Abstract: Placing virtual network functions (VNFs) in nodes of an infrastructure are disclosed. A request for services may identify a series of VNFs. Requests may be grouped by time windows such that all requests in a time window are placed together. Placing the requests includes identifying which nodes host which VNFs. Some of the VNFs may already exist in the infrastructure. Placing the requests may include generating a placement plan that accounts, for example, for available resources, links in the infrastructure, node restrictions, service level agreements, quality of service requirements, and user priorities. Once generated, the plan is executed. Chains of VNFs can be replaced when service degrades or based on user location.

Abstract: Reconstructing compressed data to reduce compression loss. Data at a device is compressed using lossy compression and metadata values are added to the compressed data. A gateway receives a package including the compressed data and the metadata values. The data is decompressed and input to a machine learning model along with the metadata values. The machine learning model is trained to reduce the compression loss. The output of the model is an improved decompressed data. Actions may be performed based on the improved decompressed data.

Abstract: Improving the quality of lossy compressed data is disclosed. Lossy compressed data may be generated at a device and transmitted to a gateway. The gateway includes a reconstruction engine that is trained to reduce the compression loss by generating a learned residue. The residue is added to the lossy decompressed data to generated corrected decompressed data that has a smaller loss than the lossy decompressed data.

Abstract: A method for scaling a virtual network function (VNF) by analyzing VNF metrics data, the method that includes obtaining, by a computing device, a VNF metrics data, identifying a queue in the VNF metrics data, calculating a constraint violation probability based on the queue, making a first determination that the constraint violation probability is greater than an upper threshold, and based on the first determination, scaling up the VNF.

Abstract: A method for scaling a virtual network function (VNF) by analyzing VNF metrics data, the method that includes obtaining, by a computing device, a VNF metrics data, identifying a queue in the VNF metrics data, calculating a constraint violation probability based on the queue, making a first determination that the constraint violation probability is greater than an upper threshold, and based on the first determination, scaling up the VNF.

Abstract: Techniques are provided for logical network resource allocation and creation. One method comprises obtaining a representation of virtualized resources to allocate for a logical network within a physical network, wherein the physical network comprises physical nodes and physical links, wherein the virtualized resources comprise virtual nodes and virtual links, and wherein the physical nodes host one or more virtual nodes and one or more of the physical links host a given virtual link; obtaining a first ordered list of the plurality of virtual nodes; obtaining a second ordered list of the plurality of physical nodes; for a virtual node in the first ordered list: selecting and allocating a physical node from the second ordered list as a host for the virtual node; and allocating a physical path for the virtual links associated with the virtual node. The logical network can be created using the allocated physical nodes and physical links.

Abstract: Placing virtual network functions (VNFs) in nodes of an infrastructure are disclosed. A request for services may identify a series of VNFs. Requests may be grouped by time windows such that all requests in a time window are placed together. Placing the requests includes identifying which nodes host which VNFs. Some of the VNFs may already exist in the infrastructure. Placing the requests may include generating a placement plan that accounts, for example, for available resources, links in the infrastructure, node restrictions, service level agreements, quality of service requirements, and user priorities. Once generated, the plan is executed. Chains of VNFs can be replaced when service degrades or based on user location.

Abstract: A method for managing a service function chain (SFC) includes obtaining, by a SFC orchestrator of a first domain, a SFC creation request for a SFC, in response to the SFC creation request: generating a domain participation list of a set of potential domains that may service the SFC creation request, sending a bid request to each of the set of potential domains, receiving bid responses from the set of potential domains, ranking the set of potential domains, sending a SFC placement request to a second domain of the set of potential domains based on the ranking, and obtaining a SFC placement response from the second domain, wherein the SFC placement response specifies success of SFC placement in the second domain.

Abstract: One example method includes creating an ILP model that includes a delay model, an energy model, and a QoS model, and modeling, using the integer linear programming model, a VNF placement problem as an ILP problem, and the modeling includes: using the delay model to identify propagation, transmission, processing, and queuing, delays implied by enabling an instance of the VNF at an edge node to accept a user VNF call; using the energy model to identify energy consumption implied by enabling an instance of the VNF at an edge node to accept a user VNF call; and using the QoS model to identify end-to-end delay, bandwidth consumption, and jitter, implied by enabling an instance of the VNF at an edge node to accept a user virtual network function call. The problem modeled by the ILP model may be resolved by a heuristic method.

