:
Taming the cost of coherency for multicore systems
Innovation Matters: Computer Architecture
When building multicore systems, computer architects must increase performance while exploiting the data-sharing benefits of parallelism. In designing the basic building block for the BlueGene/P supercomputer -- a four-way multicore node -- our architects turned to a filtering technique that recognizes and then eliminates processes that result in unnecessary cost.
Java Runtime Optimizations
Innovation Matters: Programming Languages & Software Engineering
The Java programming language has gained widespread popularity in the industry since its advent in 1995. An efficient and well-designed Java Virtual Machine (JVM) and Just-In-Time (JIT) compiler are crucial to provide high performance for Java applications. The size and complexity of Java-based server and middleware applications have been growing steadily, which creates performance challenges. IBM Research has produced key technologies in both JVMs and JIT compilation, making significant contributions to the Sovereign Virtual Machine for production use, and producing the Jikes Research Virtual Machine (RVM) for the research community.
Understanding Performance Analysis, Modeling, and Optimization
Spotlight: November 2003: Performance Modeling and Analysis
IBM Research has a distinguished history in the theory and practice of performance analysis, modeling and optimization. Fundamental and foundational contributions have been made in a number of important areas, including: product-form queueing networks; optimal control and scheduling in queueing networks; stochastic ordering and majorization; rare-event and parallel simulation; matrix-analytic analysis of stochastic models; polling systems; and performance modeling tools. These have played a critical role in understanding important problems in the design, development, management and planning of complex systems. With the advent of the new technologies such as Autonomic Computing and new business models such as On Demand, performance modeling and analysis has become increasingly important for the delivery of high-quality and continuously available services.
Data Management Research Spotlight
Spotlight: March 2004: Data Management
IBM Research has produced major contributions to the area of data management for more than three decades. This includes E. F. Codd's seminal work on relational algebra; the System R relational database management system prototype (which led to IBM's DB2®); ARIES transaction recovery and logging; Starburst extensible database technology, and DB2 parallel database technology. Today, we continue to explore new data management technology in the areas of data warehousing, object-relational features, digital libraries, multimedia content management, federated databases, and integration of structured, semi-structured, and unstructured data, as well as the emerging areas of e-commerce, Internet, and mobile applications.
HealthMiner
Innovation Matters: Computational Biology & Medical Informatics
HealthMiner is an association rule discovery application that mines biomedical data for relationships among medical conditions, test results, treatments, etc. The goals of this joint effort with IBM Healthcare and Life Sciences include discovering new relationships in biomedical data for conducting research, making diagnoses and determining prognoses.
e-Commerce
Spotlight: March 2003: Services Computing
E-Commerce and E-Business are one of the largest research activities in IBM Research. We are in the unique position of having a very comprehensive coverage of this area with our research spanning problems of real businesses to more forward looking topics such as experimental economics and agent based economies. We are also very active in promoting e-commerce/e-business research and education in the academic community.
Enterprise Search Technology
Innovation Matters: Web
WebSphere Information Integrator OmniFind Edition is an enterprise information retrieval product, that combines scalable and robust indexing of data from multiple data sources with state-of-the-art retrieval and ranking algorithms. It supports a rich and expressive search syntax, while also supporting the simple internet-style search queries that are typically submitted to Web search engines.
Fossilization: Compliant reference storage solutions
Innovation Matters: Storage Systems
Enables organizations to maximize the value of their electronic records and minimize liability by providing practical solutions for storing and managing large volumes of electronic records over extended periods of time.
Statistical methods for business intelligence
Innovation Matters: Statistics
The goal of this research project is to develop innovative statistical methods for business analytics. In modern businesses, data are regularly collected from different business processes to help make them more efficient, effective and profitable. For such data-based business intelligence, IBM Research is developing statistical methods essential for data analysis, modeling, decision making and forecasting.
The Computational Biology Center
Spotlight: August 2003: Computational Biology & Medical Informatics
Computational Biology (CompBio) including bioinformatics is the study of how computer systems can manage, analyze, and simulate the complex structures and processes inherent in living systems.
An On Demand Platform for Online Games
Innovation Matters: Communications & Networking
Online gaming is a rapidly growing segment of the video games industry and has the potential to be a killer Internet application. Current massive multi-player games support hundreds of thousands of simultaneous players, and that number is expected to continue to increase.
Everywhere Interactive Displays
Innovation Matters: Mobile Computing
IBM Research is creating a new generation of display devices that transform ordinary surfaces into interactive touch screens. Combining an overhead projector, a steerable mirror, video cameras, and advanced computer graphics and image processing, the Everywhere Interactive Displays projector augments everyday reality with projected "touch screens", animated labels, video clips, and interactive animated graphics. This new technology is being incorporated by IBM Digital Media into end-to-end solutions for the retail and the transportation sectors.
