IBM Research and Development – Ireland

City Fabric

The City Fabric project is the Open Innovation Platform which provides scalable technology and tools, allowing Cities to open data content to the Internet.

Smarter Transportation

City officials around the world are investing in their transport infrastructure, however the volume of raw transport data of diverse types is overwhelming. In collaboration with Dublin City Council we are addressing a range of system and analytics challenges to deal with the continuous assimilation of large volume of diverse raw data sources to monitor the state of transport systems in real.

Smarter Energy

Today there is a need for smarter, more dynamic power grids that can dramatically reduce outages and faults, improve responsiveness, handle current and future demand, increase efficiency and manage costs. Complexity is increased by dynamic consumption patterns and distributed generation in the electricity network. At DRL we aim to apply forecasting and scenario reduction techniques to enable domain experts to have an integrated view of the opportunities and risks. These new models will be based upon past data from smart meters, sensors on the network, and external data points (eg. Weather)

Smarter Water

We are exploring how sensor systems installed in many buildings today can be used to apportion water/energy consumption between users, leading to apportionment of constrained resources based on historical data. We are also building a computational technology that will use real time sensors to integrate, analyse and enable predictive water and environmental services delivered via the web. As such we will develop a platform that provides a single, reliable, up to the minute view of water use.

DRL Risk projects focus on making rigorous decision making under uncertainty scalable within a business environment. Indeed, IBM has identified Risk Science as a key direction for research, recognizing that next-generation decision support models and tools must allow for uncertainty in both their inputs and outputs.

Natural Language Processing:

Simplify the creation of risk models by extracting risk information from structured and unstructured data, with a focus on text.

Imprecise Probability:

Lessen the input requirements for risk models by develop efficient decision theoretic algorithms to work with imprecise probabilistic information such as ranges or orders of magnitude.

Constraint Programming applied to Stochastic Optimization:

Find effective way of solving complex stochastic problems by combining constraint programming with stochastic programming and address fundamental challenges in risk-based optimization

Risk Communication:

Experiment with novel ways for users to communicate about and interact with risk information through the development of novel risk visualization support and the refactoring of a web-based expert elicitation tool.

Escale

• The Exascale Computing team at IBM Research and Development –Ireland is advancing system technology to meet the challenging requirements introduced by very large scale emerging workloads, including stream computing applications, scale-out analytics such as map-reduce, and high-end cloud computing. The effort pursues innovation in areas such as simulation and performance modeling, networking, system software, HPC consumability, and large-scale automated management.

Simulation and Performance Modelling

• - Automatic Bottleneck Detection
  - Data movement optimization
  - Workload generation
  - Large input generation
  - Cloud workload testbed (deliverable to GTS)

Workload-optimized Systems

• - Workload-driven optical networking
  - Messaging/operating system services
  - Map-reduce task placement
  - Application management support / problem determination

Next Generation Systems team work by co-designing algorithms and software together with the hardware solution using the team Hybrid lab, which provides an incubation environment that enables design and prototyping of re-configurable and optimised system solutions.

The current main focus platform for the team is PowerEN, IBM's new Edge of Network processor. PowerEN is a blur of general purpose and network processors and is designed to provide wire-speed payload processing at general purpose server ease of use. PowerEN is a hybrid computer that employs: massive multiheading capabilities, integrated 10G Ethernet ports and unique special-purpose accelerators, such as Compression, Crypto, Regular expressions, XML and Ethernet. It also provides significant power reduction through Architecture and implementation.

We are engaged in the design and development both of the system software stack - providing the communications layer and high speed networking capability for this new hybrid system - and on the strategic application side - designing and developing of high throughput and low latency appliances for specific workloads, and leveraging the unique PowerEN capabilities.