Dynamic threshold for monitor systems on grid service environments

Abstract

Grid technology requires use of geographically distributed multiple domain's resources. Resource monitoring services or tools consisting sensors or agents will run on many systems to find static resource information (such as architecture vendor, OS name and version, MIPS rate, memory size, CPU capacity, disk size, and NIC information) and dynamic resource information (CPU usage, network usage(bandwidth, latency), memory usage, etc.). Thus monitoring itself may cause system overhead. This paper proposes push based resource notification architecture on OGSI (Open Grid Service Infrastructure) and the dynamic threshold to measure monitoring events in accurate and with less overhead. By employing the new feature (dynamic threshold), we find out unnecessary system overhead is significantly reduced and accuracy of events is still acquired. © Springer-Verlag 2004.