Internal Testing of sample analytic workloads; current as of March 28, 2014. Performance improvement figures are cumulative of all queries in the workload. Individual results will vary depending on individual
workloads, configurations and conditions.
IBM Analytics Stack: IBM Power System S824; 24 cores / 192 threads, POWER8; 3.5GHz, 384 GB memory, DB2 10.1 and Cognos v 10.2
Competitive stack: HP DL380p; 24 cores / 48 threads; Intel E5-2697 v2; 2.7 GHz; 384 GB; DB2 10.1 and Cognos v 10.2
這是Big Data + Hadoop: 2倍效能 vs. X86
POWER8 delivers 2x performance on Big Data / Hadoop
Terasort benchmark on a POWER8 doubles the system capacity of the best x86 published result
 POWER8 S822L will deliver over 2x the performance
of the best published x86 system
… and continues to offer far superior RAS
 POWER8 exploits additional cores, more threads,
larger caches, memory bandwidth
 Terasort is a popular benchmark to measure the
performance of a Hadoop solution
 Sorts a large dataset (10 TB) in parallel
 Exercises the Map-reduced framework and Hadoop
Distributed File System (HDFS)
IBM Analytics Stack: IBM Power System S822L; 24 cores / 192 threads, POWER8; 3.0GHz, 512 GB memory, RHEL 6.5, InfoSphere BigInsights 3.0
http://www.cisco.com/en/US/solutions/collateral/ns340/ns517/ns224/ns944/le_tera.pdf
2x
這是rPerf 效能, 採用P8: 對軟體授權$$$ ,省1/2
POWER8 can consolidate 2 POWER7 boxes
IBM POWER processors continue to deliver improved core performance – up to 2x versus POWER7
 In scale-out environments POWER8
delivers the same performance in less
than ½ the footprint of POWER7
Relative Core Performance
2
 ½ reduction in per-core software
licenses with POWER8 processorbased systems due to improved core
performance
 Improved core architecture
 Improved cache structure
 Improved threading
 Management & Energy Cost reduction
Based on IBM Internal Measurements/Projections on 2-socket/4U POWER8 systems versus 2-socket/4U POWER7 systems
1.5
1
0.5
0
POWER7
Floating Point
OLTP
CRM
Java
Integer
新的處理器出來, 應該是越來越快, 軟體授權才能省, 但
Intel卻….?
POWER8 moves forwards while x86 moves backwards
IBM POWER systems continue to deliver improved system performance and more value per SW $ spent
•
Infrastructure Software Price-performance has been REDUCED on Intel servers by up to 14%
–

