Approximate performance/cost analysis of m-level hierarchical multiprocessor systems

In previous work, a class of cluster-based multiprocessor systems was introduced and analyzed. The analysis was based on an approximate probabilistic model and assumed circuit-switching methodology. It showed that for a higher rate of favourite requests the cluster-based system performs fairly close to the crossbar system and outperforms the typical multiple-bus system (with buses equal to half the number of processors). In this paper we first simulate the cluster-based system based on the analytical model assumptions presented in our earlier work. We then develop a simulation model that relaxes two unrealistic assumptions of the analytical model. The results obtained from the simulation models are compared with those obtained from the approximate analytical model for the cluster-based, crossbar, and multiple-bus systems. We then introduce a packet-switched cluster-based system. The system uses packet-switched single-shared buses to increase the bus utilization and hence the system throughput. A simulation model based on the event-scheduling approach is developed to evaluate the performance of the packet-switched cluster- based system in terms of delay and throughput. The simulation results of the packet-switched cluster-based system are compared with those obtained for the crossbar and multiple-bus systems for both uniform and favourite reference models.

Large-scale MIN-based shared-memory multiprocessor systems have long shared memory latency. Private caches can improve memory access latency but they may suf fer from the cache coherence problem and potentially lower data locality due to data sharing and multiproces sor scheduling. These two problems also increase shared memory load and may result in frequent memory stalls. In this paper, we evaluate the performance of memory caches, a cache memory placed in front of shared memory, in a large-scale multiprocessor system in the presence of pro cessor caches. The memory cache is shown to have good performance and scalability.

As faults are unavoidable in large scale multiprocessor systems, it is important to be able to determine which units of the system are working and which are faulty. System-level diagnosis is a long-standing realistic approach to detect faults in multiprocessor systems. Diagnosis is based on the results of tests executed on the system units. In this work we evaluate the performance of evolutionary algorithms applied to the diagnosis problem. Experimental results are presented for both the traditional genetic algorithm (GA) and specialized versions of the GA. We then propose and evaluate specialized versions of Estimation of Distribution Algorithms (EDA) for system-level diagnosis: the compact GA and Population-Based Incremental Learning both with and without negative examples. The evaluation was performed using four metrics: the average number of generations needed to find the solution, the average fitness after up to 500 generations, the percentage of tests that got to the optimal solution and the average time until the solution was found. An analysis of experimental results shows that more sophisticated algorithms converge faster to the optimal solution.

Directory-based cache coherence scheme using number-balanced binary tree

The drastically reduced costs and improved performance of fiber optic components in recent years offers an opportunity to economically incorporate high-bandwidth optical communication links within a large-scale multiprocessor system. In this paper, the design, implementation, and initial performance of a fiber optic bus for a tightly coupled multiprocessor system are discussed. This optical bus ulitizes a multiple fiber-multiple star configuration with data rates of up to 80 MHz. It transmits bus control line information by repetitively scanning the control lines and transmitting their states as a packet of bits. It is currently being used for a 30-processor system called the Heidelberg Polyp, which has a custom bus design. However, the design approach is a general one that could easily be adapted to other buses, especially asynchronous buses such as the VME bus.

A cluster-based multiprocessor system that was introduced and analyzed by D. Agrawal and I. Mahgoub (1985, 1987) is considered. The analysis was based on an approximate analytical model, and the results were compared with those obtained for the crossbar and multiple-bus systems using the same model. A simulation model that relaxes two unrealistic assumptions of the approximate analytical model used by Agrawal and Mahgoub is developed here. The results obtained from the simulation model are compared with those obtained from the approximate analytical model for the cluster-based, crossbar, and multiple-bus systems. The simulation results confirm the conclusions arrived at in the previous studies.

