Linux Performance on IBM Power Systems How an 18-year Partnership Eases Conversion, Boosts Capabilities, and Lower Costs

IBM and Linux: Building the Future of Computing

IBM has a long history with Linux and promoting open-source operating systems. As far back as 1999, IBM issued their support, vowing to Linux-enable all of their hardware platforms. At the time, this was a surprising announcement. There was little interest in Linux from enterprise customers. As an open source operating system, it was mostly regarded as a threat to the revenue streams of major IT vendors such as Microsoft, Dell, and even IBM itself.

Linux was developed not by a company, but by thousands of programmers working in collaboration for free and sharing the software for free. IBM took a big leap and looked beyond potential software revenue streams. It planted its flag with Linux, over time making a total investment in Linux adoption at a price of US$1 billion.1

Linux struggled to gain mainstream traction in the early 2000s. Trends leaned to the old Unix market, where Unix versions were proprietary and maintained by specific hardware vendors. What IBM saw in its investment in Linux was the ability to run high-volume servers. In addition, young programmers were drawn to the open-source nature of Linux, which dovetailed nicely with the growing Open Web movement. The next wave of IT talent began to embrace Linux.

Jumping forward a decade, Linux is everywhere, from smartphones to gaming consoles to supercomputers. There’s a version of Linux that will run pretty much anything. However, the primary reason enterprise companies buy servers is to solve business problems, and Linux must fit into that equation. IBM has made the investment in Linux to ensure that each of their system and server families have a wide range of applications, middleware, and management software to give companies many options to choose from.2 IBM has benefited from this choice, particularly with mainframes at first and now with their Power Systems servers. 2

IBM’s Move to Be the Primary Enterprise Solution

Today, the growth in corporate data center environments is driven by architecture systems that offer ease of management, processing power, and affordability. IBM Power Systems’ scale-out servers are high powered, energy efficient, affordable, and easy to deploy. They deliver improved scale-out scenarios while occupying less rack space. In addition, with the OpenPower Initiative, a partnership that began with IBM, Google, Mellanox, NVIDIA, and Tyan, IBM is able to give customers greater access to a wide range of data technology choices, along with more flexibility to customize systems.3

In the spirit of Linux, IBM is sharing Power-based server processor specs, firmware, and software with partners. These developments are driving the global growth of IBM POWER8-based systems. With the ability to easily integrate IBM Power servers into existing ecosystems, run virtually any operating system including Linux, and easily migrate applications and capabilities to the cloud, IBM Power Systems are gaining ground on their competitors.

IBM is also looking to surpass Intel in server chip processing speeds and capabilities with the speed of POWER8 multiprocessors and the upcoming January 2017 release of POWER9.4 Additionally, with PowerKVM virtualization, businesses that choose IBM Power Systems avoid vendor lock-in and gain the flexibility to quickly integrate innovative technology solutions into their existing infrastructure.

The following chart from IBM provides a look at the open-source ecosystem of IBM Power Systems:5

Open Source Ecosystem Available on IBM Power

Cloud Other HA, Dev. Env Databases Big Data echnical ools Analytics Management Security Computing Stac etc.

Available Available Available Available Available Available Backbone, Bootstrap, Accumulo column, Hadoop Core, Apache eb Server BTFS ALLPATH-LG, Bedtools, Docker, Elgen lib Cassandra Hive, HBase, Accumolo, Bootstrap bfast, Bio Conductor, Erlang, CouchDB document Ambari, Avro, Falcon, Ceph, Chef server Chroma-key BioConductor-base, Ganglia, GCC, GDB, Derby Flume, Hue, Knox, Juju Juju gui Cluster Glue Blast, BOOST, Bowtie, Jenkins, Jruby, MariaDB v10 optimied Lucene-Solr, Mahout, Landscape client DBD Bowtie2, BA, bip2, keepalived, LLVM, Memcached KVS Oole, Paruet, MAAS, OpenStack Evolution data svr Cufflinks -2.2.1, FASTA, Lucene, Maven, Nagios, MongoDB document Phoenix, Pig, Soop, Puppet HAProxy FastC, HMM, HGIN, node.js, MySL Storm Te, ookeeper Apache pid Heartbeat HTSe, LibGDpartial, OpenJDK, PHP, Postgre SL Thrift keepalived libpng, Mothur, nose, phpMyAdmin, , abbitM Ldirectord NumPy, OpenSSL, Python, Python-Django, edis KVS OpenSSL PICAD,PLINK, Python-Pip ecosystem, SLite Pacemaker Python, SAMTools, , rsyslog, uby, uby Virtuoso graph samba SAMTools 1.0, SeAN, on ails rbenv, uby Tophat setuptoolsPyhton, Gems, scala, snappy, ordPress SHIMP, SOAP3-DP, Socket.io npmjs, SOAPDenovo tabix, SystemTap, Vagrant, TopHat, Trinity, Port in Progress Port in Progress Port in Progress Port in Progress V8, wireshark VelvetOases, lib Voldemort KVS Spark Ceilometer client Port in Progress Neo4J graph ABySS, Balsa, GoLong, gccgo Bioconductor, GMP, Pubsub.io 3 GOMACS, NAMD, Logstash spice, uantum Espresso