Stream Data Mining
Innovation Matters: Knowledge Discovery & Data Mining
With the advance of data gathering and communication technologies, it has become increasingly important to support real-time monitoring of large amounts of information from diverse information sources. Examples include trade surveillance for security fraud and money laundering, network monitoring for intrusion detection, and various sensor network-based monitoring applications. Because data is a continuous stream in these applications, the goal is to develop stream mining techniques that can discover temporal patterns or anomalies on the fly.
Distributed and Fault Tolerant Computing
Spotlight: June 2003:
Distributed and Fault-Tolerant Computing (DFTC) addresses many of the core issues in computer science. Early successes on fundamental topics, such as, transaction processing, recovery, clustering, etc. have been key enablers for current state of computing. Our current focus is on addressing key challenges in grid, autonomic computing, flexible integration of disparate IT and business processes for enabling on-demand services.
Research on User Interface Technologies
Spotlight: January 2003: User Interface Technologies
IBM Research is spearheading many current activities in the area of communications and networking ranging from in-depth performance evaluation of content distribution networking, server-centric quality of service support, high-performance web caching, application and query caching techniques, scalable architectures for network processing offload, network processors, innovative usage of MPLS technologies, storage networking, and several other leading-edge projects.
Artificial Intelligence research at IBM
Spotlight: August 2003: Artificial Intelligence
IBM has been a leader in artificial intelligence since AI’s earliest days, when Arthur Samuel (in the 1950s) developed an expert checkers-playing program that learned from experience. Forty years later, IBM Research’s chess-playing program DEEP BLUE made history by beating world chess champion Garry Kasparov.
Speech-to-Speech Translation
Innovation Matters:
This research develops a universal speech translator that assists human communication using natural spoken language for people who do not share a common language. Multilingual speech-to-speech translation is considered a "grand challenge" of speech and natural language technologies, demanding the best technologies in Automatic Speech Recognition (ASR), Text-to-Speech (TTS) Machine Translation (MT), Natural Language Understanding (NLU), Natural Language Generation (NLG).
Algorithms & Theory
Spotlight: October 2003: Algorithms & Theory
Researchers at IBM have a distinguished history in the development of algorithms and the theory of computation in a variety of areas including foundations of complexity theory, combinatorial optimization, randomized algorithms, cryptography, streaming algorithms, search algorithms, database theory, queuing theory, online algorithms, quantum computation, communication networks, and algebraic circuit complexity. Our researchers are widely and internationally recognized leaders in many of these areas, which is continually demonstrated by major awards, plenary and invited talks and active participation in key conferences.
Mobile & Pervasive Computing
Spotlight: September 2003: Mobile Computing
IBM’s research activities in mobile and pervasive computing span three areas: extension of existing enterprise applications and Internet content to mobile and pervasive devices; creation of new pervasive-computing paradigms enabled through existing mobile and wireless technologies; and exploration of new, revolutionary, and paradigm-shifting technologies for pervasive computing.
Natural Language Processing
Research Spotlight (February 2004): Natural Language Processing
Our mission is to offer speech and language technologies that form the core of current and future products and solutions for processing natural language. We adopt linguistic, knowledge-based, statistical, as well as hybrid methods to natural language processing.
Unstructured Information Management Architecture (UIMA)
Innovation Matters: Natural Language Processing
The UIMA project is producing a common integrating platform for unstructured information analysis. A primary focus is on facilitating the interoperation of independently developed natural language processing components for discovering semantics in natural language documents. The result is improved intelligent search applications in a variety of domains, ranging from e-commerce to call-center support to life sciences.
BlueGene/L: Benchmarking and Applications Performance
Innovation Matters: Supercomputing
This project consists of a variety of research to improve trust in our computing platforms. This research project originated with the development of the IBM secure co-processors (4758 and 4764) and continues with research that investigates how to leverage the Trusted Computing Group consortium's hardware initiative.
Web Technologies Research
Spotlight: January 2004: Web
IBM Research has been active in virtually all areas of Web-related research since the Web’s inception. Our contributions are related to the provision of dynamic data at some of the most highly accessed Web sites in the world, new techniques for searching and characterizing the Web, industry standards, most recent in the area of Web Services, infrastructures enabling the Web to move to XML, cooperative and collaborative Web applications, and Web access from mobile devices.
Egyptian Culture Project
Innovation Matters: Graphics & Visualization
An extraordinary partnership between the Egyptian government and IBM has created Eternal Egypt, an interactive Web site experience providing worldwide access to 7000 years of Egyptian history.