The core of current-generation high-performance multiprocessor systems is out-of-order execution processors with aggressive branch prediction. Despite their relatively high branch prediction accuracy, these processors still execute many memory instructions down mispredicted paths. Previous work that focused on uniprocessors showed that these wrong-path (WP) memory references may pollute the caches and increase the amount of cache and memory traffic. On the positive side, however, they may prefetch data into the caches for memory references on the correct-path. While computer architects have thoroughly studied the impact of WP effects in uniprocessor systems, there is no comparable work for multiprocessor systems. In this paper, we explore the effects of WP memory references on the memory system behavior of shared-memory multiprocessor (SMP) systems for both broadcast and directory-based cache coherence. Our results show that these WP memory references can increase the amount of cache-to-cache transfers by 32%, invalidations by 8% and 20% for broadcast and directory-based SMPs, respectively, and the number of writebacks by up to 67% for both systems. In addition to the extra coherence traffic, WP memory references also increase the number of cache line state transitions by 21% and 32% for broadcast and directory-based SMPs, respectively. In order to reduce the performance impact of these WP memory references, we introduce two simple mechanisms—filtering WP blocks that are not likely-to-be-used and WP aware cache replacement—that yield speedups of up to 37%.

Disaster risk management (DRM) is undergoing noteworthy changes, reflecting the broader shifts in global and local levels of governance. At the global level two significant changes are of interest: (1) the shift from monolithic structures of global governance to a wide range of organizations that can be brought together for specific purposes and (2) the emergence of a globalized system of DRM, with technological, organizational, and institutional capacities enhancing DRM’s ability as a unit in near real time across the globe. At the local level there is an increase in ability to govern and develop creative solutions for complex problems that follow rapid urbanization. The importance of getting the global–local interface to work in tandem has been highlighted by recent hazard events, such as the 2011 Tohoku Earthquake and Tsunami. From a broad view of global and local shifts, a strategic role is becoming clearer at the national level for enhancing the relationships between the global and local levels. Through the influence of a globalized system of DRM, the local level can significantly improve its capacity without the heavy investment that might have been required to develop these capacities in isolation. One key to achieving this is a diffusion of DRM higher education, supported by an enhanced system of information flow.

Additional investigation of the optimization of cross-sectional properties for rotor blades was performed with methodology based on the Genetic Algorithm (GA). In previous work by the authors, a multilevel optimization was conducted using gradient-based methods that are appropriate for continuous variables. In contrast to that work, here optimization is conducted with the main focus of assessing the impact of discrete variables that are inherent to the problem at both local and global levels. The GA was modified based on existing routines in MATLAB to make the code suitable for the methodology presented and to utilize interfaces with the multi-body dynamics program DYMORE, used for global analysis, and the cross-sectional analysis program VABS, used at the local level. At the first exploratory phase of the optimization, values of discrete variables are determined using the GA along with several promising regions for global minima that serve as starting points for the second phase optimization, which deals only with continuous variables. Furthermore, if more than one optimal design exists, further analysis is performed to pick the best design from among candidate configurations. ptimization of the cross-sectional layout of composite rotor blades is a challenging problem because of the large number of design parameters and the wide-ranging impact of those parameters on overall vehicle characteristics. The availability of high-fidelity structural beam analysis tools naturally allows a hierarchical decomposition of the problem that separates the local (cross-sectional) sub-problem from the global optimization. By encapsulating the intricate details of the cross-sectional design into a relatively small number of equivalent beam properties, one can significantly reduce the overall complexity of the optimization. An additional advantage of such a decoupled approach is that it follows a venerable tradition of analyzing rotorcraft systems by splitting three-dimensional (3D) problems into separate sub-problems over the crosssection (2D) and span (1D). Modern computer tools for analyzing composite rotor blades with arbitrary cross-sectional configurations such as VABS 1, 2 provide a fast and accurate means of assessing both static and dynamic structural properties of rotor blades in terms of an equivalent beam models. Calculation of the cross-sectional stiffness matrix is conducted using a finite element discretization. Results of this calculation also include a set of cross-sectional stress/strain recovery relations. Once the equivalent cross-sectional stiffnesses are obtained, a beam analysis can be conducted very rapidly, for various loading and boundary conditions as appropriate. Results from the 1D analysis can then be fed back into 2D recovering relations to obtain detailed stress and strain components over the cross-section. The overall analysis possesses accuracy comparable to that of 3D finite element methods with a reduction in computational cost by two to three orders of magnitude. 3 Optimization based on a parametric cross-sectional model was considered in Ref. 4. The layout was sufficiently realistic to allow optimization of a baseline cross-section in which the ply angles, thicknesses, and spar location were varied in order to place the shear center in a specified location while holding the crosssectional stiffnesses within a specified tolerance. This work was focused on the lower level of optimization and was continued with a still higher level of realism in Ref. 5.

