IBM’s investment in Linux
You've probably heard about IBM Watson. If not, take a look as it is truly a groundbreaking achievement in technology. To summarize some thoughts from an article in The Atlantic, IBM estimates that Watson can process up to 60 million pages of text per second. And this is just plain, simple written prose. It is completely unstructured, just like about 80% of all existing information that current IT systems need to process. The surprising part is that Watson can make sense of almost any freeform information source whether it is someone speaking or a dictation of someone's handwritten notes. It can actually learn, and quite quickly. First it was used to beat Jeopardy's best and now it is positioned to assist medical diagnosis for patients, revolutionize customer service, create new recipes for chefs, and a number of other interesting and previously insurmountable challenges. Some of these challenges are obviously critically important and others are simply fun and challenging applications that may well become highly strategic in the future. Watson's unique ability to adapt is head-shaking while the sheer amount of data that it can process is unmatched.
What is Watson when it comes to bits and bytes? Sure, it is a collection of clever software components with innovative algorithms but what is the underlying hardware infrastructure? You know it is an IBM computer, but specifically it is an IBM Power Systems™ server. And more specifically, it is running Linux on Power.
So the question becomes, if that combination can be used to create Watson, can it do what I need it to do?
To begin with, how serious is IBM about Linux on Power? IBM is serious enough to commit to invest $1 billion in it in September of 2013. This is an investment into solutions for Linux® and open source workloads and is in addition to numerous investments over the last decade on a wide range of open initiatives. The important thing to know is that it is not just an investment from IBM. It is an investment specifically from the Power Systems brand at IBM. However, it is important to note that money alone will not make the initiative successful. Skills from experienced resources are required and IBM has reassigned numerous, highly motivated sales and technical resources to focus exclusively on Linux and Power and the solutions that they can address together. IBM sells software too – lots of it. You can bet that the huge majority of that software runs on Linux on Power. You can further assume that these applications have been optimized to run quite efficiently on this platform.
Let's look at some specifics.
Every Power Systems server that IBM sells runs Linux. In fact, every server that IBM sells runs Linux, including Red Hat, SUSE and now Ubuntu. To be clear, it is the same Linux that runs on x86. It is built from the same source as x86, delivered on the same schedule and supported at the same time. What makes it better on Power? We'll discuss this in detail later, but it is the fundamental concepts such as performance, reliability, virtualization, and price. IBM thinks that these fundamental concepts are compelling advantages – hence the commitment.
In what other ways is IBM invested in Linux?
We have fully staffed Linux Technology Centers located in the United States, Europe, and Asia that provide a number of resources for customers, independent software vendors (ISVs), systems integrators, and managed service providers. Services offered at these centers include briefings, training, porting and migration assistance, benchmarking, and so on.
IBM has also created tools and processes that allow Linux to use the IBM Power technology. Some examples of these are:
Advance Toolchain: Is a set of open source development tools and runtime libraries that allow users to take full advantage of IBM Power hardware features. It is bundled, tested, and supported by IBM
Software Development Kit (SDK): Is an Eclipse-based integration of important Linux and IBM tools into a single GUI environment. It includes code scanners that recommend changes during x86 to Power porting that can improve performance (free of charge).
Migration Factory: When migrating to Linux, a trusted partner with a thorough, tested process is a key requirement. IBM has experience migrating thousands of clients to Linux on IBM systems with the The Migration Factory. It uses a five-step process that has been meticulously refined over 25 years, helping clients migrate to IBM systems running any supported operating system, including Linux.
Watson's infrastructure was based on the Power 750 server. This is an offering that fits squarely in the middle of our Power Systems family. You can see the detailed product offerings in the following chart and you can bet that any application that runs on one of these servers can run on any of these servers.
Figure 1 shows the recently announced IBM POWER8™ processor technology (April, 2014). At the time of this paper, the smaller 1- and 2-socket servers had been announced. Larger servers might be available later.
Figure 1. IBM POWER portfolio
IBM's Power division firmly believes that they can meet any customer's requirements with this portfolio. Look at the variety:
- Low-cost, scale-out servers such as the new Power System S812 and S824 and the IBM POWER7+™ processor-based Power 710 and Power 720.
- Linux-only servers such as the new Power System S812L and Power System S822L and the recent IBM PowerLinux™ 7R1 and 7R2.