AMBIENCE - Automatic Model Building using InferENCE
Innovation Matters: Performance Modeling and Analysis
Performance modeling has become increasingly important in the design, engineering and optimization of Information Technology (IT) infrastructures and applications. One of the biggest challenges in modeling complex IT systems is in the calibration of model parameters, such as the service requirements of various job classes. In this project we developed an innovative approach for solving this problem in the queueing network framework using inference techniques. This approach has been successfully applied to sizing and capacity planning in several pilot projects.
Research for Advancing Trusted Computing
Innovation Matters: Security
This research project originated with the development of the IBM secure cryptographic coprocessors (4758 and 4764) in the mid-1990's and continues with research that investigates how to use the Trusted Computing Group consortium's hardware initiative to verify the integrity of software running on a remote system.
Programming Languages and Software Engineering
Spotlight: November 2002: Programming Languages & Software Engineering
IBM Research's expertise includes all software lifecycle activities (design, development, test), program analysis and optimization. As systems become more tightly connected due to web-based and networking technologies, new opportunities arise for both programming languages and software engineering. The breadth and depth of expertise that IBM Research has in these areas uniquely positions us to play a leading role in the development of the next generation of programming systems and methodologies.
The Cell architecture
Innovation Matters: Computer Architecture
A cross-disciplinary team from Sony, Toshiba, IBM Research and IBM Systems Technology Group embarked on a project to redefine and dramatically increase the computing capabilities of game consoles. The resulting architecture (known as Cell) is based on a heterogeneous chip-multiprocessor that takes parallel processing to a new level. Previous systems had just two cores on a single chip, but the team envisioned a highly parallel system with 10 processor cores that would implement the IBM Power architecture. Novel, customized accelerator engines, called Synergistic Processing Units (SPU), exploit single-instruction, multiple-data parallelism, another innovative difference in the technology.
Personal Wizards: Learning Procedural Knowledge by Demonstration
Innovation Matters: Artificial Intelligence
Computer-based procedures are at the heart of a variety of activities including business processes, system administration, and technical support. The goal of Personal Wizards is capturing procedural knowledge (know-how) from demonstrations. By observing multiple users as they perform procedures and combining these observations, our system produces executable procedures that can be reexecuted or disseminated.
Operating Systems
Spotlight: December 2002: Operating Systems
Operating Systems are key in providing an abstraction to the hardware and in providing the essential services to the middleware and applications. Applications and middleware, such as databases, e-commerce, web hosting, and grid computing, place new demands on the services provided by the OS, such as security, scalability, availability, maintainability, and adaptability. New initiatives such as autonomic and on-demand computing require a fresh approach to operating system development.
Software Networking
Spotlight: February 2003: Communications & Networking
IBM Research is spearheading many current activities in the area of communications and networking ranging from in-depth performance evaluation of content distribution networking, server-centric quality of service support, high-performance web caching, application and query caching techniques, scalable architectures for network processing offload, network processors, innovative usage of MPLS technologies, storage networking, and several other leading-edge projects.
Computer Architecture
Spotlight: July 2003: Computer Architecture
Over the years, IBM Research has made seminal contributions to the field of computer architecture.
As advances in all areas of computing continue
Spotlight: October 2002: Graphics & Visualization
IBM Research's expertise in these areas includes the acquisition, representation, processing, and transmission of geometric and image data. To convey valuable information about this data, on devices ranging from PDAs to high-resolution displays, we also develop efficient rendering and visualization algorithms. To allow our techniques to handle all sized data sets, we design complete visualization systems, from hardware to software, which range from a single PC to massively parallel clusters of workstations.
The K42 Research Operating System
Innovation Matters: Operating Systems
A research operating system for cache-coherent 64-bit multiprocessors, K42 is being used as a high-performance framework for exploring system software innovation. Its object-oriented design has been explored to achieve scalability, fine-grain customization, and dynamic adaptation of services. K42 is an open-source project released under the LGPL license, and is Linux API and ABI compatible.
SHARK shorthand
Innovation Matters:
Researchers at the IBM Almaden Research Center are inventing and developing novel human-computer interaction methods for future client computing. SHARK shorthand is a high performance text and command input method for hand held mobile devices. It combines novel pattern recognition technology with a stylus keyboard. A new user traces the letters on the keyboard to enter a word. Over time the user remembers parts of the patterns which speeds up the text writing, making SHARK shorthand much faster and more effortless than a conventional stylus keyboard.
Multi Dimensional Clustering (MDC) in DB2
Innovation Matters: Data Management
Multi Dimensional Clustering (MDC) provides a performance oriented solution for query processing and data maintenance in a relational environment using a multi-dimensional block oriented clustering scheme for data storage. We have implemented novel techniques for query processing operations which provide significant performance improvements for MDC tables as well as techniques to maintain the physical layout efficiently over time.