Software Licensing has increased by 1.5x
–

Assumes flat system pricing
12 cores versus 8 cores
Performance has increased only 1.29-1.47x (IBM increased by 1.71-2.28x)
–
x86 publishes on 2-socket Sandy Bridge and 2-socket Ivy Bridge
x86
“Sandy
Bridge”
x86
“Ivy Bridge”
2-socket
E5-2690
2-socket
E5-2697
POWER7+
POWER8
2-socket
POWER7+
2-socket
POWER8
16
24
1.50
16
24
1.50
ERP SAP 2-Tier
7960
10240
1.29
10000
21212
2.12
SPECint_rate
693
1020
1.47
884
1750
1.98
SPECfp_rate
510
734
1.44
602
1370
2.28
SPECjEnterprise2010
8310
11260
1.35
13161
22543
1.71
# Cores
System
Performance
Ratio
1
Based on generational comparisons of SW that utilizes per core pricing and 50% more cores in per system (Power: 8c POWER7 to 12c POWER8; x86:8c E5-2690 to 12c E5-2697 v2)
2
Performance is based on published x86 data and published/projected POWER7+ & POWER8. Workloads are ERP, Integer, Floating Point, Java
System
Performance
Ratio
新處理器: 效能應該是越來越快, 但Intel卻越來越慢….? 發
生啥事?
POWER8 moves forwards while x86 moves backwards
Relative Core Performance
Relative Core Performance
IBM POWER processors continue to deliver improved core performance – up to +35% versus POWER7+ while
Intel went backwards (-10%) with Ivy Bridge versus Sandy Bridge.
1.4
1.2
1
0.8
0.6
0.4
0.2
0
P7+
P8 CRM
P8 Integer
P8
Floating
Point
P8 Java
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Sandy
Bridge
Ivy Bridge Ivy Bridge Ivy Bridge Ivy Bridge
CRM
Integer
Floating
Java
Point
1
Based on generational comparisons of SW that utilizes per core pricing and 50% more cores in per system (Power: 8c POWER7 to 12c POWER8; x86:8c E5-2690 to 12c E5-2697 v2)
2
Performance is based on published x86 data and published/projected POWER7+ & POWER8. Workloads are ERP, Integer, Floating Point, Java
SAP 的好朋友: SAP run on POWER8, 2X vs X86
Transaction Processing Workload – SAP S&D 2-tier ERP 6
IBM Power System S824 with DB2 10.5 vs.. Competition
Over 2x
better, 24 core performance than nearest x86 competitive results
Up to 2x greater performance than previous POWER generation
Exploiting more cores, more threads and L3 cache than past POWER platforms
2x
(1.0) IBM Power System S824 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 4 processors / 24 cores / 96 threads, POWER8; 3.52GHz, 512 GB memory, 21,212 SD benchmark users, running AIX® 7.1 and DB2® 10.5, dialog response: 0.98 seconds,
line items/hour: 2,317,330, dialog steps/hour: 6.952,000 SAPS: 115,870 database response time (dialog/update): 0.011 sec / 0.019sec, CPU utilization: 99%, Certification #: * Results valid as of 3/24/14. * Certification # 2014016. Source: http://www.sap.com/benchmark.
(1.1) Fujitsu RX300 S8 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 2 processors / 24 cores / 48 threads. Intel Xeon E5-2697 processor 2.70 GHz, 256 GB memory, 10.240 SD benchmark users, running Windows Server 2012 SE and SQL Server 2012,
Certification #: 2013024
(1.2) Cisco UCS c240 M3 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 2 processors / 24c ores / 48 threads. Intel Xeon E5-2697 processor 2.70 GHz, 256 GB memory, 10.045 SD benchmark users, running Windows Server 2012 DE and SQL Server
2012, Certification #: 2013038
(1.3) HP ProLiant BL460c Gen8 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 2 processors / 24 cores / 48 threads. Intel Xeon E5-2697 processor 2.70 GHz, 256 GB memory, 10.025 SD benchmark users, running Windows Server 2012 DE and SQL
Server 2012, Certification #: 2013025
(2.1 IBM Flex System p270 Compute Node on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 4 processors / 24 cores / 96 threads, POWER7+; 3.4GHz, 256 GB memory, 12.528 SD benchmark users, running AIX® 7.1 and DB2® 10 .5 Certification #:
3012019 Source: http://www.sap.com/benchmark.
(1.1)IBM Flex System p260 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 2 processors / 16 cores / 64 threads, POWER7+; 4.1GHz, 256 GB memory, 10,000 SD benchmark users, running AIX® 7.1 and DB2® 10, Certification #: 2012035
SAP 的好朋友: 2x ERP Users vs Oracle T5 Core
Double the user capacity with the same core count
POWER8 supports over 2x ERP users per x86 core (and 2.8x the users vs. Oracle T5 core)
SAP SD Standard Application Benchmark Results, 2-Tier: SD Benchmark Users per Core
SAP enhancement package 5 for SAP ERP 6.0 As of April 28, 2014
Source: http//www.sap.com/solutions/benchmark/sd2tier.epx
POWER8 per core result is
>2x greater than the best
competitors!
900
800
700
2X
600
SD
Benchmark 500
Users per 400
Core
more users
Vs x86 E5 core
2.8X
more users
Vs T5
300
200
100
0
IBM
S824
POWER8
24c/192t
Oracle
M5-32
M5
192c/1536t
Fujitsu
Primergy
E5-2697 v2
24c/48t
HP
DL580
E7-4890 v2
60c/120t
Oracle
T5-8
T5
128c/1024t
1.0) IBM Power System S824 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 4 processors / 24 cores / 96 threads, POWER8; 3.52GHz, 512 GB memory, 21,212 SD benchmark users, running AIX® 7.1 and DB2® 10.5, dialog response: 0.98 seconds, line
