Support for disconnected operation via architectural self-reconfiguration

Abstract

In distributed and mobile environments, the connections among the hosts on which a software system is running are often unstable. As a result of connectivity losses, the overall availability of the system decreases.

Improving availability of distributed event-based systems via run-time monitoring and analysis

“…The monitoring information is then provided to our DeSi tool for system visualization and analysis. We have developed several algorithms that analyze and estimate improvements in deployment architectures [5,8]. Finally, after using DeSi to select the desired deployment architecture, Prism-MW facilities are leveraged to effect the selected deployment architecture.…”

“…With the help of DistributionConnectors, AdminComponent and DeployerComponent are able to send and receive from any device to which they are connected the 1. A detailed description of the redeployment problem is given in [8]. Network Reliability Monitor events that contain application-level components (sent between address spaces using the Serializable interface).…”

“…DeSi allows an engineer to rapidly explore the space of possible deployments for a given system (real or postulated), determine the deployments that will result in greatest improvements in availability (while, perhaps, requiring the smallest changes to the current deployment architecture), and assess a system's sensitivity to and visualize changes in specific parameters (e.g., the reliability of a network link) and deployment constraints (e.g., two components must be located on different hosts). DeSi allows one to easily integrate, evaluate, and compare different algorithms targeted at improving system availability [8] in terms of their feasibility, efficiency, and precision. As will be detailed in the remainder of this section, we have extended DeSi to allow its integration with any distributed middleware platform that supports system monitoring and runtime component deployment.…”

“…In the Constraints panel, the user can specify different constraints on component locations (e.g., fixing a component to a selected host). Using the set of buttons in the Algorithms panel, different algorithms [5,8] can be invoked and the results displayed in the Results panel. Finally, the user can modify the desired system parameters (by editing the appropriate tables in Tables of parameters panel) to assess the sensitivity of a deployment architecture to parameter changes.…”

“…However, finding the actual deployment architecture that maximizes a system's availability is an exponentially complex problem that may take years to resolve for any but very small systems [8]. Also, even a deployment architecture that increases the system's current availability by a desired amount cannot be easily found because of the many parameters that influence this task: number of hardware hosts, available memory and CPU power on each host, network topology, capacity and reliability of network links, number of software components, memory and processing requirements of each component, their configuration (i.e., software topology), frequency and volume of interaction among the components, and so forth.…”