Wireless sensor networking
Innovation Matters:
In the near future, networked sensors and actuators will enable a plethora of new services and applications. The goal of sensor networks and edge server software research in at the IBM Zurich Laboratory is to create innovative concepts, efficient architectures and state-of-the-art reference designs to develop IBM’s business in the area of end-to-end enterprise solutions ranging from sensors/actuators to wireless networking, middleware, and business applications.
Service-oriented architecture for extended business collaboration
Innovation Matters:
The SOA for Extended Business Collaboration project provides an innovative framework to move the current Web services’ simple client-server model to a comprehensive service-to-service collaboration model by introducing a uniform business resource description model (e.g. Site, Project, Task, requirements, Transactions, Documents, References, etc.) and their relationships as well as their extensible message unit schema and message exchange patterns for enabling business integration and performance management in value chain environments.
Production design and operations scheduling (PDOS) for plate products in the steel industry
Innovation Matters:
Production design and operations scheduling (PDOS) integrates supply chain level planning (based on a 6 month horizon) to day to day scheduling (to the level of ten minutes) that is feasible at the plant floor and thereby provides a way for operationalizing key business perfomance indicators. The focus of this research is two-fold: (i) to provide a library of reusable kernel optimization functions that are customized for steel manufacturing, and (ii) generalize these to make them usable for a broader set of production processes in other industries.
LeakBot
Innovation Matters: Programming Languages & Software Engineering
Despite automatic reclamation of memory, memory leaks remain an important problem. We have found that diagnosing these problems requires an large (and increasing!) amount of time and expertise, and is complicated by the need to analyze machines in their production environments. We have developed an automated and scalable tool for diagnosing memory leaks, called LeakBot.
Resonant Clock Networks
Innovation Matters:
Clock distribution is one of the big challenges of high frequency energy-efficient processor design. Trends show that the global clock distribution is consuming a greater fraction of the chip power (ten percent and growing), and having a growing impact on performance for the largest fastest chips. Using an innovative on-chip distributed oscillator scheme IBM researchers have designed and built two test-chips one operating up to 4.6 GHz and the other slower, but with more built-in testability showing 70 percent recycling of global clock power and 70 percent reduction in performance-robbing jitter. While doing the first high frequency resonant clock design in a product will require a fairly radical design change, inevitably involving some risk, IBM’s strategy should be relatively non-disruptive compared to other recent resonant clock proposals because it allows testing at all frequencies to facilitate testing and debugging, delivers a low-skew constant amplitude clock, and provides power and jitter benefits over a fairly wide +/- 20 percent frequency window without tuning circuits. There are many ideas for ways to save power, and many ideas to improve performance, but resonant clocking appears to be one of the few innovations that promises to significantly reduce chip power consumption and increase performance.
Efficient Video Annotation (EVA) system
Innovation Matters:
Research and development of model-driven video and image retrieval systems is an exciting and active field in computer science research. A crucial aspect here is the availability of semantically labeled image and video data for training and evaluation. The IBM EVA Tool is a new annotation tool, developed by the Intelligent Information Analysis Group. It is a specifically designed multi-user system to allow producing high-quality annotations for large collections.
Punch-operation and die optimization for ceramic-module production
Innovation Matters:
Ceramic modules are high-tech versions of the familiar printed circuit board: they can have a hundred or more layers of metal wiring, alternating with insulating ceramic layers, with layer-to-layer connections through holes in the ceramic. These holes must be punched out mechanically, and optimizing this manufacturing process is a nice demonstration of the power of mathematics: we sped up the process by 5 to 50%, at absolutely no cost.
SPARCLE (Server Privacy ARrchitecture and CapabiLity Enablement) policy workbench
Innovation Matters:
SPARCLE (Server Privacy ARrchitecture and CapabiLity Enablement) is a research prototype of a policy workbench. The goal of the proposed system is to enable a logical and verifiable flow from natural language rules written by policy experts within an organization, through the implementation of the rules in the organization’s configuration, to the compliance audits of the enforcement decision logs.
Self-Aware distributed systems
Innovation Matters:
As distributed computer systems and networks continue to grow in size and complexity, systems management tasks such as real-time fault localization and problem diagnosis become much more expensive and complicated. Autonomic computing is a rapidly growing research area that sets an ambitious goal of automating systems-management tasks, and generates a variety of challenging application for AI field. Our project focuses on machine-learning approaches to developing Self-Aware Distributed Systems by using real-time Bayesian inference and learning combined with information-theoretic methods for optimal test selection, and develops novel cost-efficient, adaptive diagnostic algorithms. We also investigate robustness issues in large, complex networks and suggest algorithms for improving robustness in the presence of both random failures and coordinated attacks.
