page-brochureware.php

IBM Power Systems

Cloud performance proof-points

Run swiftly and smoothly on a high performing global infrastructure. Deploy global cloud infrastructure rapidly on virtual servers.

Open source Docker with IBM WebSphere Application Server Liberty

For the systems and workload compared:

IBM POWER8 delivers improved container density at lower solution price. IBM WebSphere Application Server (WAS) Liberty V9 on IBM Power System S822LC for Big Data with open source Docker delivers 1.57X more containers and 1.96X better price-performance than Intel Xeon E5-2650 v4 Broadwell.

  • Container density – 57% more containers per server
  • Server capacity – 45% more aggregate throughput
  • Reduced infrastructure costs – 1.96X better price-performance


image617


image618


image619

System configuration

IBM Power System S822LC for Big Data Competitor: HP Proliant DL380
20 cores / 160 threads, POWER8 24 cores / 48 threads, Xeon E5-2650 v4
2.92 GHz, kernel 4.4.0-12-generic processor frequency governor of performance, and hardware prefetch disabled 2.2 GHz processor frequency governor of performance, and hardware prefetch disabled.
Ubuntu 16.04, had 512 GB memory, included 1 TB SATA 7200 rpm HDD, 10 Gb 4-port, one 16 Gbps FCA Ubuntu 16.04, had 512 GB memory, included 1 TB SATA 7200 rpm HDD, 10 Gb 4-port, one 16 Gbps FCA
WebSphere Application Server V9.0 Liberty profile; WebSphere Application Server V9.0 Liberty profile;
Java options: -Xmx512m, -Xms512m, -Xgcthreads8, -Xnocompactgc, -Xnoclassgc, -Xconcurrentlevel0, -Xdisableexplicitgc Java options: -Xmx512m, -Xms512m, -Xgcthreads8, -Xnocompactgc, -Xnoclassgc, -Xconcurrentlevel0, -Xdisableexplicitgc
Open source Docker version: 1.12.0 / API: 1.24 / Go: 1.6.3 Open source Docker version: 1.12.0 / API: 1.24 / Go: 1.6.3
Docker storage driver: overlay2 Docker storage driver: overlay2

Notes:

  • Results are based on IBM internal testing of a single system and OS image running with Acme Air workload (https://github.com/acmeair) on a private network with a dedicated JMeter driver machine and dedicated MongoDB server machine: One MongoDB instance per 8 WebSphere Application Server containers.
  • Each WebSphere Application Server container was bound to a full core to run with 20 users and a 25 ms think time between transactions. The number of containers were increased for each system until average throughput dropped below 400 transactions/second or latency exceeded 25 ms.
  • Tests were run on 29-Nov-2016. Individual results might vary depending on individual workloads, configurations, and conditions.
  • Pricing is based on: Power S822LC for Big Data (https://www.ibm.com/systems/power/hardware/linux-lc.html) and HP DL380 (https://h22174.www2.hp.com/SimplifiedConfig/Index) on 07-Dec-2016.
  • Pricing for WebSphere Application Server is based on 20% discount from Passport Advantage pricing (https://www.ibm.com/software/passportadvantage/pao_customer.html) on 07-Dec-2016

MongoDB YCSB POWER8 leadership

For the systems and workload compared:

  • IBM Power® System S822LC for Big Data delivers 1.68x price performance leadership over Intel Xeon E5-2690 v4 running on cloud workload.
MongoDB YCSB POWER8 leadership

System configuration

Power S822LC for Big Data Xeon E5-2690 v4
20-c Power S822LC for Big Data 28-c HP DL380
2.9 GHz IBM POWER8®, 160 threads 2.6 GHz Intel Xeon® E5-2690 v4, 56 threads
128 GB memory 128 GB memory
(1) 2-port 10 GB Ethernet (1) 2-port 10 GB Ethernet
(1) 2-port 16 Gbps FCA (1) 2-port 16 Gbps FCA
(2) 1 TB SATA LFF HDD (2) 300 GB SATA LFF HDD
MongoDB 3.3.8 MongoDB 3.3.8
RHEL 7.2 RHEL 7.2

Notes:

  • Based on IBM internal testing of single system and OS image running Yahoo Cloud Services Benchmark (YCSB) 0.6.0, 1M record workload at 50/50 read/write factor. Results valid as of 8/24/16. Power8 total system throughput produced 288,824 ops and Xeon E5-2690 produced 205,951 ops. Tests conducted under laboratory condition; individual results can vary based on workload size, use of storage subsystems & other conditions.
  • Configurations represent the processor running in favor performance mode and the MongoDB server on 1 socket & the YCSB application workload on the 2nd socket. IBM Flash 900 storage was used on both servers for testing.
  • Pricing is based on:

Open Source Docker

For the systems and workload compared:

  • IBM Power System S824 delivers better performance than the compared Haswell for 12 cores with
    • Up to 42% better throughput
    • Up to 4X better latency
    • 2X container density
Docker graph comparing POWER8 to Xeon E5 V3

System configuration

IBM Power System S824 Xeon E5 V3 Haswell
Hardware Hardware
3.52 GHz, POWER8 split-core with SMT-ST mode, cpufreq governor=performance, 256 GB, PowerKVM 2.1.1, Storage: V7000 HDD 2.3 GHz, Haswell hyper threading enabled, cpufreq governor=performance, 768 GB, RHEL 7 with KVM, Storage: V7000 HDD
Software Software
IBM SDK v1.2, Node v0.12.1, Open Source Docker 1.6 Node v0.12.1, Open Source Docker 1.6.2

Notes:

  • The Acme Air web-based reservation system was used for testing. Acme Air: https://acmeair.github.io/acmeair/
  • The results occurred in a single KVM guest with the following characteristics:
    • Started with 20 high utilization (H) containers that support 20 users with no think time between transactions and represent 10% of active users and 192 low utilization (L) containers that support 20 users with 10 seconds of think time between transactions and represent 90% of active containers.
    • Increased density to 2X the initial container mix, while maintaining a 10L to 1H ratio.
  • Guest details: Ubuntu 15.04, 12 cores with 120 GB bound guest via XML file to single socket, virtio Ethernet with multi-queue=2,
  • Private network switch was used along with dedicated JMETER driver machine and dedicated MongoDB server machine: One MongoDB instance per 10H containers and one MongoDB instance per 50L containers.
  • Test completed November 12, 2015.

References:
IBM POWER8 with Docker Delivers Superior Cloud Performance (PDF, 482 KB)

Virtualized EnterpriseDB Postgres Advanced Server 9.5

For the systems and workload compared:

  • IBM Power® System S822LC for Big Data delivers 2.12x price performance leadership over Intel Xeon E5-2699 v4.
Virtualized EnterpriseDB Postgres Advanced Server 9.5

System configuration

Power S822LC for Big Data Xeon E5-2699 v4
20-c Power S822LC for Big Data 44-c HP DL380
2.9 GHz IBM POWER8®, 160 threads 2.2 GHz Intel Xeon® E5-2699 v4, 88 threads
256 GB memory, 22 VMs 256 GB memory, 22 VMs
(1) 2-port 10 GB Ethernet (1) 2-port 10 GB Ethernet
(1) 2-port 16 Gbps FCA (1) 2-port 16 Gbps FCA
(2) 1 TB SATA LFF HDD (2) 300 GB SATA LFF HDD
EDB Postgres Advanced Server® 9.5 EDB Postgres Advanced Server® 9.5
RHEL 7.2 Guests with KVM RHEL 7.2 Guests with KVM

Notes:

Virtualized MariaDB 10.1 Sysbench 1.0 performance

For the systems and workload compared:

  • IBM Power® System S822LC for Big Data delivers 1.84x price performance leadership per virtual machine over Intel Xeon E5-2690 v4 running on sysbench workload.
  • Power S822LC for Big Data delivers 1.74x price performance leadership per transaction over Intel Xeon E5-2690 v4 running on sysbench workload.
  • Power S822LC for Big Data delivers 42% more virtual machines per server than a Xeon 28-core E5-2690 v4 system running the sysbench workload.
  • Power S822LC for Big Data delivers 35% more transactions per second per server than a Xeon 28-core E5-2690 v4 system running the sysbench workload.
Virtualized MariaDB 10.1 Sysbench 1.0 performance