items/hour: 2,317,330, dialog steps/hour: 6.952,000 SAPS: 115,870 database response time (dialog/update): 0.011 sec / 0.019sec, CPU utilization: 99%, Certification #: * Results valid as of 3/24/14. * Certification # 2014016. Source: http://www.sap.com/benchmark.
(2) Oracle SPARC Server M5-32 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 32 processors/192 cores/1536 threads, SPARC M5; 3.60 GHz, 4,096 GB memory; 85,050 SD benchmark users, running Solaris® 11 and Oracle 11g; Certification # 2012013009.
Results valid as of April 28, 2014. Source: http://www.sap.com/benchmark.
(3) Fujitsu PRIMERGY RX300 S8 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 2 processors/24 cores/48 threads, Intel Xeon Processor 2697v2; 2.70 GHz, 256 GB memory; 10,240 SD benchmark users, running Windows Server 2012 Standard Edition and SQL
Server 2012; Certification # 2013024. Results valid as of April 28, 2014. Source: http://www.sap.com/benchmark.
(4) HP ProLiant DL580 Gen8on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 4 processors/60 cores/120 threads, Intel Xeon Processor 4890 v2; 2.80 GHz, 1024 GB memory; 24,450 SD benchmark users, running Windows Server 2012 Standard Edition and SQL
Server 2012; Certification # 2043004. Results valid as of April 28, 2014. Source: http://www.sap.com/benchmark.
(5) Oracle SPARC Server T5-8 on the two-tier SAP SD standard application benchmark running SAP enhancement package 5 for the SAP ERP 6.0 application; 8 processors/128 cores/1024 threads, SPARC T5; 3.60 GHz, 2,048 GB memory; 40,000 SD benchmark users, running Solaris® 11 and Oracle 11g; Certification # 2012013008.
Results valid as of April 28, 2014. Source: http://www.sap.com/benchmark.
SAP and all SAP logos are trademarks or registered trademarks of SAP AG in Germany and in several other countries. All other product and service names mentioned are the trademarks of their respective companies.
Java 的好朋友: 2x Ejops vs X86 vs Oracle T5 Core
 POWER8 processing cores2 are the fastest
in the industry for Enterprise Java workloads
2x more performance per core than Intel E52697 v2 offerings 3
1.77x more performance per core than Oracle
T5-2 offerings 4
EjOPS per Core
POWER8 has 2x per core performance of x86 Ivy Bridge systems in
Enterprise Java-based environments 1
1000
900
800
700
600
500
400
300
200
100
0
939
469
POWER8
1
Performance based on published SPECjEnterprise2010 results www.spec.org/jEnterprise2010/;
2
IBM Power Systems S824 (24 cores)
3
2 Intel E5-2697 v2 (24 cores)
4
1 Oracle T5-2 (32 cores)
Intel
532
Oracle
Oracle EBS的好朋友: 2x Check/hr vs X86 core
IBM Power System S824 delivers Best of Breed eBS 12.1.3 Payroll
performance
Over 2x more performance per-core than Cisco and
with higher overall through-put on fewer cores
IBM Power S y st em S 824 Per for ma nce
(1 )
IBM Power S y st em S 824 Per for ma nce (1 )
100000
2X !
80000
60000
42402
40000
49345
4 0 0 ,0 0 0
0
All results use Oracle eBS 12.1.3 Payroll Batch Extra Large Kit and are current as of 3/24/2014.
For more information go to http://www.oracle.com/us/solutions/benchmark/apps-benchmark/results-166922.html
7 8 9 ,5 1 5
6 0 0 ,0 0 0
0
Oracle
BL460c
16-core
1 ,0 1 7 ,6 3 9
8 0 0 ,0 0 0
2 0 0 ,0 0 0
Cisco UCS
C240 M3
24-core
1 ,0 9 0 ,9 0 9
1 ,0 0 0 ,0 0 0
20000
IBM Power
S824
12-core
(1)
1 ,2 0 0 ,0 0 0
90909
Checks per hour
Checks per hour/core
(pe r c o r e )
IBM Power
S824
12-core
Cisco UCS
C240 M3
24-core
Oracle
BL460c
16-core
Sibel CRM的好朋友: 3x DB Perf vs X86 core
IBM Power System S824 delivers new High-water Siebel CRM Release 8.1.1.4
performance
Over 3 times the DB performance per-core than previous results
Highest overall users supported on fewer cores!
New #1
IBM Power
S824
6-core
Oracle
SPRAC T4-2
16-core
Cisco
UCS B200 M3
16-core
(1) All results use Siebel 8.1.1.4 PSPP Kit and are current as of 3/24/2014. For more information go to http://www.oracle.com/us/solutions/benchmark/white-papers/siebel-167484.html
3.3 X
IBM Power
S824
6-core
Oracle
SPRAC T4-2
16-core
Cisco
UCS B200 M3
16-core
POWER8 保證: 給我愈多Workload,達使用率保證65%下.
效能不會下降. (沒有任何一家Vendor 敢掛保證)
POWER8 Performance Utilization Guarantee
System Performance Utilization
IBM has designed these POWER8 systems to operate at
industry-leading levels of efficiency, guaranteeing the system will
perform as warranted while at a sustained 65% utilization – a
rate higher than common competitive platform utilization levels
System Performance Utilization Guarantee
When a Client acquires a POWER8 one or two-socket 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.
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, 2014, 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 - ninety (90) 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.
POWER8 四大高手: 東邪, 西毒, 南帝,北丐 vs. X86
ERP , JAVA , SPECint_rate , SPECfp_rate
Up to TWICE the performance across key workloads
ERP – SAP 2-Tier (Users)
2.1x Performance