VLSI critical area analysis via Voronoi diagrams
Innovation Matters:
The Voronoi diagram of polygonal objects in the plane is a powerful mathematical object that encodes proximity information in a compact form. Critical Area is a widely accepted measure that reflects the sensitivity of a VLSI design to random manufacturing defects, used to predict yield. We have developed a new Design Automation tool that, based on the geometric concept of Voronoi diagrams, extracts fast and accurately the critical area of a VLSI design for all types of random manufacturing defects.
Business Insights Workbench (BIW)
Innovation Matters:
Business Insights Workbench incorporates interactive text analytics methodologies including clustering and taxonomy generation, text mining and analysis such as relationship analysis, trending, etc., classification, and visualization to derive values out of structured and unstructured information.
SoulPad
Innovation Matters:
The pocket-size SoulPad with a three layer software stack can be attached to and detached from a PC in the vicinity and provide its user the functionality of today's laptop computer, namely access to his data and applications and convenient suspend and resume of his running application state, and eliminates the need to carry a heavy computer around.
Estimating the difficulty of queries submitted to search engines
Innovation Matters:
How often have you entered keywords in a search engine, only to be rewarded with a huge list of items that are way off the mark? Our research is aimed at predicting the chances of a search engine providing the correct documents for a given query. Such prediction is useful as feedback to the user and the search engine administrator. It is also used for uniting data from different sources, as well as to predict when relevant documents cannot be retrieved simply because they do not exist in the repository.
RNA interference: small molecules that carry a big stick
Innovation Matters:
RNA interference (RNAi) is a recently discovered phenomenon whereby small size RNAs called microRNAs, typically 22 nucleotides long, disrupt the expression of specific genes by either degrading the genes’ messenger RNA (mRNA) or disrupting the gene’s translation by the ribosome. Many aspects of the RNAi mechanism still remain unclear, but increasingly more scientists agree that it holds the key to our understanding of many important biological processes. At one end of the spectrum of biological processes, studies in one small model organism (fruitfly) microRNAs were shown to regulate stem cell division, whereas in another (worm) to regulate lifespan. And at the other, several microRNAs were shown to be involved in processes leading to the onset of cancer or to act as oncogenes themselves.
Print verification system
Innovation Matters:
A Print Verification System scans and processes printed pages in real time to verify their print quality. Verifying pages printed at more than 750 per minute, it replaces previous manual inspection of randomly selected pages and thus reduces operational cost and delivers more consistent results.
Analytics-driven solution helps sales teams identify opportunities
Innovation Matters:
IBM researchers developed two Web-based tools that are helping sales teams address two perennial challenges: (1) How do you identify an opportunity at existing and prospective client accounts? (2) When should sellers focus their efforts on accounts that could generate revenue somewhere down the road?
DB2 pureXML: Tackling technology issues in a hybrid relational and XML database engine
Innovation Matters: XML goes in search of storage
An IBM Research data management team enhanced the performance of database queries by focusing on three technology issues.
Thermal-aware scheduling at the system level
Innovation Matters: Operating Systems
Overheated chips can have adverse effects on packaging costs, processor power and chip reliability. IBM Research is working to lower on-chip unit temperatures by scheduling workloads on operating systems and hardware.
Using data-parallel SIMD architecture in video games and supercomputers
Innovation Matters: Power Architecture with SIMD execution becomes the processing power-of-choice
By combining application data-level parallelism with data-parallel SIMD execution, IBM Research catapulted the Power Architecture into the premier architecture for graphics and digital media processing, digital content creation, video game consoles and supercomputers.
Melody summarizes computer system descriptions automatically
Innovation Matters: Artificial Intelligence
Melody, an analytic tool for systems managers, provides automatic and self-adapting summarization of computer system information.
Preservation DataStores: Storage assist for preservation environments
Innovation Matters: Storing digital assets forever
Preservation DataStores -- a project of IBM Haifa -- is an infrastructure component of CASPAR, a European Union project to preserve scientific, artistic and cultural artifacts.
Massive parallelism for power and performance efficiency
Innovation Matters: Supercomputing
While conventional manufacturers chase Gigahertz, IBM is conducting research that will change the way systems are conceived across the semiconductor industry. In the Innovator's Corner: Valentina Salapura. (14 November 2007)
Telling TALES to address translation challenges
Innovation Matters: Translingual Automatic Language Exploitation System
TALES is a multilingual, multi-modal analytic system that translates foreign-language news broadcasts and websites into English. (December 21, 2007)
Semantic Web technologies enhance enterprise Master Data Management
Innovation Matters: Web
IBM computer scientists are exploring semantic Web technologies to help businesses maintain customer, supplier and product lists that they use on a repeated basis. Innovator's Corner showcases IBM researcher Li Ma. (January 14, 2008)
How we can predict patient response to anti-HIV treatment
Innovation Matters: The EuResist Project
The Sixth Framework European Union funded the EuResist project in years 2006-2008; the IBM Haifa Research Lab was one of the partners in this project. It a research project that is developing an integrated European system for the clinical management of antiretroviral drug resistance.
