Ramakrishna Soma

Fine-grained dynamic voltage and frequency scaling for precise energy and performance trade-off based on the ratio of off-chip access to on-chip computation times

This paper presents an intra-process dynamic voltage and frequency scaling (DVFS) technique targeted toward non real-time applications running on an embedded system platform. The key idea is to make use of runtime information about the external memory access statistics in order to perform CPU voltage and frequency scaling with the goal of minimizing the energy consumption while translucently controlling the performance penalty. The proposed DVFS technique relies on dynamically-constructed regression models that allow the CPU to calculate the expected workload and slack time for the next time slot, and thus, adjust its voltage and frequency in order to save energy while meeting soft timing constraints. This is in turn achieved by estimating and exploiting the ratio of the total off-chip access time to the total on-chip computation time. The proposed technique has been implemented on an XScale-based embedded system platform and actual energy savings have been calculated by current measurements in hardware. For memory-bound programs, a CPU energy saving of more than 70% with a performance degradation of 12% was achieved. For CPU-bound programs, 15/spl sim/60% CPU energy saving was achieved at the cost of 5-20% performance penalty.

Parallel Inferencing for OWL Knowledge Bases

We examine the problem of parallelizing the inferencing process for OWL knowledge-bases. A key challenge in this problem is partitioning the computational workload of this process to minimize duplication of computation and the amount of data communicated among processors. We investigate two approaches to address this challenge. In the data partitioning approach, the data-set is partitioned into smaller units, which are then processed independently. In the rule partitioning approach the rule-base is partitioned and the smaller rule-bases are applied to the complete data set. We present various algorithms for the partitioning and analyze their advantages and disadvantages. A parallel inferencing algorithm is presented which uses the partitions that are created by the two approaches. We then present an implementation based on a popular open source OWL reasoner and on a networked cluster. Our experimental results show significant speedups for some popular benchmarks, thus making this a promising approach.

Dynamic voltage and frequency scaling under a precise energy model considering variable and fixed components of the system power dissipation

This work presents a dynamic voltage and frequency scaling (DVFS) technique that minimizes the total system energy consumption for performing a task while satisfying a given execution time constraint. We first show that in order to guarantee minimum energy for task execution by using DVFS it is essential to divide the system power into active and standby power components. Next, we present a new DVFS technique, which considers not only the active power, but also the standby component of the system power. This is in sharp contrast with previous DVFS techniques, which only consider the active power component. We have implemented the proposed DVFS technique on the BitsyX platform - an Intel PXA255-based platform manufactured by ADS Inc., and report detailed power measurements on this platform. These measurements show that, compared to conventional DVFS techniques, an additional system energy saving of up to 18% can be achieved while satisfying the user-specified timing constraints.

Off-chip latency-driven dynamic voltage and frequency scaling for an MPEG decoding

This paper describes a dynamic voltage and frequency scaling (DVFS) technique for MPEG decoding to reduce the energy consumption using the computational workload decomposition. This technique decomposes the workload for decoding a frame into on-chip and off-chip workloads. The execution time required for the on-chip workload is CPU frequency-dependent, whereas the off-chip workload execution time does not change, regardless of the CPU frequency, resulting in the maximum energy savings by setting the minimum frequency during off-chip workload execution time, without causing any delay penalty. This workload decomposition is performed using a performance-monitoring unit (PMU) in the XScale-processor, which provides various statistics such as cache hit/miss and CPU stall, due to data dependency at run time. The on-chip workload for an incoming frame is predicted using a frame-based history so that the processor voltage and frequency can be scaled to provide the exact amount of computing power needed to decode the frame. To guarantee a quality of service (QoS) constraint, a prediction error compensation method, called inter-frame compensation, is proposed in which the on-chip workload prediction error is diffused into subsequent frames such that run time frame rates change smoothly. The proposed DVFS algorithm has been implemented on an XScale-based Testbed. Detailed current measurements on this platform demonstrate significant CPU energy savings ranging from 50% to 80% depending on the video clip.

