Abstract: Apparatuses and methods for anomaly detection. In one embodiment, a method is implemented in a computing device for building a tree structure to represent a system behavior includes obtaining one or more training log records; and building a tree structure using the one or more training log records. The tree structure includes a plurality of tree nodes. Each successive tree node in a root-to-leaf path of the tree structure representing successive log elements of the one or more training log records. Each of the one or more training log records includes one or more log elements. In one embodiment, a method implemented in a computing device for fault detection includes obtaining a live log record and determining an anomaly in the live log record by comparing corresponding successive elements of the live log record to successive nodes in a root-to-leaf direction of the tree structure.

Abstract: A system and method for creating and updating a distributed indexing table for peer nodes in a peer-to-peer network. In one embodiment, a method includes providing, from a first node to a second node of the peer-to-peer network, a first level of a first plurality of levels for a first designator identifying the first node and entered in a first distributed indexing table stored in the first node. The method also includes receiving, at the first node from the second node, a corresponding first level of a second plurality of levels for a second designator identifying the second node and entered in a second distributed indexing table stored in the second node, and updating the corresponding first level of the second plurality of levels for the second designator within the first distributed indexing table of the first node.

Abstract: A system and method for creating and updating a distributed indexing table for peer nodes in a peer-to-peer network. In one embodiment, a method includes providing, from a first node to a second node of the peer-to-peer network, a first level of a first plurality of levels for a first designator identifying the first node and entered in a first distributed indexing table stored in the first node. The method also includes receiving, at the first node from the second node, a corresponding first level of a second plurality of levels for a second designator identifying the second node and entered in a second distributed indexing table stored in the second node, and updating the corresponding first level of the second plurality of levels for the second designator within the first distributed indexing table of the first node.

Abstract: It is disclosed an agent of a first cloudlet and a method performed therein, the first cloudlet having one or more service properties, the first agent having associated a first group of one or more other cloudlets. Based on an exchange of information about service properties of cloudlets accessible to each agent, a comparison is made between service properties from different cloudlets, and if a certain cloudlet has at least one service property that exceeds a service property of any one of the first group of cloudlets, the certain cloudlet is added to a group of cloudlets, which efficiently represents service properties available to the first agent.

Abstract: A method and job scheduling node (200) for scheduling data processing across a set of processing machines (202), each processing machine comprising at least one data processor. When receiving (2:2) a request for a data processing job, the job scheduling node (200) obtains (2:5) a current processor resource status of each processing machine, the processor resource status indicating at least current usage of each data processor in said processing machine. The job scheduling node (200) further selects (2:6) at least one vacant data processor in the set of processing machines based on the obtained processor resource status, for executing the data processing job, and dispatches (2:7) the data processing job to at least one processing machine (202a) where the at least one selected data processor is located.

Abstract: It is disclosed an agent (30) of a first cloudlet and a method performed therein, the first cloudlet having one or more service properties, the first agent having associated a first group of one or more other cloudlets. Based on exchange (202, 204) of information about service properties of cloudlets accessible to each agent, a comparison (206) is made between service properties from different cloudlets, and if a certain cloudlet has at least one service property that exceeds a service property of any one of the first group of cloudlets, said certain cloudlet is added (212) to the a group of cloudlets, which efficiently represents service properties available to the first agent.

Abstract: A method and job scheduling node (200) for scheduling data processing across a set of processing machines (202), each processing machine comprising at least one data processor. When receiving (2:2) a request for a data processing job, the job scheduling node (200) obtains (2:5) a current processor resource status of each processing machine, the processor resource status indicating at least current usage of each data processor in said processing machine. The job scheduling node (200) further selects (2:6) at least one vacant data processor in the set of processing machines based on the obtained processor resource status, for executing the data processing job, and dispatches (2:7) the data processing job to at least one processing machine (202a) where the at least one selected data processor is located.