- High-end enterprise servers with unmatched performance similar to Power 780 and Power 795
A final area worth mentioning is the level of contribution that IBM provides to the open source community. This might surprise many of you, but here is a partial list of the number of changes (contributions) made by those that focus on Linux.
Table 1. Contributions to Linux by company
|Company name||Number of changes||Percent of total|
As you can see, it is more than Watson when it comes to IBM's commitment to Linux on Power. Even though Watson is a great example of what IBM can do in this space, we know that there are many other application areas that we need to use with this technology combination.
Linux on Power is best positioned to handle customer workloads
Let's check out some trends that were identified in Gartner's 2013 Worldwide CIO Agenda. When asked to identify their top technology priority, analytics was at the top followed closely by cloud, and mobile. Other sources show supporting data such as the fact that 83% of CIO's cite business intelligence (BI) and analytics as part of their visionary plan and that 54% of companies use analytics to provide their business with explicit competitive advantages. The technologies behind big data and analytics are diverse, but one requirement is consistent. You need to provide hardware to run the application and that hardware must be able to excel in some very specific areas. First of all, these applications require extremely high throughput in many different places. Where specifically? Processor speed, memory bandwidth, and I/O throughput to name a few. Also, big data and analytics demand threads – lots of threads – and they also have some serious dependencies on Java™. Generally speaking, we are talking about balanced, efficient systems.
Let's not forget other workloads that are important to many customers. Email and collaboration (social and business-centric) remain strategic. While email client components are obviously transitioning away from the desktop and increasingly moving to mobile devices, the back-end infrastructure to support any form of client must still exist. Security becomes paramount with these applications. The ability to scale quickly and to interface with many other technologies is also important. Finally, yes, just like everything else, performance is something customers look for with these workloads.
Also, business applications including enterprise resource planning (ERP), supply chain, and customer relationship management (CRM) are still in use by nearly every major corporation in the world. What kinds of unique infrastructure requirements exist for these heavy-hitters? We need to be able to quickly create new environments to test new functionality. We need to be able to adjust system resources (such as processor, memory, I/O) dynamically to handle peak-processing periods such as financial closes and demand plans. These applications also have some of the highest uptime requirements of all. Finally, application performance must meet acceptable levels.
If we compile a list of the most important system requirements for the most popular workloads, it would probably look something as shown in the following list (all of this in an ultra-secure, cost-optimized solution):
- Flexibility (we call this virtualization)
Let's make sure we can get all of this in a cloud offering too!
It is no coincidence that the name of this paper is simply "Why Linux on Power?" All of the applications mentioned above need a place to run and we want to show you why, from a technical standpoint, Power servers running Linux offer a compelling business case. Part of that business case is price and we will get to that point, but let's dive a little deeper into the Power technology.
Comparing Power with Intel
Power has been the platform of choice for mid-range database needs for many years. However, when IT people think of Linux, they almost immediately think of the Intel® x86 platform. Take a look at the features of Power and Intel for a moment. We at IBM are very proud of the Power platform for its outstanding reliability, availability, serviceability, scalability and most importantly security.
If we look at the features of Power and compare them to the current line of Intel servers (see below), we can see superior functionality and flexibility. Additionally, from a cost standpoint, Linux on Power is extremely attractive. This is covered more fully in the following section of this article. However, the following charts show some of the benefits of Power over Intel.
Figure 2. Some benefits of IBM Power over Intel
Figure 3: Cache and bandwidth differences between Power and Intel
For server technology, Power has the following three specific advantages over Intel:
- Performance – POWER7+ offers a full line of servers with up to 256, high-performance cores and four threads per core. The new POWER8 processor-based servers have doubled the thread capability by offering eight threads per core. Intel sells smaller systems with current offerings topping out at 60 cores and two threads per core that require much more heating, cooling, and management.
- Virtualization – IBM PowerVM® is the only hypervisor in the market that can boast of zero security vulnerabilities. No other x86-based hypervisor can make that claim. PowerVM resides in firmware, therefore, reducing latency and allowing for much higher overall utilization of the hardware.
- Resilience – Because Power is engineered by the same team who has engineered a very reliable system, the IBM Mainframe, that same resilience is built in allowing for much higher uptimes and guaranteeing 99.997% uptime in a calendar year. Intel cannot come near that claim. Power hardware is self-healing and the most-critical updates can be done without any downtime through features, such as Live Partition Mobility (LPM). x86 servers currently do not have the same mission-critical resilience characteristics.
