Abstract: The present disclosure is directed to a system for monitoring and analyzing operation of a widely distributed service operated by an Infrastructure-as-a-Service (IaaS) tenant but deployed on a set of virtual resources controlled by an independent IaaS provider. The set of virtual resources provided to the IaaS tenant by the IaaS provider is selected by the IaaS provider and can change rapidly in both size and composition (i.e., the virtual resources are “ephemeral”). The monitoring system can integrate system-level metrics collected directly from virtual resources with infrastructure metadata characterizing the virtual resources collected from the IaaS provider to report on operation of the virtual resources. The infrastructure metadata can contain a resource type, a resource role, an operational status, an outage history, or an expected termination schedule. By integrating different sources of data, the monitoring system can avoid inaccurate analysis.

Abstract: The present disclosure is directed to a monitoring system for rapidly updating a service architecture of a widely distributed service operated by an Infrastructure-as-a-Service (IaaS) tenant but deployed on a set of virtual resources controlled by an independent IaaS provider. The set of virtual resources provided to the IaaS tenant by the IaaS provider is selected by the IaaS provider and can change rapidly in both size and composition (i.e., the virtual resources are “ephemeral”). The monitoring system can infer from infrastructure metadata and/or system-level metric data how the virtual resources should be organized into groups, clusters and hierarchies. The monitoring system can also update the service architecture frequently to capture an expected rate of change of the resources, e.g., every five minutes. The monitoring system can then run analytics based on this inferred service architecture to report on service operation.

Abstract: A method, system and computer program product for detecting conditions of network instability. An attribute indicative of network instability, e.g., processor and/or co-processor utilization, packet arrival rates, packet peak rates, packet size distribution, packet clustering tendencies, buffer usage patterns, occurrence of peak utilization, out-of-buffer conditions, packet discard rates, may be monitored for a network device, e.g., router. The monitored attribute may be associated with a plurality of labels (variables) where a portion of those labels may be stored in a cache. A hit ratio for the cache storing labels associated with the monitored attribute may be tracked within a period of time based on the number of requested items, i.e., labels, that are currently stored in the cache. A condition of network instability may then be detected based on the hit ratio by determining if the absolute value of the acceleration of the hit ratio exceeds a threshold.

Abstract: The present invention provides an improved method and system for managing peripheral bus timeouts. The method includes: sending a request to a peripheral in a performance of a task; determining that a peripheral bus timeout occurred; sending an interrupt signal; retrying for an access to the peripheral; completing the request if the retry for the access to the peripheral is successful; and resuming the performance of the task. The present invention provides a software method in which an input/output (I/O) request to a peripheral is sent by a recoverable access routine. When a bus timeout occurs, a hung bus detector asserts an interrupt signal. The peripheral bus is freed from the hung condition. The recoverable access routine is suspended, and a bus timeout interrupt handler is then loaded. The interruption and the freeing of the peripheral bus frees the microprocessor from the processing of the I/O request so that other requests may be performed.

Abstract: A method, system and computer program product for detecting conditions of network instability. An attribute indicative of network instability, e.g., processor and/or co-processor utilization, packet arrival rates, packet peak rates, packet size distribution, packet clustering tendencies, buffer usage patterns, occurrence of peak utilization, out-of-buffer conditions, packet discard rates, may be monitored for a network device, e.g., router. The monitored attribute may be associated with a plurality of labels (variables) where a portion of those labels may be stored in a cache. A hit ratio for the cache storing labels associated with the monitored attribute may be tracked within a period of time based on the number of requested items, i.e., labels, that are currently stored in the cache. A condition of network instability may then be detected based on the hit ratio by determining if the absolute value of the acceleration of the hit ratio exceeds a threshold.