This paper describes a view-based outdoor navigation method. Navigation in outdoor can be divided into two levels; the global level deals with localization and subgoal selection, while the local level deals with safe navigation in a local area. We adopt an improved version of SeqSLAM method for global-level localization, which can cope with changes of robot’s heading and speed as well as view changes using very wide-angle images and a Markov localization scheme. The global level provides the direction to move and the local level repeatedly sets subgoals with local mapping using 3D range sensors. We implemented these global and local level methods on a mobile robot and conducted on-line navigation experiments.

The recent rise of jihadist movements in West Africa, including Boko Haram in the Lake Chad region and Al-Qaeda in the Islamic Maghreb and its affiliates in the Sahel-Saharan region, has puzzled many observers. The easy spread of the jihadist ideology, the jihadist movements’ success in massively recruiting followers among local populations as well as their ability to conquer and administer territories, are unprecedented in the region’s contemporary history. This paper sheds light on the factors and processes that contribute to the emergence of these movements. It argues that the phenomenon of jihadist insurgencies in West Africa emerges as a result of a series of processes at the global, local, and individual level. At the global level, there is the formation and dissemination of the global ideology of jihadism, conceptualised by Muslim activists and scholars based on a particular understanding of Islam and the challenges that are facing contemporary Muslim societies. At the local level, the appropriation of jihadist ideologies by “Muslim activists” who then use it to formulate a discourse which taps into local social and political demands in order to mobilise followers, is key. For a wide range of reasons, certain regions of Africa have experienced weakened state capacity and increased local conflict, and it is in these areas that jihadist insurgencies have emerged. At the individual level, the process by which African individuals decide to enrol in jihadist groups include ideological, situational, and strategic motivations, and these have all been facilitated by deteriorating conditions of life in marginalised areas.

Former President Nelson Mandela, Bono, Peter Gabriel and other superstars stood together on the stage at Greenpoint Stadium in Cape Town in front of billions of television viewers around the world, watching the “46664” music extravaganza in support of the fight against AIDS in Africa. AIDS is clearly a global pandemic and responses to it have inevitably been on a global scale. At the same time, the disease has highly localised aspects to it. AIDS activists have had to address both the global dimensions and the local specificities of this epidemic. This article focuses on new conceptions and arenas of civic action promoted by a Cape Townbased AIDS activist group, the Treatment Action Campaign (TAC), and the new forms of citizenship that emerge from their engagement with AIDS treatment policy. TAC is an example of a “new” social movement that has constructed its own arenas of action and spaces of participation in response to state-driven AIDS policies. Unlike the formal, hierarchical structure and conventional class politics and Marxism of the “old” political and labour movements, TAC, like other new social movements, draws on grassroots, bottom-up, network based modes of organisation that operate simultaneously in local, national and global spaces (Touraine 1981; Melucci 1989). While it is problematic to assert a strict divide between “old” and “new” social movements, organisations such as TAC have introduced tactical and organisational innovations that take advantage of the increased global reach and instantaneity of the media, the Internet, email and other circuits of telecommunications (Cohen and Rai 2000; Wasserman 2003). While TAC speaks to audiences well beyond the border of the nation– state, its main objective has been to lobby and pressurise the South African government to provide AIDS treatment. TAC has been forced to address a wide range of issues at global, national and local levels: it has tackled the global pharmaceutical industry in the media, the courts and the streets; it has fought discrimination against HIV-positive people in schools, hospitals and at the workplace; it has challenged AIDS dissident science in South Africa and abroad; and taken the government to court for refusing to provide prevention of mother-to-childtransmission treatment (PMTCT) programmes in public health facilities. In addition to these high profile activities, TAC has also launched AIDS literacy campaigns in black townships throughout South Africa challenging AIDS myths, silence, denial and misinformation (see Robins, forthcoming). This article shows how these diverse TAC activities and interventions have contributed towards creating new political spaces for engagement at local, national and global levels. TAC relies both on transnational advocacy networks and grassroots mobilisation in ways that are similar to modes of activism that are increasingly described as ‘globalisation from below’ or ‘grassroots globalisation’ (Appadurai, 2002a). TAC also provides examples of organisational practices that cut across institutional and noninstitutional spaces, and that are capable of generating multiple relations to the state. In doing

