Several characteristics distinguish today's complex software systems from "traditional" ones. Examples in different areas show that these characteristics, already the focus of agent-oriented software engineering research, influence many application domains. These characteristics will impact how software systems are modeled and engineered.

Adaptive Mesh Refinement (AMR) is a type of multiscale algorithm that achieves high resolution in localized regions of dynamic, multidimensional numerical simulations. One of the key issues related to AMR is dynamic load balancing (DLB), which allows large-scale adaptive applications to run efficiently on parallel systems. In this paper, we present an efficient DLB scheme for Structured AMR (SAMR) applications. Our DLB scheme combines a grid-splitting technique with direct grid movements (e.g., direct movement from an overloaded processor to an underloaded processor), for which the objective is to efficiently redistribute workload among all the processors so as to reduce the parallel execution time. The potential benefits of our DLB scheme are examined by incorporating our techniques into a parallel, cosmological application that uses SAMR techniques. Experiments show that by using our scheme, the parallel execution time can be reduced by up to 47 % and the quality of load-balancing can be improved by a factor of four. 1

Load balancing is a key issue in the development of parallel algorithms with irregular structures. Existing load balancing systems each support only one specific programming paradigm and thus are of limited use. The system VDS presented here allows concurrent use of various paradigms such as fork-join, weighted tasks, and static dags (directed acyclic graphs that are known in advance). The system provides visual performance evaluation tools to facilitate the efficient application of the system. VDS supports various communication interfaces including PVM and MPI. Thus, VDS-applications can be run on architectures ranging from workstation clusters to massively parallel systems.

In this paper, we identify and analyze a set of characteristics that increasingly distinguish today's complex software systems from "traditional" ones. Several examples in different areas show that these characteristics are not limited to a few application domains but are widespread. Then, we discuss how these characteristics are likely to impact dramatically the very way software systems are modeled and engineered. In particular, we appear to be on the edge of a radical shift of paradigm, about to change our very attitudes in software systems modeling and engineering. Keywords: Distributed Systems, Design Paradigms, Multiagent Systems. To be published in: The Knowledge Engineering Review Corresponding Author: Franco Zambonelli 2 1

Adaptive Mesh Refinement (AMR) is a type of multiscale algorithm that dynamically achieves high resolution in localized regions of multidimensional numerical simulations. A dynamic load balancing(DLB) scheme for structured AMR applications was proposed in [19]. Unfortunately, the overhead introduced by this DLB scheme is significant. Further, a parameter called threshold is used in this scheme which determines whether a load balancing process should be invoked; its value directly influence the efficiency of the overall DLB scheme. In this paper, we first present two improvements on this DLB scheme to reduce its overhead. Then a detailed sensitivity analysis is provided to identify an optimal value for the parameter threshold. Experiments show that by interleaving grid splitting with direct grid movement and by employing nonblocking communication, the execution time can be improved by up to and the overhead can be reduced as much as �;andsetting threshold to be around � � is optimal. KEY WORDS dynamic load balancing, adaptive mesh refinement, sensitivity analysis 1.

Although a large number of users are using P2P systems, the ability of these systems to provide services with quality is questioned. A load balanced P2P system can provide services with smaller failure rate and better performance; hence, service quality of the system can be improved. Cluster systems have been adopted for services which are tolerant to faults. Although a cluster structure could be adopted for improving the reliability and robustness of a P2P system, the load unbalancing problem still remains because of the heterogeneities of nodes and requests. Existing dynamic load balancing mechanisms in P2P systems require extra connections and overhead on aggregating the load status from the nodes. We propose diffusive load balancing in a clustered P2P systems, where a global balance is achieved through balancing the neighbourhoods of all clusters within the existing overlay network. We simulated three load balancing schemes: directory-initiated, sender-initiated, and receiver-initiated; from an initially unbalanced situation, the results show that diffusive load balancing can achieve a global balance comparable to a centralized directory scheme, and the distributed directory-initiated scheme provides better results than the sender- or receiver-initiated schemes. 1.

Abstract. IP-networked streaming media storage has been increasingly used as a part of many applications. Random placement of data blocks has been proven to be an effective approach to balance heterogeneous workload in multi-disk steaming architectures. However, the main disadvantage of this technique is that statistical variation can still result in short term load imbalances in disk utilization. We propose a packet level randomization (PLR) technique to solve this challenge. We quantify the exact performance trade-off between PLR approach and the traditional block level randomization (BLR) technique through both theoretical analysis and extensive simulation. Our results show that the PLR technique can achieve much better load balancing in scalable streaming architectures by using more memory space. 1

Abstract This paper proposes a framework to translate certain subclasses of differential equation systems into practical protocols for distributed systems. The generated protocols are intended for large-scale distributed systems that contain several hundreds to thousands of processes. The synthesized protocols are state machines containing probabilistic transitions and actions, and they are proved to show equivalent stochastic behavior to the original equations. The protocols are probabilistically scalable and reliable, and have practical applications in large-scale distributed systems, e.g., peer to peer systems. In order to illustrate the usefulness of the framework, it is used to generate new solutions for the problems of (i) responsibility migration (giving rise to a novel model of dynamic replication), and (ii) majority selection. We present mathematical analysis of these two protocols, and experimental results from our implementations. These two protocols are derived from natural analogies that are represented as differential equations- endemics and the Lotka-Volterra model of competition respectively. We believe the design framework could be effectively used in transforming, in a very systematic manner, well-known natural phenomena into protocols for distributed systems.

Petascale machines with hundreds of thousands of cores are being built. These machines have varying interconnect topologies and large network diameters. Computation is cheap and communication on the network is becoming the bottleneck for scaling of parallel applications. Network contention, specifically, is becoming an increasingly important factor affecting overall performance. The broad goal of this dissertation is performance optimization of parallel applications through reduction of network contention. Most parallel applications have a certain communication topology. Mapping of tasks in a parallel application based on their communication graph, to the physical processors on a machine can potentially lead to performance improvements. Mapping of the communication graph for an application on to the interconnect topology of a machine while trying to localize communication is the research problem under consideration. The farther different messages travel on the network, greater is the chance of

Abstract—Node clustering is an effective solution for achieving good performance and high reliability for peer-topeer (P2P) systems. To improve the performance of a clustered P2P system, it is important to balance the service load among the clusters in the system. In this paper, we describe a diffusive load balancing scheme for clustered P2P systems, which dynamically adjusts the size of the clusters, by moving nodes among the clusters, based on their service demands and node resource capacities. Our simulations show that the proposed load balancing scheme significantly improves the performance of a P2P system in terms of balanced available capacity. Keywords-Load balancing; diffusive load balancing; peerto-peer systems; distributed algorithms; dynamic resource allocation; performance management; server clusters; clustered peer-to-peer systems I.