Embedded Gossip: Lightweight Online Measurement for Large-Scale Applications

Abstract

For large-scale parallel applications, lightweight online monitoring can enable a wide range of online adaptations, including load balancing, power management, and progress monitoring. The processing and monitoring overhead of centralized global tracing techniques make them unsuitable for such tasks. Purely local tools, on the other hand, fail to provide the global information necessary for many desirable online adaptations of large-scale applications.In this paper, we describe a novel distributed online measurement method for large-scale applications called Embedded Gossip (EG). EG works by piggybacking performance information about application behavior on existing application messages and merging received information with previously known data in a fashion customized to the needs of a particular monitoring task. EG thus provides each process with both local and global views of application behavior with low overhead.To illustrate the capabilities of Embedded Gossip, we also show that it disseminates global information in a timely fashion for a wide range of monitoring tasks, including critical path profiling, workload imbalance monitoring, and progress monitoring. This global information has a wide range of potential uses, including imbalance detection for load balancing and energy management tools, progress monitoring for batch schedulers, and a wide range of other performance debugging and optimization techniques.

Online Performance Data Introspection with IPM

“…The Distributed Performance Consultant in Paradyn [12] in conjunction with MRNet [13] provides a scalable framework for online automated performance diagnosis. Embedded Gossip [14] spreads performance data among processes including it transparently into the MPI messages sent by the application. Although these tools manage performance data online, only a few of them such as TAUg or TAUoverSupermon provide online access to this performance data as the Performance Introspection API does.…”

Autonomous aggregation in location aware ad hoc wireless networks

“…Decentralized methods for consensus problem that are known to work well in a relatively static environment, for example, for parallel applications , are studied in . One approach to improving fault tolerance in dynamic distributed systems, which is measured by the maximum radius of impact caused by a given fault, is presented by S. Pike .…”

Self-organized aggregation in irregular wireless networks

“…Decentralized methods for consensus problem that are known to work well in a relatively static environment, e.g., for parallel applications [25], are studied in [5], [19], [20]. One approach to improving fault tolerance in dynamic distributed systems, which is measured by the maximum radius of impact caused by a given fault, is presented by S. Pike [21].…”