Issues raised in the authors’ previous works are further explored, introducing a multilevel, multiphase approach with a surrogate model. Previous approaches did not show a clear indication of the communication between global and local levels because both levels reached the optimal solution without iteration between levels. To guarantee the need for an iterative process, the blade model was changed, and surrogate models were embedded in the procedure to ensure the communication between global and local levels. Concomitantly, values of the design variables were forced to stay in the feasible design space. Candidate surrogate models include Kriging, and linear and quadratic Response Surface Regression (RSM). These models are compared and the most suitable surrogate model was adopted to finalize the iterative process of the methodology. The new example is the Elastic Articulation (EA) rotor system which is the rotor system of the Generic Georgia Tech Helicopter. The EA rotor is a soft-in-plane bearingless rotor with 10% effective flapping hinge offset and forward sweep. To ensure the dynamic stability of such a system, the methodology utilizes high-fidelity blade structural analysis tools DYMORE and VABS, which reduces the computation time by allowing hierarchical decomposition of the problem. This separates the local (cross-sectional) sub-problem, with constraints on local quantities as maximum stresses, from the global optimization, with constraints on global parameters such as on the Autorotation Index (AI). The procedure is implemented using MATLAB as the overall optimizer.

Through analysis and experiments, this paper investigates two-phase waiting algorithms to minimize the cost of waiting for synchronization in large-scale multiprocessors. In a two-phase algorithm, a thread first waits by polling a synchronization variable. If the cost of polling reaches a limit L poll and further waiting is necessary, the thread is blocked, incurring an additional fixed cost, B . The choice of L poll is a critical determinant of the performance of two-phase algorithms. We focus on methods for statically determining L poll because the run-time overhead of dynamically determining L poll can be comparable to the cost of blocking in large-scale multiprocessor systems with lightweight threads. Our experiments show that always-block ( L poll = 0) is a good waiting algorithm with performance that is usually close to the best of the algorithms compared. We show that even better performance can be achieved with a static choice of L poll based on knowledge of likely wait-time distributions. Motivated by the observation that different synchronization types exhibit different wait-time distributions, we prove that a static choice of L poll can yield close to optimal on-line performance against an adversary that is restricted to choosing wait times from a fixed family of probability distributions. This result allows us to make an optimal static choice of L poll based on synchronization type. For exponentially distributed wait times, we prove that setting L poll = 1n(e-1) B results in a waiting cost that is no more than e/(e-1) times the cost of an optimal off-line algorithm. For uniformly distributed wait times, we prove that setting L poll =1/2(square root of 5 -1) B results in a waiting cost that is no more than (square root of 5 + 1)/2 (the golden ratio) times the cost of an optimal off-line algorithm. Experimental measurements of several parallel applications on the Alewife multiprocessor simulator corroborate our theoretical findings.

Although considerable attention has been given to the definition of topic from both a global and local level, most of this work has been done entirely from the perspective of researchers. We examined perceptions of global and local topic change from the point of view of conversation participants. The results indicate that participants readily agree on points of topic change at the global level, but demonstrate relatively little agreement at the local level. These results were discussed in terms of the current debate over global versus local topic definition as well as other implications for future research.