Shivashis Saha

Thermal-Constrained Energy-Aware Partitioning for Heterogeneous Multi-core Multiprocessor Real-Time Systems

Next-generation multi-core multiprocessor real-time systems consume less energy at the cost of increased power density. This increase in power-density results in high heat density and may affect the reliability and performance of real-time systems. Thus, incorporating maximum temperature constraints in scheduling of real-time task sets is an important challenge. This paper investigates thermal-constrained energy-aware partitioning of periodic real-time tasks in heterogeneous multi-core multiprocessor systems. We adopt a power model which considers the impact of temperature and voltage on a processors static power consumption. Two types of thermal models are used to respectively capture negligible and non-negligible amount of heat transfer among cores. We develop a novel genetic-algorithm based approach to solve the heterogeneous multi-core multiprocessor partitioning problem. Extensive simulations were performed to validate the effectiveness of the approach. Experimental results show that integrating a worst-fit based partitioning heuristic with the genetic algorithm can significantly reduce the total energy consumption of a heterogeneous multi-core multiprocessor real-time system.

HyScaleII: A high performance hybrid optical network architecture for data centers

Tremendous growth in data-intensive cloud applications have resulted in an increased demand for highly scalable data center network (DCN) architectures with high throughput and low network complexity. In this paper, we propose HyScaleII to improve the performance of HyScale [17]. HyScaleII is a switch-centric high performance hybrid optical network based DCN architecture that has most of the desirable properties of a data center, e.g. high scalability, low diameter, high bisection width, fault-tolerance, and low network complexity. We also present an efficient and simple routing scheme called HySII routing, which exploits the structural properties of HyScaleII. In our experiments, HyScaleII has lower packet loss ratio and higher average aggregate throughput by an average of 50% and 13.8% respectively as compared to HyScale [17].

Energy models driven green routing for data centers

Energy consumption is becoming a serious issue for data centers. Minimizing the energy consumption of a data center is a complex problem. It consists of developing energy models for servers and network elements and investigating energy-aware (green) routing strategies using these models. In this paper, we adapt energy models for servers and network elements in a data center. The server energy model incorporates the impact of temperature and voltage of a server on its leakage energy. Using the server and network energy models, we propose a green routing scheme that minimizes the total combined energy consumption of servers and network elements in a data center under dynamic traffic. The proposed green routing scheme uses dynamic voltage scaling, rate adaptation, and anycast transmission for minimizing the total energy consumption. Extensive simulation results validate the effectiveness of the proposed green routing in minimizing the total energy consumption of a data center as compared to well-known existing approaches.

HyScale: A hybrid optical network based scalable, switch-centric architecture for data centers

Data Center Network architectures (DCN) are evolving for increased scalability, performance, and low network complexity. In this paper, we propose HyScale, a switch-centric DCN architecture using hybrid optical networks. HyScale employs Optical Burst Switching and Optical Circuit Switching technologies for transmitting low and high volumes of data respectively in a data center. The proposed architecture is highly scalable, recursively defined, fault-tolerant, and has low network complexity. It also has a multitude of desirable graph-theoretic properties like high bisection width, and low diameter. By exploiting the structural properties of HyScale, we propose a highly efficient and simple routing scheme. In our experiments, the proposed routing scheme gives a lower packet loss ratio by an average of 23% as compared to the Shortest Path Routing with almost negligible increase in the length of the routes.

Minimizing network cost in all-optical networks

The problem of minimizing the total network cost of an optical network topology by efficient selection of switching sites, size of optical switches, and optical links is investigated in this paper. The problem investigated is NP—hard. Therefore, we develop an efficient heuristic to approximate the solution in polynomial time. A mixed integer quadratic programming (MIQP) formulation of the problem is also presented to find the optimal network cost and compute the efficiency of the heuristic. The total network cost calculated by the heuristic in the experiments is within 19% of its optimal value. Moreover, the total network cost in half of the test problems is within 6% of its optimal value. The heuristic solves the problem with 20 node topologies in less than a second. However, the commercial optimization software can not solve any problem with more than 10 nodes even in two weeks.

Adaptive energy-efficient task partitioning for heterogeneous multi-core multiprocessor real-time systems

The designs of heterogeneous multi-core multiprocessor real-time systems are evolving for higher energy efficiency at the cost of increased heat density. This adversely effects the reliability and performance of the real-time systems. Moreover, the partitioning of periodic real-time tasks based on their worst case execution time can lead to significant energy wastage. In this paper, we investigate adaptive energy-efficient task partitioning for heterogeneous multi-core multiprocessor realtime systems. We use a power model which incorporates the impact of temperature and voltage of a processor on its static power consumption. Two different thermal models are used to estimate the peak temperature of a processor. We develop two feedback-based optimization and control approaches for adaptively partitioning real-time tasks according to their actual utilizations. Simulation results show that the proposed approaches are effective in minimizing the energy consumption and reducing the number of task migrations.