IBM has been consistently delivering on the Power roadmap for over 20 years producing revolutionary, leading-edge processors and transistor technology. Research and development have continued to be a focus of IBM investment resulting in industry-leading innovations. Power continues to have a clear and detailed technology roadmap.
The current version of the Power chip has built-in cryptography and an accelerator for hardware transactional memory. Hardware transactional memory enables Power servers to run highly threaded Java workloads with higher performance capabilities.
Power also has superior memory and cache performance. With three levels of cache and two channels per memory socket, Power provides superior performance for workloads requiring high internal bandwidth.
The proliferation of x86 servers often leads to the following customer challenges:
- Increasing system administration and management costs
- Excessive energy usage and heating problems
- Inadequate power and cooling infrastructure
- Skyrocketing software costs
- Growing staffing costs
- More inexplicable outages
- Increased unplanned downtime
- Reduced security
- Little automation, no flexibility
Power server's enhanced security features and lack of security vulnerabilities allows IT managers to avoid costly consequences of security breaches such as:
- Strengthening existing IT security and carrying out additional training
- Contacting those whose records may have been exposed
- Credit monitoring for those affected
- Legal action taken by people who may have suffered a financial loss
- Damage to the company/brand reputation
- Email blacklisting
- Impact on share price
- Costs to regain market position
One common myth about Linux is that it is synonymous with x86 or commodity hardware – why would one consider running Linux on Power? The common perception is that Power is too expensive to run Linux.
As mentioned earlier, all Power servers run Linux. However, IBM introduced Linux-only servers to provide customers another solid platform alternative for Linux at a competitive price point. Options include the IBM POWER7® processor-based one-, two-, and four-socket servers – IBM PowerLinux 7R1, 7R2, and 7R4 – and now the POWER8 one-socket and two-socket Power S812L and Power S822L models. Consider the following charts. If you look at the total cost of acquisition (TCA), which includes the server list price, virtualization, Linux operating system subscription and support list price, the TCA pricing for Linux on Power servers is very comparable to Intel x86 options if not less expensive.
Figure 4. IBM Power 822L pricing comparison (in U.S. dollars)
Figure 5. Comparing the TCA pricing of POWER7+ with Intel x86 for Linux
In addition to TCA, there are other cost factors to consider as well. Let's compare the virtualization capabilities available on Power with x86; specifically, PowerVM against VMware. As Figure 5 demonstrates, PowerVM offers superior capabilities and value in terms of providing customers with virtualization features that are more flexible, more scalable, and more secure. Why should customers care about these benefits and what does it mean in terms of cost? Note that PowerVM allows for flexible configurations for virtual processors and memory, and the number of virtual processors per VM is only limited by the number of cores available on the physical server. Contrast that with VMware where you do have limited flexibility in shifting processor and memory resources, and there is a hard limitation to 32 virtual processors per socket. What this means is that customers deploying with PowerVM have the flexibility to modify their system configuration based on their needs. In today's world, IT needs are fluid and dynamic. PowerVM allows customers to adjust their server environment as their needs change. Furthermore, superior scalability characteristics of PowerVM can help customers to consolidate more workloads onto a single server. Instead of acquiring additional physical servers to accommodate expanding or new workloads, PowerVM customers can run more workloads within a single server, dramatically increasing overall system utilization. PowerVM enables customers to do more with less. With fewer physical servers, there are savings in terms of power and cooling, data center rack space, and management costs. Software licensing cost is also a major factor. Most commercial software applications are licensed by number of cores. If you can do more with fewer cores, then that directly translates into less software licensing expense, which can be a significant portion of IT budgets. Next, let's explore the area of security. PowerVM is integrated into the Power hypervisor with zero reported vulnerabilities. Contrast that with VMware which has 561 reported vulnerabilities. Imagine a system with several hosts. If the hypervisor is vulnerable to security-related issues, all of the hosts that run on it are at risk. This can mean serious impacts to customer businesses in terms of downtime. Finally, if you examine cost from a virtualization license cost alone, you would notice that PowerVM licenses are very competitive as compared to VMWare.