Logic optimization techniques improve cycle-time of integrated circuits
Innovation Matters: Design Automation
A design automation team at IBM Research is using logic optimization techniques to enhance the productivity of high-performance integrated circuits in IBM microprocessors.
Statistics researchers predict road traffic conditions
Innovation Matters: Statistics at IBM Research
IBM Traffic Prediction Tool (TPT) is a statistical model that predicts near-future traffic conditions. It's now used in Singapore.
Ensuring access control in virtualized storage environments
Innovation Matters: Storage Systems
Capability-based Command Security is an access control protocol for the virtualized data center.
Taming the cost of coherency for multicore systems
Innovation Matters: Supercomputing
When building multicore systems, computer architects must increase performance while exploiting the data-sharing benefits of parallelism. In designing the basic building block for the BlueGene/P supercomputer – a four-way multicore node – our architects turned to a filtering technique that recognizes and then eliminates processes that result in unnecessary cost.
Developments in EIP
Spotlight: December 2002: Electrical Interconnect and Packaging
We are working on providing leadership packaging for supercomputers, servers, and client machines. We provide best-of-breed designs and simulation tools, develop new characterization techniques, and develop leading-edge thermal and mechanical solutions for IBM servers. We also build systems to explore the peak supercomputer performance obtainable using new architectural concepts, and we support supercomputing on existing high-end IBM machines.
VLSI Design
Spotlight: June 2003: VLSI Design
VLSI Design is the process of specifying the details of all transistors and their interconnections in a complete chip. Contemporary designs can contain millions of such transistors in the logic part of the chip. The overall task of VLSI designers is to specify those devices and their interconnections, within a set of technology dependent design rules, so that the final chip performs its intended function at required performance and power levels.
Applying digital halftoning to digital watermarking and LCD displays
Innovation Matters: Signal Processing
For authentication, security and other purposes, images often need to be embedded into other images. IBM Research has developed a unified way to do the embedding, using digital image watermarking, image authentication, self-repairing images, and viewing angle characteristics in LCD displays.
Electrical Interconnect and Packaging
Innovation Matters:
With the demand for increased hardware performance and a decrease in signal rise times, accurate electromagnetic (EM) modeling becomes of great concern to IBM engineers. Improving accuracy, however, means increasing the complexity of the electromagnetic model, and the performance of existing modeling tools cannot keep pace with the needs of the engineers. The goal of the EMSurf project is to provide a new EM modeling tool based on state of the art numerical techniques to handle larger, more complicated problems with increased efficiency.
The Cell architecture
Innovation Matters: Computer Architecture
A cross-disciplinary team from Sony, Toshiba, IBM Research and IBM Systems Technology Group embarked on a project to redefine and dramatically increase the computing capabilities of game consoles. The resulting architecture (known as Cell) is based on a heterogeneous chip-multiprocessor that takes parallel processing to a new level. Previous systems had just two cores on a single chip, but the team envisioned a highly parallel system with 10 processor cores that would implement the IBM Power architecture. Novel, customized accelerator engines, called Synergistic Processing Units (SPU), exploit single-instruction, multiple-data parallelism, another innovative difference in the technology.
X-Gen
Innovation Matters: Verification Technology
Functional verification is the process of validating that a hardware design conforms to its specification, and is deemed one of the bottlenecks of the hardware design cycle. X-Gen is a system-level test-case generator, used for the verification of some of IBM’s flagship hardware products – from the Cell processor powering Sony’s next generation game console to the CPU chip of XBOX 360 to Power5, the heart of a high-end pSeries eServer machines.
Resonant Clock Networks
Innovation Matters:
Clock distribution is one of the big challenges of high frequency energy-efficient processor design. Trends show that the global clock distribution is consuming a greater fraction of the chip power (ten percent and growing), and having a growing impact on performance for the largest fastest chips. Using an innovative on-chip distributed oscillator scheme IBM researchers have designed and built two test-chips one operating up to 4.6 GHz and the other slower, but with more built-in testability showing 70 percent recycling of global clock power and 70 percent reduction in performance-robbing jitter. While doing the first high frequency resonant clock design in a product will require a fairly radical design change, inevitably involving some risk, IBM’s strategy should be relatively non-disruptive compared to other recent resonant clock proposals because it allows testing at all frequencies to facilitate testing and debugging, delivers a low-skew constant amplitude clock, and provides power and jitter benefits over a fairly wide +/- 20 percent frequency window without tuning circuits. There are many ideas for ways to save power, and many ideas to improve performance, but resonant clocking appears to be one of the few innovations that promises to significantly reduce chip power consumption and increase performance.