A Semantic Framework for Integrated Asset Management in Smart Oilfields

Integrated asset management (IAM) is the vision of IT- enabled transformation of oilfield operations where information integration from a variety of tools for reservoir modeling, simulation, and performance prediction will lead to rapid decision making for continuous production optimization. This paper describes the design of a model-based IAM system for production forecasting. Domain knowledge is captured through a formal modeling language that forms the basis for an intuitive user interface to the system. An IAM metacatalog captures domain knowledge as well as metadata about computational resources and data sets in a single ontological framework, thereby providing a unified mechanism for application, data, and workflow integration . The framework is designed to be portable across oilfield assets, to allow different classes of end users to interact with the integrated system, and to accomodate new domain knowledge, software applications, data sets, and workflows for IAM.

Semantic web technologies for smart oil field applications

In model based oil field operations, engineers rely on simulations (and hence simulation models) to make important operational decisions on a daily basis. Three problems that are commonly encountered in such operations are: on-demand access to information, integrated view of information, and knowledge management. The first two problems of on-demand access and information integration arise because a number of different kinds of simulation models are created and used. Since these models are created by different processes and people, the same information could be represented differently across models. A unified view of the models and their simulations is desirable for decision making, and thus the necessity for information integration. Knowledge management refers to a systematic way to capture the rationale (knowledge) behind the various analyses performed by an engineer and decisions taken based on the analyses. It is critical to capture this knowledge for auditing, archiving, and training purposes. In this paper, we propose the application of semantic web technologies to address these problems. The key elements of the semantic web approach are the ontologies or the information schemas that model various elements from the domain, and a knowledge base (KB) which is a central repository of the instance information in the system. We present a modular approach for organizing the ontologies and outline the process that was followed to define the ontologies. We also describe the workflow that was used to populate the KB and briefly discuss some of our prototype applications that address the problems mentioned above. Based on our experience, semantic web technologies appear to be a highly promising approach to deal with these information management issues in the oilfield domain, although performance and tool support remain the key areas of concern at this stage.

A Service Oriented Data Composition Architecture for Integrated Asset Management

This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s).

Towards an integrated modeling and simulation framework for freight transportation in metropolitan areas

Freight transportation at distribution nodes such as marine ports, airports and rail yards has been putting tremendous environmental pressure in metropolitan areas. A prerequisite for proposing any solution that would make the existing systems more efficient is an accurate analysis and understanding of freight movements. A single model cannot fully capture aspects of freight transportation which interact and affect each other in a complex manner. Rather, integration of a variety of legacy simulation and analysis tools along with holistic optimization is a necessity for freight transportation system design. This paper proposes an integrated modeling and simulation framework for freight transportation using semantic web technology which offers benefits of modularity, extensibility and reusability of both code and design to the applications. We discuss the implementation strategies and methods to achieve these goals and identify some of the key research challenges in realizing our framework vision.

An Architecture of a Workflow System for Integrated Asset Management in the Smart Oil Field Domain

Integrated asset management (IAM) is the vision of IT-enabled transformation of oilfield operations where information integration from a variety of tools for reservoir modeling, simulation, and performance prediction will lead to rapid decision making for continuous optimization of oil production. In this paper, we discuss the similarities and differences of IAM applications and typical e-Science applications. We then propose an architecture for a workflow system for IAM based on the four key requirements: support for creation, orchestration and management of workflows including those involving legacy tools, support for audit trails and data quality indicators for data objects, usability and extensibility. Our architecture builds upon current research in the scientific workflow area and applies many of its learnings to address the requirements of our system. We propose some implementation strategies and technologies and identify some of the key research challenges in realizing our architectural vision.

A model-based framework for developing and deploying data aggregation services

Data aggregation services compose, transform, and analyze data from a variety of sources such as simulators, real-time sensor feeds, etc. This paper proposes a methodology for accelerating the development and deployment of data aggregation modules in a service-oriented architecture. Our framework allows existing semantic web-service techniques to be embedded into a programming language thereby leveraging ease of use and flexibility enabled by the former with the expressiveness and tool support of the latter. In our framework data aggregations are written as regular Java programs where the data inputs to the aggregations are specified as predicates over a rich ontology. Our middleware matches these data specifications to the appropriate web-service, automatically invokes it, and performs the required data serialization-deserialization. Finally the data aggregation program is deployed as yet another web-service. Thus, our programming framework hides the complexity of web-service development from the end-user. We discuss the design and implementation of the framework based on open standards, and using state-of-art tools.