Design of an all-optical WDM lightpath concentrator

A concentrator is a switch in which the number of input fibers is strictly greater than the number of output fibers. Thus, a concentrator drops some input signals and switches each of the remaining signals to output fibers on possibly different wavelengths. If the switching is done in the electronic domain, it is called an electronic concentrator or simply a concentrator. If the switching is performed in all-optical domain, then it is called an all-optical lightpath concentrator or simply lightpath concentrator. With the advancement of wavelength division multiplexing (WDM) technology, there has been an exponential growth in the size of interconnects, their complexity, the number of necessary optical-electrical-optical (O/E/O) converters, and their cost. There is a large mismatch between the capacity and speed of optical transmission compared with electronic transmission which results in a bottleneck of the electronic switches and concentrators in WDM networks. This mismatch motivates design of all-optical switches and lightpath concentrators which can keep data in the optical domain, eliminating the need for costly and inefficient O/E/O conversions. There are several practical deployments of optical networks where it is necessary to switch signals from N fibers to M fibers, where N > M [1]. For example, the function of a hub in metro networks is to collect N lightpaths from several regional hubs and transfer them to M lightpaths in the backbone network. Thus, a hub is a lightpath concentrator [1]. Lightpath concentrators have also been used in the design of all-optical switch architectures [2], [3], [4]. With the advancement of WDM technologies, there has been a need for cost effective WDM interconnects. WDM crossconnects typically have separate stages for space and wavelength switching. These designs are expensive, and the switching costs increase significantly with the increase in the number of lightpaths carried by the fiber. A new paradigm for WDM interconnect design is evolving in which space and wavelength switching is performed simultaneously and seamlessly [2], [3]. This design for WDM crossconnects eliminates the need for separate wavelength conversion stages. In this paper, we present a design of an all-optical lightpath concentrator based on wavelength exchanging optical crossbars (WOC) and WDM Crossbar Switches [2], [3]. The design of an all-optical lightpath concentrator presented in this paper makes use of WOC properties and can similarly achieve a balance between wavelength switching and conversion costs and does not require a separate wavelength conversion stage.

A novel thermal-constrained energy-aware partitioning algorithm for heterogeneous multiprocessor real-time systems

Next-generation multiprocessor real-time systems consume less energy at the cost of increased power density. This increase in power density results in high heat density and may affect the reliability and performance of real-time systems. Thus, incorporating maximum temperature constraints in scheduling of real-time task sets is an important challenge. This paper investigates a novel algorithm for thermal-constrained energy-aware partitioning of periodic real-time tasks in heterogeneous multiprocessor systems. When designing our new algorithm, we have applied insights gained from a famous knapsack problem solution. Both simulation and experimental results show that our new branch-and-bound based partitioning algorithm can significantly reduce the total energy consumption of multiprocessor real-time systems.

Hybrid all-optical networks: Routing and wavelength assignment

Hybrid all-optical networks that integrate OCS and OBS technologies have been recently proposed to facilitate next-generation applications like cloud and grid computing. In this paper, we propose an ILP formulation and a heuristic based on genetic algorithm for routing and wavelength assignment problem in hybrid all-optical networks. The proposed model aims to minimize wavelength contentions by minimizing the overlapping paths in the hybrid networks. Experimental simulations show that the heuristic effectively minimizes the wavelength contention in the hybrid networks.

Minimizing Resource Blocking Rate in GoOBS

The state of the resources at a destination in Grid computing over OBS architecture (GoOBS) may change between a tasks selection of a destination and its arrival at the destination. These changes in the availability of the resources requested at the destination may lead to blocking of tasks, and thus increase the resource blocking rate. In this paper, we investigate the resource scheduling problem in GoOBS. Our objective is to minimize the resource blocking rate by containing the impact of the changes in the availability of the resources at a destination. We propose a non-selfish destination selection paradigm to minimize the resource blocking rate. The selection of a destination by a request is called non-selfish, if the selected destination has sufficient resources available to simultaneously process one or more additional requests. Extensive simulations were performed to validate the effectiveness of the heuristics based on the non-selfish destination selection paradigm. Among the proposed heuristics, the NFFD heuristic is most effective in minimizing the resource blocking rate. Compared to the best existing approach, the NFFD heuristic reduces the resource blocking rate by 21% to 73% in our experiments.