System configuration

Power S822LC for Big Data Xeon E5-2690 v4
20-c Power S822LC for Big Data 28-c HP DL380
2.9 GHz IBM POWER8®, 160 threads 2.2 GHz Intel Xeon® E5-2690 v4, 56 threads
384 GB memory, 20 VMs 256 GB memory, 14 VMs
(1) 2-port 10 GB Ethernet (1) 2-port 10 GB Ethernet
(1) 2-port 16 Gbps FCA (1) 2-port 16 Gbps FCA
(2) 1 TB SATA LFF HDD (2) 300 GB SATA LFF HDD
MariaDB 10.1.16 MariaDB 10.1.16
Ubuntu 16.04 Guests with KVM Ubuntu 16.04 Guests with KVM

Notes:

  • Results are based on IBM internal testing of single system running multiple virtual machines with Sysbench read-only workload and are current as of August 22, 2016. Performance figures are based on running 24 M record scale factor per VM. Power8 total system throughput produced 15,903 tps and Xeon E5-2690 produced 11,727 tps. Individual results will vary depending on individual workloads, configurations and conditions.
  • Power S822LC for Big Data; 20 cores / 160 threads, POWER8; 2.9 GHz , 384GB memory MariaDB 10.1.16, 20 8vcpu VMs of Ubuntu 16.04 with KVM compared to competitive stack: HP Proliant DL380 28 cores / 56 threads; Intel E5-2690 v4, 2.6 GHz; 256 GB memory, MariaDB 10.1.16, 14 4vcpu VMs of Ubuntu 16.04. with KVM
  • Each system was configured to run at similar per VM throughput levels and the number of VMs were increased for each system until total system throughput showed maximum throughput levels.
  • Pricing is based on:

Cloud internal workload with LAMP

For the systems and workload compared:

  • IBM Power Systems S822L supports 1.59X more users per hour at 65% lower cost with comparable performance compared to x86 running a LAMP application stack from Canonical, Apache, MariaDB, and Zend
Cloud internal workload with LAMP graph comparing POWER8 against Xeon E5-2650 V3

System configuration

Power S822L Xeon E5-2650 V3 Haswell
Hardware Hardware
3.52 GHz, 8-cores, 64 threads, 128 GB, Ubuntu 14.04 2.0 GHz, 16-cores, 128 GB, Ubuntu 14.04
Software Software
Magento 1.4, Zend 8, MariaDB 10.1 Magento 1.4, Zend 8, MariaDB 10.1
  • Results based on testing from February 15-21, 2016 and are based on the Magento e-commerce benchmark.

IBM Corporation 2017®

IBM, the IBM logo, ibm.com, POWER and POWER8 are trademarks of the International Business Machines Corp., registered in many jurisdictions worldwide. Intel, Intel logo, Intel Inside, Intel Inside logo, Intel Centrino, Intel Centrino logo, Celeron, Intel Xeon, Intel SpeedStep, Itanium, and Pentium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. Linux is a registered trademark of Linus Torvalds in the United States, other countries, or both. Java and all Java-based trademarks and logos are trademarks or registered trademarks of Oracle and/or its affiliates. Other product and service names may be the trademarks of IBM or other companies.

The content in this document (including any pricing references) is current as of July 22, 2015 and may be changed by IBM at any time. Not all offerings are available in every country in which IBM operates

THE INFORMATION CONTAINED ON THIS WEBSITE IS PROVIDED ON AN “AS IS” BASIS WITHOUT ANY WARRANTY EXPRESS OR IMPLIED INCLUDING WITHOUT ANY WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE ANDY ANY WARRANTY OR CONDITION OF NON-INFRINGEMENT.

In no event will IBM be liable for damages arising directly or indirectly from any use of the information contained in this document.

All information contained on this website is subject to change without notice. The information contained in this website does not affect or change IBM product specifications or warranties. IBM’s products are warranted according to the terms and conditions of the agreements under which they are provided. Nothing in this website shall operate as an express or implied license or indemnity under the intellectual property rights of IBM or third parties.

All information contained on this website was obtained in specific environments, and is presented as an illustration. The results obtained in other operating environments may vary.

No licenses, expressed or implied, by estoppel or otherwise, to any intellectual property rights are granted by this website.