Watson—IBM’s Jeopardy-winning computer—showcased the refined capabilities of the computer giant’s technologically advanced, workload-optimized systems. During its three-night stretch on the TV game show, Watson ran on the latest IBM Power7technology, which can manage the most demanding applications, such as real-time analytics. While Watson’s QA (question-answering) technology was an IBM research breakthrough and will emerge into the market over time, Watson’s other IBM systems are commercially available and in use in companies worldwide today. For instance, the IBM Smart Analytics System contains components that are alive in Watson. The Smart Analytics System represents a high-performance, easy-to-maintain, component-based infrastructure that leverages pre-tuned and pre-integrated IBM hardware, software and storage capabilities to create business-changing insights. The IBM analytics “appliance” is but one of a wide variety of systems technologies that relate to Watson. This eWEEK slideshow takes a look at 16 things you may not know about the muscle behind Watson.
Watson is powered by 10 racks of IBM Power 750 servers, all running the Linux operating system. The Power 750 is commercially available and is being used by customers around the world today.
Watson contains 15 terabytes of RAM and 2,870 processor cores. It can operate at 80 teraflops (80 trillion operations) per second.
POWER7 has the computing power to process an enormous number of concurrent transactions and data while allowing for real-time analysis. This enables the massively parallel analytical capabilities Watson needs to match the speed at which Jeopardy contestants must decipher clues and provide accurate responses through analyzing subtle meaning, irony, riddles, and other complexities of natural language.
The entire Watson data store resides in memory in order to achieve the sub-3-second response time needed for Jeopardy. The large memory support of Power systems enables the development of high-performance, workload-optimized applications. Watson was not connected to the Internet during Jeopardy—all the data was inside Watson, in memory.
IBM’s Power 750 servers have twice the amount of bandwidth than other commercially available systems, and include intelligent energy features that make Watson one of the most highly energy-efficient systems in its class.
To save energy, IBM’s Power7 systems allow customers to power on and off various parts of the system or to dynamically increase or decrease processor clock speeds based on thermal conditions and system use. All that can be done on a single server or across a pool of multiple servers.
IBM’s Power7 systems are the first four-processor servers in the industry to qualify for the EPA’s Energy Star status. Energy Star is a U.S. government program that rates the energy efficiency of many products, including computers and other electronics.
Watson also includes IBM’s SONAS (Scale-out Network Attached Storage), a file system invented by IBM Research that in Watson includes a total of 21.6 terabytes of raw capacity. The actual size of the data Watson uses to generate answers is less than 1 TB.
Investment in Power
Over the past four years, IBM has invested $3.2 billion in Power systems, leading to a year-long rollout in 2010 of workload-optimized systems for the demands of emerging business models such as smart electrical grids, real-time analytics in financial markets and health care, mobile telecommunications and smarter traffic systems.
Retailer Tupperware Australia relies on Power7 for a Web-ordering system used by its 6,000-person sales force and 32 Australian and New Zealand distributors.
Other Power7 Customers
Financial customer GHY International, which offers customs brokerage services in the United States and Canada, and Russian pharmaceutical distributor Pro-Tek are both running on Power7 systems.
Education and Research
Rice University in Houston is using Power7 for a major cancer research project involving researchers working in locations around Texas. The system allows for multiple streams of information to be processed in real time so that teams of researchers in disparate locations are working off the latest information.
IBM designed Power7 to meet the global data explosion. The amount of digital information that is being generated, stored, processed and analyzed each year is increasing at an exponential rate. IDC predicts that the total data volume will reach 35,000 exabytes in 2020, compared with 1,200 exabytes in 2010, representing a 29-fold increase in the next 10 years.
Power7 systems use more cores, or CPUs, and add more threads—or virtual cores, which are resources that manage computational tasks—per chip. Each new Power7 processor can now run 32 simultaneous tasks—with eight cores and four threads per core—quadruple the maximum number of cores of Power6 systems and eight times the number of threads per chip as Power6.
TurboCore mode, which is highly optimized for databases or other transaction-oriented workloads, provides more cache memory and memory bandwidth, and allows the clock speed of the chip to be increased. TurboCore mode can give financial benefit to customers, maximizing the return on investment from software by potentially cutting software costs in half for applications that are licensed per core, while increasing per-core performance of that software.
Power7 technology features "Intelligent Threads" that can dynamically vary based on workload demand. With more threads, Power7 can deliver more total capacity as more tasks are accomplished in parallel, such as monitoring millions of individuals’ household energy use by the minute in a smart grid.