Abstract: Techniques are provided for data discovery and data integration in a data lake. One method comprises obtaining data files from a data lake, wherein each data file comprises multiple records having multiple fields; selecting multiple candidate fields from a data file based on a record type; determining a relevance score for each candidate field from the data file based on multiple features extracted from the data file; and clustering the scored candidate fields into clusters of similar domains using a hashing algorithm, wherein a given cluster comprises candidate fields, wherein multiple data files can be integrated based on a domain of the candidate fields in the given cluster. The relevance score for each candidate field is based on multiple features comprising, for example, features that take into account a morphological or semantic similarity between file name, file metadata and/or file records and features that consider statistics of candidate fields in a data file.

Abstract: Techniques are provided for data discovery and data integration in a data lake. One method comprises obtaining data files from a data lake, wherein each data file comprises multiple records having multiple fields; selecting multiple candidate fields from a data file based on a record type; determining a relevance score for each candidate field from the data file based on multiple features extracted from the data file; and clustering the scored candidate fields into clusters of similar domains using a hashing algorithm, wherein a given cluster comprises candidate fields, wherein multiple data files can be integrated based on a domain of the candidate fields in the given cluster. The relevance score for each candidate field is based on multiple features comprising, for example, features that take into account a morphological or semantic similarity between file name, file metadata and/or file records and features that consider statistics of candidate fields in a data file.

Abstract: Techniques are provided for dataflow execution time estimation for distributed processing frameworks. An exemplary method comprises: obtaining an input dataset for a dataflow for execution; determining a substantially minimal data unit for a given operation of the dataflow processed by the given operation; estimating a number of rounds required to execute a number of data units in the input dataset using nodes assigned to execute the given operation; determining an execution time spent by the given operation to process one data unit; estimating the execution time for the given operation based on the execution time spent by the given operation to process one data unit and the number of rounds required to execute the number of data units in the input dataset; and executing the given operation with the input dataset. A persistent cost model is optionally employed to record the execution times of known dataflow operations.

Abstract: Techniques are provided for dataflow execution time estimation for distributed processing frameworks. An exemplary method comprises: obtaining an input dataset for a dataflow for execution; determining a substantially minimal data unit for a given operation of the dataflow processed by the given operation; estimating a number of rounds required to execute a number of data units in the input dataset using nodes assigned to execute the given operation; determining an execution time spent by the given operation to process one data unit; estimating the execution time for the given operation based on the execution time spent by the given operation to process one data unit and the number of rounds required to execute the number of data units in the input dataset; and executing the given operation with the input dataset. A persistent cost model is optionally employed to record the execution times of known dataflow operations.

Abstract: Pattern queries are evaluated in parallel over large N-dimensional datasets to identify features of interest.

Abstract: Techniques are provided for automatic placement of cache operations in a dataflow. An exemplary method obtains a graph representation of a dataflow of operations; determines a number of executions and a computational cost of the operations, and a computational cost of a caching operation to cache a dataset generated by an operation; establishes a dataflow state structure recording values for properties of the dataflow operations for a number of variations of caching various dataflow operations; determines a cache gain factor for dataflow operations as an estimated reduction in the accumulated cost of the dataflow by caching an output dataset of a given operation; determines changes in the dataflow state structure by caching an output dataset of a different operation in the dataflow; and searches the dataflow state structures to determine the output datasets to cache based on a total dataflow execution cost.

Abstract: An architecture, methods and apparatus are provided for managing sensor data. Sensor networks comprised of a plurality of sensors are managed by obtaining measurement data and context data from the plurality of sensors; storing the obtained measurement data and context data using a Massively Parallel Processing Database Management System (MPP DBMS); and managing the sensor network from outside of the sensor network using the MPP DBMS. Context-aware adaptation of sensors is based on context regarding a state of the sensor network and context regarding a state of one or more applications. The sensor nodes are optionally clustered based on semantic similarities among sensor readings from different sensor nodes and a distance among the sensor nodes. A subset of the sensor nodes is optionally selected to be active based on a residual energy of the sensor nodes and a relevance of the sensor nodes to an application. Data prediction models are generated and employed for data sensing and analytics.

Abstract: An architecture, methods and apparatus are provided for managing sensor data. Sensor networks comprised of a plurality of sensors are managed by obtaining measurement data and context data from the plurality of sensors; storing the obtained measurement data and context data using a Massively Parallel Processing Database Management System (MPP DBMS); and managing the sensor network from outside of the sensor network using the MPP DBMS. Context-aware adaptation of sensors is based on context regarding a state of the sensor network and context regarding a state of one or more applications. The sensor nodes are optionally clustered based on semantic similarities among sensor readings from different sensor nodes and a distance among the sensor nodes. A subset of the sensor nodes is optionally selected to be active based on a residual energy of the sensor nodes and a relevance of the sensor nodes to an application. Data prediction models are generated and employed for data sensing and analytics.