Figure 6. Comparing PowerVM for IBM PowerLinux with VMWare (in U.S. dollars)
In Q4 of 2013, IBM made two additional announcements that make IBM PowerLinux an even more attractive option for customers. These announcements are the introduction of the Power Integrated Facility for Linux (IFL) and reduced processor value unit (PVU) for all IBM PVU licensed software products for all Power cores running Linux.
Power IFL is an enterprise Power offering to consolidate and integrate Linux applications with IBM AIX® and IBM i applications and data in large enterprises. Customers with Power 770, 780, or 795 servers can add Linux workload capacity through capacity on demand (add additional processor and memory resources using activation key, no outage required). This offering is scalable to 32-sockets through 4-core, 32 GB memory bundles. It is intended for enterprise customers who may have inactive capacity on their existing enterprise Power servers. Instead of acquiring additional x86-based servers to run Linux workloads, customers can now just activate the inactive resources on existing Power servers, with no outage required to activate the resources. Customers can enjoy the performance, security, and availability that are inherent in the Power enterprise servers, while streamlining the purchase of Linux workloads at new low prices. Power IFL allows customers to reduce time to market as there is no need to add additional equipment to the data center. You can just enter a key to activate the needed resources. By using IFL, customers can co-locate applications and data confidently while reducing latency and maximizing utilization.
Figure 7. Power IFL pricing
The second announcement is the reduced IBM Software Group PVU for all PVU-licensed software products on all Power cores running Linux. A PVU is a unit of measure used to differentiate software licensing on distributed processor technologies. For Linux on Power, the IBM Software Group PVU is 70 PVUs per core across the board, even for the largest Power servers. Compare this to 100 to 120 PVUs per core for 4-socket and larger Intel servers. Because software licenses can be the largest cost in solution acquisition (far higher than server costs) there can be significant costs savings when running such workloads on Linux on Power
The myth that Power is too expensive to run Linux does not hold true. Not only is it competitive in TCA pricing, but it also provides superior features that customers would value above commodity solutions.
Here is a fact-based analysis video by a leading research firm, Solitaire Interglobal Ltd. In this video, Solitaire compares the x86 architecture with IBM Power Systems and explains why the foundation for your next generation applications should be IBM Power:
IBM has been an active Linux participant since 1999 and is one of the leading commercial contributors to Linux. As we mentioned earlier, the IBM Power division invested another 1 billion U.S. dollars towards enabling solutions for Linux and open source workloads. This investment includes the opening of new Power Systems Linux Center and Innovation Centers around the world. The Power Systems Linux Center is available to Linux developers around the world. Locations include Beijing/China, Austin/TX, New York/NY, Tokyo/Japan, and Montpellier/France. Linux Center enables developers to build and deploy new applications using Linux and the latest IBM Power servers. Resources available at the center include tailored customer briefings, Linux training workshops, porting assistance, and hands-on assistance for developers on how to take advantage of Power capabilities and optimize performance. IBM Innovation Centers are state-of-the art facilities dedicated to helping ISVs. The services offered at the Innovation Centers include architectural design and implementation consulting, porting, migration, and testing services, support for application integration and proof of concepts, and expertise in the latest technologies.
As an example of IBM's commitment to Linux on Power, IBM has dedicated teams specifically around the mission of enabling clients and ISVs on PowerLinux. Furthermore, IBM has the Linux Technology Center (LTC), which is an IBM team of open source software developers who work on cooperation with the Linux open source development community. The LTC serves as a center of technical competency for Linux and is the technical liaison to IBM's Linux Distribution partners.
IBM offers a number of other community resources as well. Assume that you need to leverage IBM to help you port your application. IBM has the Chiphopper program, which is IBM Systems Application Advantage for Linux. This program engages the appropriate resources within IBM to work with you and your team in assessing potential porting challenges and assists with the port. Another IBM program is the Migration Factory. IBM Migration Factory can help you develop a migration roadmap, an estimate of the migration effort, and a customized business case for the migration. If you require a platform for testing but do not have access to a server, IBM has the Power Development Cloud, or what was previously known as the virtual loaner program, available. This program offers no-charge remote access to IBM hardware. Finally, IBM has the Linux on Power community wiki, which is your one-stop-shop for helpful FAQs, video demos, best practices, how-tos, and access to the Linux on Power forum. Questions or discussion points posted on the discussion board is automatically forwarded to the entire Linux on Power development team – allowing you to tap into the deepest Linux on Power technical knowledge through our experts.