Port in Progress IGV, iODS, NAStar, ISAAC, SOAPAligner 3

Supported Linux Distributors of IBM Power Systems

There are three main distributors for IBM Power Systems: Ubuntu, SUSE, and . Due to the flavor preferences of developers, it’s necessary to offer not only a full portfolio of open-source software, but also the specific distributions that are in demand today. Ubuntu, SUSE, and Red Hat each have their various strengths.

Choosing Like Linux, Ubuntu started with free distribution. Under the umbrella or Canonical, it grew to offer professional services through its distribution. Ubuntu for IBM Power Systems performs well for modern workloads. Certain Ubuntu developers have also shown a preference for Ubuntu, particularly those building early container-based applications. Additionally, Ubuntu attracts developers due to its early use with OpenStack private clouds.6

Ubuntu and IBM Power’s benefits include offering scalability comparative to public cloud infrastructure in a pay- for-consumption model. Users are able to incorporate next-gen applications in a system that’s reliable, secure, and, with POWER8 processors, fast and efficient. One point to note is the need to install the Long Term Support version of Ubuntu when using it as a server platform. This is due to Ubuntu offering enterprise support and patch systems that are available for at least 7 years.

Choosing SUSE and IBM have been working together since IBM first embraced Linux in 1999. The SUSE Linux Enterprise Server was the first Linux distribution for SAP HANA to run on Power Systems servers.7 The current solution offers SUSE tailored data center integration, giving users the option of preselected configurations that work with existing resources, as well as the opportunity to modify the installation to match custom environments.

IBM POWER8 supports both big and little-endian modes. On little endian, Power Systems with POWER8 processing help simplify source code porting from x86 Linux applications to little-endian Linux on Power, allowing for easier and faster access to Linux applications.8 In addition, IBM Power Systems and SAP HANA on SUSE provide up to 1,536 threads per system, 16 TB of memory, 224 MB cache per socket, and 230 GB per second of sustained memory bandwidth.9

Choosing As the dominant player when estimating market share of enterprise distributions, Red Hat employs more developers than other Linux-supported vendors. There has been a long commitment between IBM and Red Hat Red Hat to build flexible infrastructure based on open hardware and software solutions, notably Red Hat Enterprise Virtualization and Red Hat Cloud.

Red Hat Virtualization allows for more resource-intensive workloads to run on fewer servers and is a popular choice for Linux workloads. Red Hat Cloud lets users build and manage an open, private IaaS cloud at a competitive cost. Together, Red Hat and IBM Power Systems provide enhanced capabilities to companies wanting to improve scale-out big data and cloud deployments for a broader range of enterprise applications. 4

Linux on Power Performance Advantages

Regardless of which distribution channel is used, adopting Linux on Power provides performance advantages for a variety of workloads, including data collection that translates to new revenue streams. IBM Power Systems with POWER8 processing speeds allow for a customizable infrastructure solution that integrates software for big data and analytics. IBM Data Engine is ideal for workloads involving Open Platform and Spark. These two frameworks allow companies to optimize storage and computing resources, while IBM InfoSphere® Streams™ handles high-ingest streaming analytics.