Efficient Video Annotation (EVA) system
Innovation Matters:
Research and development of model-driven video and image retrieval systems is an exciting and active field in computer science research. A crucial aspect here is the availability of semantically labeled image and video data for training and evaluation. The IBM EVA Tool is a new annotation tool, developed by the Intelligent Information Analysis Group. It is a specifically designed multi-user system to allow producing high-quality annotations for large collections.
Punch-operation and die optimization for ceramic-module production
Innovation Matters:
Ceramic modules are high-tech versions of the familiar printed circuit board: they can have a hundred or more layers of metal wiring, alternating with insulating ceramic layers, with layer-to-layer connections through holes in the ceramic. These holes must be punched out mechanically, and optimizing this manufacturing process is a nice demonstration of the power of mathematics: we sped up the process by 5 to 50%, at absolutely no cost.
VLSI critical area analysis via Voronoi diagrams
Innovation Matters:
The Voronoi diagram of polygonal objects in the plane is a powerful mathematical object that encodes proximity information in a compact form. Critical Area is a widely accepted measure that reflects the sensitivity of a VLSI design to random manufacturing defects, used to predict yield. We have developed a new Design Automation tool that, based on the geometric concept of Voronoi diagrams, extracts fast and accurately the critical area of a VLSI design for all types of random manufacturing defects.
Estimating the difficulty of queries submitted to search engines
Innovation Matters:
How often have you entered keywords in a search engine, only to be rewarded with a huge list of items that are way off the mark? Our research is aimed at predicting the chances of a search engine providing the correct documents for a given query. Such prediction is useful as feedback to the user and the search engine administrator. It is also used for uniting data from different sources, as well as to predict when relevant documents cannot be retrieved simply because they do not exist in the repository.
RNA interference: small molecules that carry a big stick
Innovation Matters:
RNA interference (RNAi) is a recently discovered phenomenon whereby small size RNAs called microRNAs, typically 22 nucleotides long, disrupt the expression of specific genes by either degrading the genes’ messenger RNA (mRNA) or disrupting the gene’s translation by the ribosome. Many aspects of the RNAi mechanism still remain unclear, but increasingly more scientists agree that it holds the key to our understanding of many important biological processes. At one end of the spectrum of biological processes, studies in one small model organism (fruitfly) microRNAs were shown to regulate stem cell division, whereas in another (worm) to regulate lifespan. And at the other, several microRNAs were shown to be involved in processes leading to the onset of cancer or to act as oncogenes themselves.
Print verification system
Innovation Matters:
A Print Verification System scans and processes printed pages in real time to verify their print quality. Verifying pages printed at more than 750 per minute, it replaces previous manual inspection of randomly selected pages and thus reduces operational cost and delivers more consistent results.
Fast quasi-static capacitance extraction using CSurf
Innovation Matters:
The Electrical Interconnect and Packaging (EIP) group provides best-of-breed electrical designs, modeling and simulation tools, as well as characterization techniques for new products and future packaging needs in supercomputers and servers. One of the recent additions to this suite of electromagnetic (EM) modeling tools is an arbitrary shape 3D solver named Csurf, which stands for quasi-static capacitance extraction based on surface integral equation (or boundary element method). Benchmarks have demonstrated that Csurf can out-perform Ansoft Q3D by over 100 times for via array structures in multilayer dielectrics.
Statistical methods for business intelligence
Innovation Matters: Statistics
The goal of this research project is to develop innovative statistical methods for business analytics. In modern businesses, data are regularly collected from different business processes to help make them more efficient, effective and profitable. For such data-based business intelligence, IBM Research is developing statistical methods essential for data analysis, modeling, decision making and forecasting.
Stream Data Mining
Innovation Matters: Knowledge Discovery & Data Mining
With the advance of data gathering and communication technologies, it has become increasingly important to support real-time monitoring of large amounts of information from diverse information sources. Examples include trade surveillance for security fraud and money laundering, network monitoring for intrusion detection, and various sensor network-based monitoring applications. Because data is a continuous stream in these applications, the goal is to develop stream mining techniques that can discover temporal patterns or anomalies on the fly.
Algorithms & Theory
Spotlight: October 2003: Algorithms & Theory
Researchers at IBM have a distinguished history in the development of algorithms and the theory of computation in a variety of areas including foundations of complexity theory, combinatorial optimization, randomized algorithms, cryptography, streaming algorithms, search algorithms, database theory, queuing theory, online algorithms, quantum computation, communication networks, and algebraic circuit complexity. Our researchers are widely and internationally recognized leaders in many of these areas, which is continually demonstrated by major awards, plenary and invited talks and active participation in key conferences.
