International Business Machines Corporation

Bhuvan Bamba

Madhukar R Korupolu

Kun-Lung Wu

Kaladhar Voruganti

Sougata Mukherjea

Bugra Gedik

Aameek Singh

Prashant Pandey

A method for processing a first data stream specifying locations of a user at different times and a second data stream specifying values of a monitored attribute at a location of interest at different times includes: receiving a location-centric trigger specifying at least one spatial predicate condition relative to the location of interest and at least one non-spatial predicate condition relevant to the location of interest, calculating a safe region that includes locations whose probability of satisfying the spatial predicate condition falls below a first threshold, calculating a safe value container that includes values whose probability of satisfying the non-spatial predicate condition falls below a second threshold, and processing the first data stream and the second data stream against the location-centric trigger, by considering only those locations that are not contained within the safe region and only those values that are not contained within the safe value container.

Selective processing of location-sensitive data streams

A scalable, performance-based, volume allocation technique that can be applied in large storage controller collections is disclosed. A global resource tree of multiple nodes representing interconnected components of a storage system in a plurality of component layers is analyzed to yield gap values for each node (e.g., a bottom-up estimation). The gap value for each node is an estimate of the amount in GB of the new workload that can be allocated in the subtree of that node without exceeding the performance and space bounds at any of the nodes in that subtree. The gap values of the global resource tree are further analyzed to generate an ordered allocation list of the volumes of the storage system (e.g., a top-down selection). The volumes may be applied to a storage workload in the order of the allocation list and the gap values and list are updated.

Scalable performance-based volume allocation in large storage controller collections

A storage placement planning system receives a resource graph describing SAN's resources and virtual machine applications requiring a particular amount of a processing resource element and a storage resource element. The system then determines a coupled placement of the processing element and storage element for each of the applications on a coupled pair of the resource nodes based on a specified throughput and a distance factor between coupled pairs of resource nodes. The coupled placement is determined using an algorithm that implements a cost function that determines affinities between processing nodes and storage nodes for each of said applications of a particular workload. The coupled placement for each of said applications identifies the particular amount of processing resource element placed on a first node for providing a processing resource and the particular amount of storage resource element placed on a second node for providing a storage resource for that application.

Integrated placement planning for heterogenous storage area network data centers

A scalable, performance-based, volume allocation technique that can be applied in large storage controller collections is disclosed. A global resource tree of multiple nodes representing interconnected components of a storage system is analyzed to yield gap values for each node (e.g., a bottom-up estimation). The gap value for each node is an estimate of the amount in GB of the new workload that can be allocated in the subtree of that node without exceeding the performance and space bounds at any of the nodes in that subtree. The gap values of the global resource tree are further analyzed to generate an ordered allocation list of the volumes of the storage system (e.g., a top-down selection). The volumes may be applied to a storage workload in the order of the allocation list and the gap values and list are updated.

The importance of semantic web resources is determined. Some resources are classes (of the type RDFS:Class); others are non-class resources. Non-class resources belong to one or more classes while class resources are subclasses of one or more parent classes. A subjectivity score is determined for each resource of a set of resources based on the number of Resource Description Format (RDF) triples of which the resource is the subject and predefined weights of properties of the triples. An objectivity score is determined for each resource based on the number of RDF triples of which the resource is the object and predefined weights of the properties of the triples. The importance of a class is determined from the respective subjectivity score and objectivity score, and a factor relating to the importance of the class's parents.

Growing community of inventors

Larchmont, NY, United States of America

Bhuvan Bamba