With the explosion of NoSQL and the challenges of cost and scalability, IBM Data Engine for NoSQL creates a new tier of memory, attaching up to 40 TB of auxiliary flash memory to POWER8 systems without the latency issues of traditional storage I/O.10 The auxiliary flash memory is attached by way of the POWER8’s Coherent Accelerator Processing Interface (CAPI). This setup provides a scalable, hybrid solution that delivers performance well beyond I/0 attached acceleration engines. As flash memory continues to drop in price—now being 15x less than the cost of DRAM memory—cost savings are realized as well.11

Linux on Power Cloud Manager with OpenStack leverages Power’s enterprise security and reliability to deliver cloud virtualization without risk. OpenStack manages PowerKVM environments directly. It increases efficiency through heterogeneous cloud management across Power Systems, System z, and x86 environments. Other features include a self-service portal for the provisioning of services, virtualized image management, virtual image oversight, automated approval processing and provisioning, and basic usage metering supporting pay- per-use business models.

IBM Watson is also built on Linux on Power. With the high processing speeds of POWER8—and the January 2017 release of POWER9— Power Systems built on Linux on Power allows Watson to use a set of transformational technologies that leverage natural language, hypothesis generation, and evidence-based learning to cognitively “think” and problem solve. The end result is deep content analysis and evidence-based reasoning to accelerate and improve decisions and optimize business outcomes.

In addition, Linux on Power drives high-performance computing. Configurable into highly scalable Linux clusters, Power Systems for high performance computing are deployed in demanding workloads like genomics, finance, oil and gas exploration, and performance data analytics. High-performance servers function above a teraflop, or 1012 floating-point operations a second. Job orchestration and automatic storage policies enable fast storage that is tightly coupled to the compute node as a managed cache, which preserves disk space and bandwidth. Smart caching optimizes storage resources and bandwidth for improved processing performance without the need for additional investments in infrastructure.12 Converting to Linux on Power

In most cases, porting Linux applications from the x86 platform to Linux on Power is a simple endeavor. By focusing on porting first and optimizing later, the difficulty of porting from Intel to Linux on Power can be greatly minimized. Both platforms are based on a common Linux version. Porting only requires a GNU Compiler Collection (GCC) with some minor changes to some compiler and linker switches.

Advanced tools are also available for free to help the conversion. These tools include the IBM Software Development Kit for PowerLinux (SDK), which integrates C/C++ source development with Advance Toolchain, post-link optimization, and standard Linux performance tools, including OProfile, Perf, and Valgrind.

In addition, IBM Power Systems Linux Centers, IBM Innovation Centers and IBM Client Centers provide a global network of resources to help developers build and deploy custom applications. IBM Innovation Centers offer both on-site and remote support for building and testing. Free Linux workshops are also available to increase skills and knowledge. Resources Overview

• Many resources are widely available for additional support with IBM Power Systems. • Tailored customer briefings provide details about using Power Systems and to deploying Linux applications. • Linux training workshops teach users how to program, port, and optimize applications using Ubuntu, SUSE, and Red Hat. • Developers can get hands-on assistance when optimizing POWER8 parallel processing and Power Systems’ advanced virtualization capabilities. • Customized business development and technical workshops highlight how Power technology can help grow business.

For more information, contact ABC Services at 888.505.6222 or email us at [email protected]

1 IBM | http://www-03.ibm.com/ibm/history/ibm100/us/en/icons/linux/ 2 IBM | ftp://public.dhe.ibm.com/linux/pdfs/GCG_IBM_and_Linux-9_years_later.pdf 3 IBM | https://www.ibm.com/blogs/think/2015/10/16/ibm-open-power/ 4 The Motley Fool | http://www.fool.com/investing/general/2015/11/24/ibm-vs-intel-corporation-the-data-center-battle-es.aspx 5 IBM 6 IBM | http://www-01.ibm.com/common/ssi/cgi-bin/ssialias?htmlfid=LUW12345USEN 7 SUSE | https://www.suse.com/docrep/documents/kgu61iyowz/suse_and_ibm_power_best_for_demanding_workloads_and_fastest_sap_hana_server_flyer.pdf 8 SUSE | https://www.suse.com/docrep/documents/jf1q7ngmde/sles_12_on_ibm_power8_faq.pdf 9 IBM Redbooks | http://www.redbooks.ibm.com/redpapers/pdfs/redp5248.pdf 10 Brech, Brad; Hollinger, Michael; Rubio, Juan. United States. IBM Systems and Technology Group. Data Engine for NoSQL – IBM Power Systems™ Edition. White Paper; October2014 11 Stack Exchange | http://electronics.stackexchange.com/questions/210576/why-dram-costs-much-more-than-flash-memory 12 United States. IBM Spectrum Computing. Fast results and less investment with data-centric HPC – Reduce costs and speed results with smart, data-aware solutions. White Paper. August 2016