No abstract

Modeling large-scale stochastic systems of heterogeneous individuals and their interactions, where multiple behaviors and contagions co-evolve with multiple interaction networks, requires high performance computing and agent-based simulations. We present graph dynamical systems as a formalism to reason about network dynamics and list phenomena from several application domains that have been modeled as graph dynamical systems to demonstrate its wide-ranging applicability. We describe and contrast three tools developed in our laboratory that use this formalism to model these systems. Beyond evaluating system dynamics, we are interested in understanding how to control contagion processes using resources both endogenous and exogenous to the system being investigated to support public policy decision-making. We address control methods, such as interventions, and provide illustrative simulation results.

We describe a web services based computational tool for studying large commodity markets. The software architecture is based on three important guiding principles: (i) efficiency and scalability, (ii) extensibility to incorporate several different clearing mechanisms and commodities, (iii) modularity for platform independent operations and use. The computational model has several distinguishing features. They include: (i) the ability to generate individualistic, demographics based, time-varying demand profiles, (ii) a highly configurable system that supports different market clearing mechanisms, strategies and matching algorithms for buyers and sellers (Atkins, Marathe, and Barrett 2007), (iii) ability to aggregate individuals into different hierarchy of classes and (iv) ability to physically clear flow based commodities. From the software standpoint, the architecture has several unique features, including use of web services for loosely coupling individual elements of the system, an easy to use web based graphical user interface for specifying input parameters as well as viewing the results, and a work flow language for modeling various markets and market mechanism (Atkins et al. 2004). The system provides users the unique ability to experiment with a variety of markets such as market for communication spectrum, Internet bandwidth, electricity, as well as traditional commodities like corn, cotton etc. THE MARKET MODELING FRAMEWORKWe describe a web services based computational tool for studying commodity markets. SIGMA (Simulation of Generic Markets) is a service oriented, high fidelity, agent-based, computational market modeling tool. The software architecture is based on three important guiding principles: (i) efficiency and scalability, (ii) extensibility to incorporate several different clearing mechanisms and commodities, (iii) modularity for platform independent operations and use. More specifically, SIGMA functionality is facilitated and enhanced through the following software based features.

No abstract

The need to allow threads to abort an attempt to acquire a lock (sometimes called a timeout) is an interesting new requirement driven by state-of-the-art database applications with soft real-time constraints. This paper presents a new composite abortable lock (CAL), a combination of abortable queue-based (QL) and test-andset based backoff (BL) lock mechanisms, which provides non-blocking aborts while ensuring low space requirements without need for a memory reclamation scheme. The key observation motivating our approach is that the fast lock hand-off achieved by QLs only requires the first few threads to be queued (not all waiting threads), and that the remaining threads can run as in a BL. We developed an algorithm that uses only a short fixed size structure for queueing, allowing most threads to back-off. This reduces worst-case space overhead dramatically, and improves performance by eliminating the need for expensive and complicated memory management mechanisms.Experimental results show that our new CAL algorithm not only saves on space, it actually outperforms Scott's state-of-the-art nonblocking abortable QL under contention, and even more so when there are more threads than processors. Moreover, as the rate of lock aborts increases, the CAL continues to perform well, while Scott's algorithm deteriorates rapidly.