25000
Java - SPECjEnterprise2010 (EjOPS)
2.0x Performance
25000
20000
20000
SPECint_rate2006
1.7x Performance
2000
1600
1800
1400
1600
1200
1400
15000
15000
10000
10000
1200
1000
1000
800
800
600
600
5000
5000
400
400
200
200
0
Fujitsu RX300 S8
2s/24c/48t
Intel Xeon Ivy Bridge
1)
2)
3)
4)
POWER S824
2s/24c/192t
IBM POWER8
0
Oracle Sun X4-2
2s/24c/48t
Intel Xeon Ivy Bridge
POWER S824
2s/24c/192t
IBM POWER8
SPECfp_rate2006
1.9x Performance
0
SuperMicro
SuperServer 6027AX-TRF
2s/24c/48t
Intel Xeon Ivy Bridge
POWER S824
2s/24c/192t
IBM POWER8
0
SuperMicro
POWER S824
SuperServer 6027AX-TRF
2s/24c/192t
2s/24c/48t
IBM POWER8
Intel Xeon Ivy Bridge
Results are based on best published results on Xeon E5-2697 v2 .
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 April 28, 2014. Source: http://www.sap.com/benchmark
SPECjEnterprise2010 results are valid as of 4/22/2014. For more information go to http://www.specbench.org/jEnterprise2010/results/
SPECcpu2006 results are submitted as of 4/22/2014. For more information go to http://www.specbench.org/cpu2006/results/
POWER8 四大高手: 東邪, 西毒, 南帝,北丐 vs. X86
ERP , JAVA , SPECint_rate , SPECfp_rate
Up to TWICE the performance across key workloads
ERP – SAP 2-Tier (Users)
2.1x Performance
25000
Java - SPECjEnterprise2010 (EjOPS)
2x Performance
25000
20000
20000
SPECint_rate2006
1.8x Performance
2000
1600
1800
1400
1600
1200
1400
15000
15000
10000
10000
1200
1000
1000
800
800
600
600
5000
5000
400
400
200
200
0
Fujitsu RX300 S8
2s/24c/48t
Intel Xeon Ivy Bridge
1)
2)
3)
4)
POWER S824
2s/24c/192t
IBM POWER8
0
Oracle Sun X4-2
2s/24c/48t
Intel Xeon Ivy Bridge
POWER S824
2s/24c/192t
IBM POWER8
SPECfp_rate2006
2x Performance
0
Dell PowerEdge T620
2s/24c/48t
Intel Xeon Ivy Bridge
POWER S824
2s/24c/192t
IBM POWER8
0
Dell PowerEdge T620
2s/24c/48t
Intel Xeon Ivy Bridge
Results are based on best published results on Xeon E5-2697 v2 from the top 5 Intel system vendors.
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 April 28, 2014. Source: http://www.sap.com/benchmark
SPECjEnterprise2010 results are valid as of 4/22/2014. For more information go to http://www.specbench.org/jEnterprise2010/results/
SPECcpu2006 results are submitted as of 4/22/2014. For more information go to http://www.specbench.org/cpu2006/results/
POWER S824
2s/24c/192t
IBM POWER8
POWER8 業界性能標竿測試 vs X86 Core (實驗室調優)
Performance Comparison – POWER8 vs. x86 E5
IBM POWER8 core and system performance is 2x the x86 Xeon E5-2697 v2 core performance
•
•
Published Benchmarks –
ALL data is PUBLISHED or SUBMITTED (SPECint_rate2006 and SPECfp_rate2006 on the POWERS824)
x86
“Ivy Bridge”
IBM
POWER S824
Intel Xeon E5-2697 v2
(except where noted)
POWER8 @ 3.5 GHz
24
24
SAP 2-Tier
10240
21212
2.1
SPECint_rate2006 (peak)
1020
1750
1.7
SPECfp_rate2006 (peak)
734
1370
1.9
11260
22543
2.0
1090909
2.1
# Cores
SPECjEnterprise2010
Oracle eBS 12.1.3 Payroll
Siebel CRM Release 8.1.1.4
1)
2)
3)
4)
5)
1017639
POWER8 vs. x86 Core
Performance Ratio
(12-core)
10000
50000
(16-core E5-2690)
(6-core)
13.3
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 April 28, 2014. Source: http://www.sap.com/benchmark
SPECcpu2006 results are submitted as of 4/22/2014. For more information go to http://www.specbench.org/cpu2006/results/
SPECjEnterprise2010 results are valid as of 4/22/2014. For more information go to http://www.specbench.org/jEnterprise2010/results/
Oracle eBS 12.1.3 Payroll Batch Extra Large Kit and are current as of 3/24/2014. For more information go to http://www.oracle.com/us/solutions/benchmark/apps-benchmark/results-166922.html
Siebel 8.1.1.4 PSPP Kit and are current as of 3/24/2014. For more information go to http://www.oracle.com/us/solutions/benchmark/white-papers/siebel-167484.html
POWER8 業界性能標竿測試 vs X86 Core (現實人生使用)
Core Performance Comparison – POWER8 vs. x86
IBM POWER8 core performance is up to 6.7x the x86 Xeon E5-2697 v2 core performance (typical customer
utilization)
•
•
•
Industry Standard Benchmarks –
All Ivy Bridge performance numbers are IBM internal projections and publishes where available
IBM S824 data is published/projected
x86
“Ivy Bridge”
IBM
Power S824
Intel Xeon E5-2697 v2
Power 8 @ 3.5 GHz
P8 Util: 100%
x86 Util: 100%
P8 Util: 65%
x86 Util: 40%
P8 Util: 65%
x86 Util: 20%
24
24
Benchmark Utilization
Utilization with
virtualized x86
Utilization without
virtualized x86
OLTP
2100
3585
1.7
2.8
5.5
ERP SAP 2-Tier
10240
21212
2.1
3.4
6.7
SPECjbb (k JOPS)
2120
3777
1.8
2.9
5.8
SPECint_rate
1020
1750
1.7
2.8
5.6
SPECfp_rate
734
1370
1.9
3.0
6.0
11260
22543
2.0
3.3
6.5
# Cores
SPECjEnterprise2010
POWER8 vs. x86 Core Performance Ratio
LEGEND:
Published
Projected
25
POWER8 快4x 到 9x vs X86 Core (一場效能遊戲12種玩法)
POWER8 delivers up to 4x the performance of x86 sweet spot
Don’t allow x86 vendors to mix top of the line performance claims with low-performance chip & system pricing
SPECint2006_rate
(2-socket)
Intel Ivy Bridge-EP