IBM has developed tools specifically for Linux on Power to provide additional value. One tool is the IBM Installation Toolkit for PowerLinux. It includes the Simplified Setup Tool for popular workloads, Linux on Power expert experience based tuning, and installation of over 20 reliability, availability, and serviceability and productivity tools. Other tools include the SDK and Advance Toolchain. All of these tools are offered free of charge and are optimized specifically for Power.
Whether you're a customer, Business Partner, or an ISV, IBM has the team and resources to assist you with all of your Linux on Power needs.
NC State University (NCSU) – big data – helping companies solve real information problems
In the video, researcher Dr. Michael Kowolenko discusses how IBM's cloud-based big data and analytics solutions on PowerLinux empowered NCSU to advise businesses on making high-impact investments
NC State Center of Innovation Management Studies wanted to teach their students how to solve complex business questions and give businesses actionable information to help their bottom line. The businesses that they used as examples had many sources of data in disparate forms.
They needed a hardware platform that can keep up with the volume of data and not get bogged down in very complex queries and allow the students and businesses to make the right decisions based on accurate output. They also wanted to make it easy for those folks who would be running the system in the Computer Science department. So, they chose Linux on Power. Using several IBM software products (Big Insights, Content Analytics, IBM SPSS®, IBM DB2®) they were able to greatly enhance the ability of their students to solve problems that would affect the bottom line of these businesses. Moving from an x86 platform to a Power platform, the running of indexes was reduced from 72-96 hours to 14 hours.
Watson helps with cancer diagnosis at MD Anderson
Professor Lynda Chin of MD Anderson Hospital discusses how by using Watson, they were able to create an application that greatly speeds the diagnosis and treatment of cancer patients.
IBM Watson technology is currently being deployed at MD Anderson Hospital in Houston, TX to assist with the diagnosis and treatment of all kinds of cancer.
Watson gives practitioners access to information and data from sources they never dreamed of in the past. They are now no longer isolated in their own practice but can have instantaneous information for sources all over the globe to be better able to treat and diagnose their patients.
Watson addresses the fundamental challenges in how medicine is practiced today.
- Access to quality of care
- Physician shortage
- Lack of standardization
- Cost of access
- Inefficiency in adoptions
- Inefficiency in knowledge sharing and exchange
At its core, Watson is a Hadoop cluster and runs on Linux on Power systems.
We hope that we convinced you that IBM is serious about Linux on Power. Our investment into this technology should make it clear. Among these: a billion dollars of additional investment was committed in mid 2013 including a many of new, dedicated Linux-focused people and the continued commitment to build applications that use and improve Linux on Power. We also hope that we have dispelled some of the myths commonly in place about this product combination. First, Linux on Power is the same Linux as x86. Next, it does not cost more to run Linux on Power. In fact, in most cases, the total cost of acquisition and the total cost of ownership will favor running it on Power, sometimes dramatically. Finally, Linux surely does not run better on x86. The IBM POWER® processor and the server infrastructure surrounding it are unmatched in their ability to provide the fastest total system throughput with the highest levels of reliability and security.
If you have questions or want more in-depth information on any of the topics discussed in this article, you can contact one of the authors. Grace, Ann, and Mark are all part of a newly defined organization within IBM whose primary mission is to act as technical ambassadors for all aspects of Linux running on Power. These resources and others in their organization are focused on applications and initiatives such as big data and analytics that thrive in Linux environments. They size and design optimal Power-based architectures, translate designs, and provide comparative sizing from competitive hardware providers, recommend porting paths from alternate platforms, and educate customers and ISVs on the underlying technologies.
- IBM POWER8 systems facts and features
- Power Systems Quick Reference Guide
- Draft IBM Redbooks - IBM Power Systems S812L and S822L Technical Overview and Introduction
- Draft Redbooks - IBM Power Systems S814 and S824 Technical Overview and Introduction
- Draft Redbooks - IBM Power Systems S822 Technical Overview and Introduction
- Draft Redbooks - Performance Optimization and Tuning Techniques for IBM Processors, including IBM POWER8
- Youtube - Linux on Power: The Elements of Innovation
- Youtube - Infor: Enabling Client Success on the POWER8 Platform