Large-Scale Strategic Budgeting for Wildfire Management
Innovation Matters: Operations Research
Research developed a strategic planning model for determining preparedness budgets necessary for managing wildfires during the initial response (a.k.a. initial attack) period. The model uses a linear optimization (integer programming) approach to maximizing weighted acres managed for any given cost level, an effectiveness metric used in this business problem. The model is solved iteratively across a range of total cost constraints to establish the effectiveness frontier of weighted acres managed. The uncertainty is captured in a scenario-based stochastic optimization approach rather than traditional simulation forecasting methods, which were inefficient and intractable, and even failed to solve a small deterministic instance of the problem. Research has succeeded in developing new algorithms and approaches for managing the data, for restating the problem and its constraints, and for optimizing the resulting problem in the presence of uncertainty. In addition to developing a tractable model, Research has succeeded in increasing the performance speed to 159 times over its initial model.
Statistical Analysis
Spotlight: September 2003: Statistics
Statistical analysis serves as a foundation for informed decision making in a number of areas central to IBM's business, such as quality assurance, manufacturing, pricing, marketing, business intelligence and delivery of services.
Continual Evolution of Automation, Models and Process
Spotlight: May 2003: Operations Research
We now have an unprecedented opportunity to analyze and model complex systems with increasing accuracy and precision, due to dramatic growth in the availability of processing power, advances in software architectures, the availability of robust software libraries, increases in the availability of real time digital data and fast, high-bandwidth communications capability.
Production design and operations scheduling (PDOS) for plate products in the steel industry
Innovation Matters:
Production design and operations scheduling (PDOS) integrates supply chain level planning (based on a 6 month horizon) to day to day scheduling (to the level of ten minutes) that is feasible at the plant floor and thereby provides a way for operationalizing key business perfomance indicators. The focus of this research is two-fold: (i) to provide a library of reusable kernel optimization functions that are customized for steel manufacturing, and (ii) generalize these to make them usable for a broader set of production processes in other industries.
Punch-operation and die optimization for ceramic-module production
Innovation Matters:
Ceramic modules are high-tech versions of the familiar printed circuit board: they can have a hundred or more layers of metal wiring, alternating with insulating ceramic layers, with layer-to-layer connections through holes in the ceramic. These holes must be punched out mechanically, and optimizing this manufacturing process is a nice demonstration of the power of mathematics: we sped up the process by 5 to 50%, at absolutely no cost.
Self-Aware distributed systems
Innovation Matters:
As distributed computer systems and networks continue to grow in size and complexity, systems management tasks such as real-time fault localization and problem diagnosis become much more expensive and complicated. Autonomic computing is a rapidly growing research area that sets an ambitious goal of automating systems-management tasks, and generates a variety of challenging application for AI field. Our project focuses on machine-learning approaches to developing Self-Aware Distributed Systems by using real-time Bayesian inference and learning combined with information-theoretic methods for optimal test selection, and develops novel cost-efficient, adaptive diagnostic algorithms. We also investigate robustness issues in large, complex networks and suggest algorithms for improving robustness in the presence of both random failures and coordinated attacks.
Estimating the difficulty of queries submitted to search engines
Innovation Matters:
How often have you entered keywords in a search engine, only to be rewarded with a huge list of items that are way off the mark? Our research is aimed at predicting the chances of a search engine providing the correct documents for a given query. Such prediction is useful as feedback to the user and the search engine administrator. It is also used for uniting data from different sources, as well as to predict when relevant documents cannot be retrieved simply because they do not exist in the repository.
RNA interference: small molecules that carry a big stick
Innovation Matters:
RNA interference (RNAi) is a recently discovered phenomenon whereby small size RNAs called microRNAs, typically 22 nucleotides long, disrupt the expression of specific genes by either degrading the genes’ messenger RNA (mRNA) or disrupting the gene’s translation by the ribosome. Many aspects of the RNAi mechanism still remain unclear, but increasingly more scientists agree that it holds the key to our understanding of many important biological processes. At one end of the spectrum of biological processes, studies in one small model organism (fruitfly) microRNAs were shown to regulate stem cell division, whereas in another (worm) to regulate lifespan. And at the other, several microRNAs were shown to be involved in processes leading to the onset of cancer or to act as oncogenes themselves.
A Statistical Approach to Capacity Planning for On-Demand Computing Services
Innovation Matters:
This research project explores the problem of capacity
planning for on-demand computing services in a shared environment, with the aim
of developing effect methods and algorithms to determine the capacity requirement
and optimal deployment strategies that fulfill service commitments with improved
resource utilization.
Production design and operations scheduling (PDOS) for hot strip mill scheduling
Innovation Matters:
Hot strip mill scheduling in the steel industry is a difficult problem involving several types of optimization problems: a knapsack problem, a constraint satisfaction problem, and a traveling salesperson problem. A coarse-to-fine approach was applied to the problem, and a scheduling system that outperforms experienced scheduling experts was delivered to a Taiwanese steel company.