Ivy Bridge (E5 family of chips) is the most current
commodity 2-socket x86 offering
 Varies from 4 to 12 cores and $200 to $2600 per chip
 x86 performance marketing messages are focused
on 12-core E5-2697 v2 (top of the line)
POWER8 performance is 2x versus E5-2697 v2 (top
of the line) across multiple benchmarks (i.e.
SPECjEnterprise2010, SAP 2-tier, etc.)
 2.5x to 4x the performance of the x86 sweet spot
(based on SPECint2006_rate)
 Up to 9x compared to the bottom of the line Intel
chip (based on SPECint2006_rate)
(1)
(2)
Proc.
# Cores
Freq.
List Price
(kUnits)
Result
S824 vs. x86
SKU
E5-2697 v2
12
2700
2614
1020
1.72
E5-2695 v2
12
2400
2336
921
1.90
E5-2690 v2
10
3000
2057
900
1.94
E5-2680 v2
10
2800
1723
860
2.03
E5-2670 v2
10
2500
1552
816
2.14
E5-2660 v2
10
2200
1329
748
2.34
E5-2650 v2
8
2600
1166
689
2.54
E5-2640 v2
8
2000
885
542
3.23
E5-2630 v2
6
2300
612
506
3.46
E5-2620 v2
6
2100
406
430
4.07
E5-2609 v2
4
2500
294
250
7.00
E5-2603 v2
4
1800
202
186
9.41
Performance data based on best published peak SPECint_rate2006 results. All results can be found at www.spec.org
Intel chip prices based on Intel published list prices. All chips and chip prices can be found at http://www.intc.com/pricelist.cfm
POWER8 業界性能標竿測試 vs X86 Core (最新E7 Intel)
Performance Comparison – POWER8 vs. x86 E7
IBM POWER8 core performance is 2x the x86 Xeon E7-4890 v2 core performance
•
•
現實環境使用率對比
Published Benchmarks –
ALL data is PUBLISHED or SUBMITTED (SPECint_rate2006 and SPECfp_rate2006 on the POWERS824)
x86
“Ivy Bridge”
IBM
POWER S824
Intel Xeon E7-x890 v2
POWER8 @ 3.5 GHz
P8 Util: 100%
x86 Util: 100%
P8 Util: 65%
x86 Util: 40%
P8 Util: 65%
x86 Util: 20%
60
24
Benchmark
Utilization
Utilization with
virtualized x86
Utilization without
virtualized x86
SAP 2-Tier
25000
21212
2.1
3.4
6.9
SPECint_rate2006
2400
1750
1.8
3.0
5.9
SPECfp_rate2006
1770
1370
1.9
3.1
6.3
# Cores
POWER8 vs. x86 Core Performance Ratio
LEGEND:
1)
2)
6x vs X86
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 April 28, 2014. Source: http://www.sap.com/benchmark
SPECcpu2006 results are submitted as of 4/22/2014. For more information go to http://www.specbench.org/cpu2006/results/
Published
Projected
POWER8 業界性能標竿測試 vs T5 Core
Core Performance Comparison – POWER8 vs. T5
IBM POWER8 core performance is up to 9x the T5 core performance (Typical customer utilization)
•
•
•
Industry Standard Benchmarks –
All Oracle performance numbers are IBM internal projections/measurements or publishes
IBM S824 data is published/projected
現實環境使用率對比
Oracle
T5-8
IBM
Power S824
T5 @ 3.6 GHz
Power 8 @ 3.5 GHz
# Cores
128
24
OLTP
8552
ERP SAP 2-Tier
7x+ vs T5
POWER8 vs. T5 Core Performance Ratio
P8 Util: 100%
T Util: 100%
P8 Util: 65%
T Util: 40%
P8 Util: 65%
T Util: 20%
Benchmark Utilization
Utilization with
virtualized T
Utilization without
virtualized T
3585
2.2
3.6
7.3
40000
21212
2.8
4.6
9.2
SPECint_rate
3750
1750
2.5
4.0
8.1
SPECfp_rate
3020
1370
2.4
3.9
7.9
SPECjEnterprise2010
57422
22543
2.1
3.4
6.8
LEGEND:
1)
2)
3)
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 April 28,
2014. Source: http://www.sap.com/benchmark
SPECcpu2006 results are submitted as of 4/22/2014. For more information go to http://www.specbench.org/cpu2006/results/
SPECjEnterprise2010 results are valid as of 4/22/2014. For more information go to http://www.specbench.org/jEnterprise2010/results/
Published
Projected
POWER8 業界性能標竿測試 vs M6 Core
Core Performance Comparison – POWER8 vs. M6
IBM POWER8 core performance is up to nearly 10x the M6 core performance (typical customer utilization)
•
•
•
Industry Standard Benchmarks –
All Oracle performance numbers are IBM internal projections as no publishes are available
IBM S824 data is published/projected
現實環境使用率對比
Oracle
M6-32
IBM
Power S824
M6 @ 3.6 GHz
Power 8 @ 3.5 GHz
384
24
OLTP
20000
ERP SAP 2-Tier
8x+ vs M6
POWER8 vs. M6 Core Performance Ratio
P8 Util: 100%
M Util: 100%
P8 Util: 65%
M Util: 40%
P8 Util: 65%
M Util: 20%
Benchmark Utilization
Utilization with
virtualized M
Utilization without
virtualized M
3585
2.9
4.7
9.3
150000
21212
2.3
3.7
7.4
SPECjbb (k JOPS)
23000
3777
2.6
4.3
8.5
SPECint_rate
11000
1750
2.5
4.1
8.3
SPECfp_rate
7000
1370
3.1
5.1
10.2
No Publishes
22543
N/A
N/A
N/A
# Cores
SPECjEnterprise2010
LEGEND:
Published
Projected
29
POWER8 業界性能標竿測試 vs HP Itanium Core
Core Performance Comparison – POWER8 vs. Itanium
IBM POWER8 core performance is up to 15x the Itanium core performance (typical customer utilization)
•
•
•
Industry Standard Benchmarks –
All HP Superdome performance numbers are IBM internal projections as no publishes are available
IBM S824 data is published/projected
現實環境使用率對比
10x+ vs
Itanium
HP
Superdome (8s)
IBM
Power S824
Intel Itanium 9560 @
2.4GHz
Power 8 @ 3.5 GHz
P8 Util: 100%
Itan. Util: 100%
P8 Util: 65%
Itan. Util: 50%
P8 Util: 65%
Itan. Util: 20%
64
24
Benchmark Utilization
Utilization with
virtualized Itan.
Utilization without
virtualized Itan.
OLTP
3270
3585
2.9
3.8
9.5
ERP SAP 2-Tier
18000
21212
3.1
4.1
10.2
SPECjbb (k JOPS)
2760
3777
3.6
4.7
11.9
SPECint_rate
1481
1750
3.2
4.1
10.2
SPECfp_rate
1467
1370
2.5
3.2
8.1
No publishes
22543
N/A
N/A
N/A
# Cores
SPECjEnterprise2010
POWER8 vs. Itanium Core Performance Ratio
LEGEND:
Published
Projected
30
舊換新: P8 - HW 減量, SW精省, but X86 - HW 減量, SW授權高昇
Intel: Reduce HW infrastructure in a virtualized environment but increase per-core SW costs…
POWER8: Reduce HW infrastructure EVEN MORE and REDUCE per-core SW costs up to 54%
Current Configuration

100 2-socket x86 servers

Xeon X5690 processor

12 cores per server, 2
threads per core

3.46 GHz

1200 total cores

VMware
POWER8
Platform
(2014)
Power S822L Configuration

23 2-socket POWER8 servers

POWER8 processor

24 cores per server, 8 threads
per core

3.0 GHz

552 total cores
Per-core SW Costs DECREASE up to 54%!!!
Current x86
Platform
(circa 2011)
1.5X
increased
throughput
POWER8 DELIVERS EQUAL CAPACITY with
1/3 of the Servers (LOWER MANAGEMENT COSTS)
1/3 of the Cores (LOWER SW COSTS)
1/3 of the Space (LOWER INFRASTRUCTURE COSTS)
ALL for <50% of the x86 HW TCA
New x86
Platform
(2014)
Source: Capacity based on IBM Sizing of typical SPECint_rate landscape and 3rd party analysis of system utilization.
Ivy Bridge Configuration

69 2-socket x86 servers

Xeon E5-2697 v2 processor

24 cores per server, 2 threads
per core

2.8 GHz

1656 total cores

VMware
Per-core SW Costs INCREASE 34%!!!
來看看建置雲端IaaS的TCA & SW Cost:
採用P8 vs X86/VMware 誰省???
POWER8 delivers lower cloud infrastructure costs
34 2-socket Power S822L servers do the job of 100 2-socket x86 (HP DL380) servers
running equal virtualized capacity
58%lower total HW TCA
66% Less Systems & Cores
vs..
Ivy Bridge w/ VMware
Lower SW License Fees
Reduced Management Costs
Reduced Floor Space
Total HW TCA
Power S822L
 TCA/TCO is for 34 servers,
816 cores
 3x better virtualized throughput
vs.. an HP 2 socket Ivy Bridge
 2S, 24 cores each
 POWER8, 3.0GHz
 PowerVM
HP DL 380 G8
$4,000,000
$2,000,000
$2,951,257
$1,307,776
$S822L/24c
HP DL380p
IvB (2s)
HW TCA
 TCA/TCO is for 100 servers,
2400 cores
 100 HP servers needed for ~
equal virtualized throughput of 34
Power S822L
 2S, 24 cores each
 Ivy Bridge, 2.7GHz
 VMware vSphere Ent
Source: Capacity based on IBM Sizing of typical SPECint_rate landscape and 3rd party analysis of system utilization. Pricing from www.hp.com.
This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment. IBM's internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As
such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on actual configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are
based on published US list prices for both IBM and competitor.
來看看建置Java 應用伺服器性價比:
採用P8 vs X86/VMware 誰省??? (10台更省,愈多愈省)
POWER8 based systems have price-performance leadership versus x86
Price-Performance (EjOPS/$)
Power S822L




0.6
40% Better Performance/$
with
Power S822L
vs..
HP DL380p G8
0.5
0.4
0.3
0.56
0.4
0.2
HP DL 380p G8
0.1
0
Power S822L
Source: Capacity based on IBM sizing of typical Java based enterprise landscape
Competitive Pricing from www.hp.com
1 servers, 24 cores
2S, 24 cores per server
POWER8, 3.0GHz
96 GB Memory
HP DL380p IvB (2s)




1 servers, 24 cores
2S, 24 cores per server
Ivy Bridge, 2.7GHz
96 GB Memory
來看看建置DB2 DB ERP伺服器TCA & TCO成本:
採用P8 vs X86/VMware 誰省???
Save nearly $1M in your DB2 DB ERP environment with POWER8 vs. x86
One 2-socket Power S824 server does the job of four 2-socket HP DL380p Gen 8servers running equal virtualized ERP
capacity with a DB2 Database
Total TCA
60%
lower total TCA
Vs.
HP Ivy Bridge w/ VMware
$1,200,000
$1,000,000
$800,000
$600,000
$400,000
$200,000
$-
Power S824
$909,216
SW TCA
3-yr. TCO savings
Vs.
HP Ivy Bridge w/ VMware
$118,050
$85,431
S824/24c
$993,296
HW TCA
$227,304
HP DL380p IvB
(2s)
 TCA/TCO is for one servers, 24
cores
 4x better virtualized throughput vs.
an HP 2-socket Ivy Bridge
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 DB2 Database
3-yr TCO
HP DL 380 G8
$1,500,000
$1,000,000
$1,405,953
3-yr TCO
$500,000
$412,657
$S824/24c
HP DL380p IvB
(2s)
 TCA/TCO is for four servers, 96
cores
 4 HP servers needed for ~ equal
ERP throughput of 1 Power S824
 2S, 24 cores each
 Ivy Bridge, 2.7GHz
 VMware vSphere Ent.
 DB2 Database
Source: Capacity based on IBM Sizing of typical ERP landscape and 3rd party analysis of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment. IBM's
internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on actual
configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor.
來看看建置Oracle DB ERP伺服器TCA & TCO成本:
採用P8 vs X86/VMware 誰省???
Save $500K in your Oracle DB ERP environment with POWER8 vs. x86
One 2-socket Power S824 server does the job of four 2-socket HP DL380p Gen8 servers running equal
virtualized ERP capacity with a Oracle EE Database
Total TCA
47%
lower total TCA
Vs.
HP Ivy Bridge w/ VMware
Power S824
$1,000,000
$800,000
$600,000
$400,000
$200,000
$684,000
SW TCA
$342,000
HW TCA
$118,050
$85,431
$S824/24c
$531,099
3-yr. TCO savings
Vs.
HP Ivy Bridge w/ VMware
HP DL380p IvB
(2s)
 TCA/TCO is for one servers, 24
cores
 4x better virtualized throughput vs.
an HP 2-socket Ivy Bridge
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 Oracle EE Database
3-yr TCO
HP DL 380 G8
$1,200,000
$1,000,000
$800,000
$600,000
$400,000
$200,000
$-
$1,118,010
3-yr TCO
$586,911
S824/24c
HP DL380p IvB
(2s)
 TCA/TCO is for four servers, 96
cores
 4 HP servers needed for ~ equal
ERP throughput of 1 Power S824
 2S, 24 cores each
 Ivy Bridge, 2.7GHz
 VMware vSphere Ent.
 Oracle EE Database
Source: Capacity based on IBM Sizing of typical ERP landscape and 3rd party analysis of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment. IBM's
internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on actual
configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor.
來看看建置DB2 vs Oracle DB ERP伺服器TCA & TCO成本:
採用P8 vs X86/VMWare 誰省???
Save $700K using DB2 in your ERP environment with POWER8 vs. Oracle DB with x86
One 2-socket Power S824 server with DB2 Database does the job of four 2-socket HP DL380p Gen8 Servers with Oracle EE
Database running equal virtualized ERP capacity
Total TCA
61%
Power S824
$1,000,000
$800,000
lower total TCA
vs..
HP Ivy Bridge w/ VMware
$600,000
$400,000
$200,000
$684,000
SW TCA
$342,000
$S824/24c
$705,353
3-yr. TCO savings
vs..
HP Ivy Bridge w/ VMware
HW TCA
$118,050
$85,431
HP DL380p IvB
(2s)
 TCA/TCO is for one servers, 24
cores
 4x better virtualized throughput vs.
an HP 2-socket Ivy Bridge
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 DB2 Database
3-yr TCO
$1,200,000
$1,000,000
$800,000
$600,000
$400,000
$200,000
$-
HP DL 380 G8
$1,118,010
3-yr TCO
$586,911
S824/24c
HP DL380p IvB
(2s)
 TCA/TCO is for four servers, 96
cores
 4 HP servers needed for ~ equal
ERP throughput of 1 Power S824
 2S, 24 cores each
 Ivy Bridge, 2.7GHz
 VMware vSphere Ent.
 Oracle EE Database
Source: Capacity based on IBM Sizing of typical ERP landscape and 3rd party analysis of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment. IBM's
internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on actual
configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor.
來看看建置WAS vs Weblogic App伺服器TCA & TCO成本:
採用P8 vs X86/VMWare 誰省???
POWER8 / WAS drives down costs when compared with x86 / Weblogic
One 2-socket Power S824 server with WAS running equal virtualized Java-based applications as four Oracle
Sun X4-2 servers with Weblogic SE
Total TCA
45%
lower HW TCA
vs..
Oracle Ivy Bridge w/
VMware
Power S824
$400,000
$300,000
$144,000
$200,000
$100,000
SW TCA
$115,920
HW TCA
$157,517
$85,431
$S824/24c
Oracle X4-2 IvB (2s)
 TCA/TCO is for one server, 24
cores
 4x better virtualized throughput vs.
an Oracle Sun 2-socket Ivy Bridge
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 IBM WAS
3-yr TCO
$123,158
3-yr. TCO savings
vs..
Oracle Ivy Bridge w/
VMware
Oracle Sun X4-2
$400,000
$300,000
$379,877
$200,000
$256,719
$100,000
$S824/24c
Oracle X4-2 IvB (2s)
3-yr TCO
 TCA/TCO is for four servers, 96
cores
 4 Oracle servers needed for ~
equal Java throughput of 1 Power
S824
 2S, 24 cores each
 Ivy Bridge, 2.7GHz
 VMware vSphere Ent.
 Oracle Weblogic SE
Source: Capacity based on IBM Sizing of typical SPECjEnterprise2010 and IBM estimates of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment. IBM's
internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on actual
configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor. Pricing from www.oracle.com
來看看建置WAS vs Weblogic APP伺服器TCA & TCO成本:
採用P8 vs Oracle T5 誰省???
POWER8 / WAS drives down costs when compared with T5 / Weblogic
Two 2-socket Power S824 servers with WAS running equal virtualized Java-based applications as four T5-2
Servers with Weblogic SE
Total TCA
34%
lower HW TCA
vs..
Oracle T5-2 w/ OVM
Power S824
$500,000
$400,000
$300,000
$192,000
$231,840
SW TCA
$200,000
$100,000
$259,372
HW TCA
$170,862
 TCA/TCO is for two servers, 48
cores
 2x better virtualized throughput vs..
an Oracle T5-2
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 IBM WAS
$S824/24c
3-yr TCO
$40,414
3-yr. TCO savings
vs..
Oracle T5-2 w/ OVM
T5-2
Oracle T5-2
$560,000
$540,000
$553,852
$520,000
$500,000
$513,438
$480,000
S824/24c
T5-2
3-yr TCO
 TCA/TCO is for four servers,
128 cores
 4 Oracle servers needed for ~
equal Java throughput of 2 Power
S824
 2S, 16 cores each
 T5, 3,6GHz
 Oracle OVM
 Oracle Weblogic SE
Source: Capacity based on IBM Sizing of typical SPECjEnterprise2010 and IBM estimates of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer
environment. IBM's internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may
vary based on actual configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor. Pricing from www.oracle.com
解決巨量資料,建置Big Data/Hadoop, 你要左邊, 還是…...右邊
“In-Memory” NoSQL/KVS Infrastructure via CAPI-attached Flash
12TB, 6K users, 2GB/User Example – Power Systems delivers 24:1 consolidation over x861
WWW
WWW
Today (x86)
10Gb Uplink
Load Balancer
10Gb Uplink
POWER8 + CAPI Flash
POWER8 4u Server
2u Flash w/ 12TB
500GB Cache Node
24 nodes
500GB Cache Node
500GB Cache Node
500GB Cache Node
500GB Cache Node
500GB Cache Node
Backup storage req’d for data
persistence
Backup Node
26 servers, switches, etc = >1 rack
In-memory database w/ 12TB DRAM
~18kW
1)
2)
4u, 1.5kW
In-memory database with 12TB Flash
Sizings
Power + CAPI Flash Advantage
• 24:1 physical server consolidation
• >1 rack to 4U Density (>12x)
• 18kW 1.5kW Energy (12x)
• $24/user $7.5/user Cost (3.2x)
• 4.5k 60k Users per rack (13x)2
24:1 system consolidation ratio (12:1 rack density improvement) based on a single IBM S824, (24 cores, POWER8 3.5 GHz), 256GB RAM, AIX 7.1 with 40 TB memory based Flash replacing 24 HP DL380p, 24 cores, E52697 v2 2.7 GHz), 256GB RAM, SuSE Linux 11SP3 . Inbound network limits performance to 1M IOPs in both scenarios, equal capacity (#user, data) in both cases. x86 cost includes 10k$ for 2x 1U switches
Normalized to a single rack
總結: POWER8 為巨量資料和雲而生
POWER8 scale-out systems offer compelling reasons
to deploy in Big Data & Analytics, Mobile/Java and Cloud
Get insights up to 50X faster with
P8 systems designed for Big Data
and BLU
24:1 server footprint reduction
POWER8
Scale out Systems
through CAPI Flash for BD&A – no
SQL
and compression for BD&A - Hadoop
Power S822L
5X Watson scaling through
Power S824L
Software Optimization and
POWER8 for Cognitive
Power S822
10X ops/sec via Mellanox Fabric
exploitation for Data-centric
workloads
4X reduction in storage through GPFS
Non-disruptive
Application mobility from P6/P7 to
P8
65% Guaranteed system utilization
Power S824
8X IBM Java advantage through
Nvidia GPU acceleration
without application degradation
Open Innovation to Put Data to Work
POP03497-USEN-01
Intel: Reduce HW infrastructure in a non-virtualized environment but increase per-core SW costs…
POWER8: Reduce HW infrastructure EVEN MORE and REDUCE per-core SW costs up to 74%
Current Configuration

100 2-socket x86 servers

Xeon X5690 processor

12 cores per server, 2
threads per core

3.46 GHz

1200 total cores

No VMware
Power S822L Configuration

13 2-socket POWER8 servers

POWER8 processor

24 cores per server, 8 threads
per core

3.0 GHz

312 total cores
POWER8
Platform
(2014)
Current x86
Platform
(circa 2011)
1.5X
increased
throughput
Per-core SW Costs DECREASE up to
74%!!!
POWER8 DELIVERS EQUAL
CAPACITY with
1/5 of the Servers (LOWER MANAGEMENT COSTS)
1/5 of the Cores (LOWER SW COSTS)
1/5 of the Space (LOWER INFRASTRUCTURE COSTS)
ALL for <33% of the x86 HW TCA
New x86
Platform
(2014)
Source: Capacity based on IBM Sizing of typical SPECint_rate landscape and 3rd party analysis of system utilization.
Ivy Bridge Configuration

69 2-socket x86 servers

Xeon E5-2697 v2 processor

24 cores per server, 2 threads
per core

2.8 GHz

1656 total cores

No VMware
Per-core SW Costs INCREASE 34%!!!
Save nearly $5M in your DB2 DB ERP environment with POWER8 vs. x86
Three 2-socket Power S824 servers do the job of five 4-socket HP DL580 Gen8 servers running equal virtualized ERP
capacity with DB2 Database
Total TCA
79%
lower total TCA
vs..
HP Ivy Bridge w/ VMware
$5,000,000
Power S824
$4,000,000
 TCA/TCO is for three servers, 72
cores
 1.67x better virtualized throughput
vs. an HP 4 socket Ivy Bridge
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 DB2 Database
$3,000,000
$4,059,000
SW TCA
$2,000,000
HW TCA
$1,000,000
$681,912
$256,293
$319,051
$S824/24c
HP DL580/E74890 v2 (4s)
3-yr TCO
$4,808,682
3-yr. TCO savings
vs..
HP Ivy Bridge w/ VMware
HP DL 380 G8
$8,000,000
$6,000,000
$4,000,000
$2,000,000
$6,046,651
$1,237,970
$S824/24c
HP DL580/E74890 v2 (4s)
3-yr TCO
 TCA/TCO is for five servers,
300 cores
 5 HP servers needed for ~ equal
ERP throughput of 3 Power S824
 4S, 60 cores each
 Ivy Bridge, 2.8GHz
 VMware vSphere Ent.
 DB2 Database
Source: Capacity based on IBM Sizing of typical ERP landscape and 3rd party analysis of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment. IBM's
internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on actual
configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor.
POWER8 delivers 60% more ERP performance at 36% less HW TCA than T5-8
Two 2-socket Power S824 servers with Oracle DB ERP as a T5-8 Server have lower TCA and TCO while
delivering better performance
Total TCA
30%
Lower HW TCA
Power S824
$1,200,000
$1,000,000
$800,000
$600,000
$912,000
$684,000
SW TCA
HW TCA
$400,000
26%
$200,000
$170,862
 TCA/TCO is for two servers, 48
cores
 2S, 24 cores each
 POWER8, 3.5GHz
 PowerVM
 Oracle EE
$242,643
$S824/24c
T5-8
Lower Total TCA
3-yr TCO
$392,101
3-yr. TCO savings
vs..
Oracle T5-8
Oracle T5-8
$2,000,000
$1,500,000
$1,000,000
$1,565,923
$1,173,822
$500,000
3-yr TCO
 TCA/TCO is for 1 server, 128
cores
 1 Oracle servers
 8S, 32 cores each
 T5, 3,6GHz
 Oracle OVM
 Oracle EE
$S824/24c
T5-8
Source: Capacity based on IBM Sizing of typical ERP landscape and IBM estimates of system utilization. This is an IBM sizing designed to replicate a typical IBM customer workload used in the marketplace. The results are calculated and not an actual customer environment.
IBM's internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other systems variations or conditions may produce different results and may vary based on
actual configuration, applications, specific queries and other variables in a production environment. Prices, where applicabl e, are based on published US list prices for both IBM and competitor. Pricing from www.oracle.com
讓POWER8 釋放
Reduce batch window requirements
 85% better response time and 1.7x more
throughput with POWER8 (Single Thread
mode) than POWER7+ (SMT4)
4
90
3.5
80
70
3
60
2.5
50
2
40
1.5
30
1
20
0.5
10
0
0
P7+ SMT4
P8 ST
P8 SMT2
Relative Capacity
1
Based on internal IBM batch processing benchmark running on Power 750+ at 3.4GHz and Power S824 at 4.1GHz
P8 SMT4
Run Tim e
P8 SMT8
Response Time (Seconds)
 3.5x more throughput and 44% better
response time with POWER8 (SMT8) than
POWER7+ (SMT4)
Relative Throughput
Up to 3.5x better core batch throughput performance at 44% better response time than P7+
Download