Abstract: Techniques are disclosed for providing high performant packets processing capabilities in a virtualized cloud environment that enhance the scalability and high availability of the packets processing infrastructure. In certain embodiments disclosed herein, the VNICs functionality performed by network virtualization devices (NVDs) is offloaded from the NVDs to a fleet of computers, referred to as VNIC-as-a-Service System (or VNICaaS system). VNICaaS system is configured to provide Virtual Network Interface Cards (VNICs)-related functionality or service for multiple compute instances belonging to multiple tenants or customers of the CSPI. The VNICaaS system is capable of hosting multiple VNICs to process and transmit traffic in a distributed virtualized cloud networks environment. A single VNIC executed by the VNICaaS system can be used to process packets received from multiple compute instances.

Abstract: A computer-implemented method to determine form factors of a tree includes building an input tree, wherein the input tree includes nodes. The method further includes implementing a first top-down pass to determine a universal number for each node in the input tree. The method further includes implementing a second top-down pass to determine form factors for each node of the input tree, wherein a form factor includes a depth and a width of the tree with a root in a corresponding node. The method further includes storing the form factors as part of node metadata or in a separate table.

Abstract: Some embodiments are directed to an improved approach to implement deployments where a client can get application-level redirects to different servers, where the service is running in a different cloud environment. Dynamic port mapping may be performed at runtime. Routes may be added to IP tables to implement redirects from a first cloud to a second cloud.

Abstract: Techniques are described herein for efficiently managing and transmitting changes to a standard object within a data processing system while preserving the integrity of customer-defined custom views on that standard object. The proposed systems and methods introduce a novel approach to handling Create, Update, and Destroy (hereinafter “CUD”) operations on a standard object, allowing customers to define custom views that encapsulate specific fields of interest. The systems and methods involve the creation of a replication object, triggered by changes to the standard object, which captures these alterations without storing actual data. This mechanism, coupled with the generation of change logs and transmission to a message bus, enables changes to the standard object to be performed without impacting the customer-defined views.

Abstract: A data management system receives updates to records of a source dimension. Some records of the source dimension reference target dimensions. The data management system identifies template records from existing records in the source dimension for modeling changes to connections with the target dimensions based on the updated records in the source dimension. The template records are discovered using rules-driven processes, AI-driven processes, or a serial or parallel hybrid processes including rules and AI. These processes use ancestor information from the updated records to find best-matching template records. The rules-driven processes additionally rely on matching fields, and the AI-driven processes additionally rely on vector embeddings and optionally clustering. Updates are made to the target records in the target dimensions, including any roll-up structures indicated for data propagation, identified using the template records, and downstream applications using the target records may consume the updates.

Abstract: Techniques for navigating interactive diagrams are provided. In one technique, a first content of an interactive diagram is presented in a content display area of a user interface. The interactive diagram also includes second content that is outside the content display area of the user interface. A boundary is identified that encapsulates the first content and the second content. While displaying a scrolling control on the user interface, first user input is received that causes a portion of the second content to be presented in the content display area and a portion of the first content to no longer be presented in the content display area. In response receiving second user input to view a portion of the interactive diagram that is outside the boundary, a visual appearance of one or more elements of the scrolling control is updated.

Abstract: Techniques discussed herein include dynamically providing synchronous and/or asynchronous data processing by a machine-learning model service. The machine-learning model service (“the service”) executes a stream manager application, a web interface, and a machine-learning model via a common container. The stream manager application can obtain input data (e.g., from an input data stream, a partition of an input data stream, etc.) and provide the data to the machine-learning model through the web interface using a local communication channel (e.g., a loopback interface that bypasses local network interface hardware of the computing device on which the model executes). Prediction results from the model may be provided as output data (e.g., to an output data stream, to a partition of an output data stream, etc.).

Abstract: Herein is high availability for online transaction processing with redundancy and redo for a federation of pluggable databases and container databases. In an embodiment of a federation of container database management systems that includes a first container database, first redo data of a first pluggable database in a second container database is obtained based on a database dictionary in the first container database. To the first pluggable database in the first container database, the first redo data of the first pluggable database in the second container database is applied. Based on the database dictionary in the first container database, second redo data of a second pluggable database in a third container database is obtained. To the second pluggable database in the first container database, without modifying content of the first pluggable database in the first container database, the second redo data of the second pluggable database in the third container database is applied.

Abstract: Techniques are disclosed for a smart network interface card (smartNIC) performing a unified logging process. In one example, an accelerator transmits a packet to a programming data plane of the smart network interface card. The programming data plane determines whether the packet is to be forwarded. In accordance with a determination that the packet is not be forwarded, the programming data plane modifies the packet to include an instruction that instructs the accelerator to log the packet and to not forward the packet. The programming data plane transmits the modified packet to the accelerator. The accelerator logs data associated with the modified packet to a unified log based at least in part on the instruction.

Abstract: Techniques for generating terms to replace an initial set of search terms for a query are disclosed. A system generates a training data set for training a machine learning model. Generating the training data set includes generating search value vectors for each of a set of labels based on sets of search values associated respectively with the labels in the set of labels. The system trains a machine learning model to predict a target label for a target search vector based on the set of labels and the respectively associated search value vectors. The system generates a target search value vector based on an initial set of search values. The system then applies the trained machine learning model to the target search value vector to predict the target label. The target label is used as a search term, that replaces the initial set of search values, for executing the query.

Abstract: A regional NRF may register NFs, and the registration information may include locality information for each registered NF. During synchronization with a root NRF, the regional NRF provides the locality information for each registered NF. During discovery request routing by the root NRF, the root NRF prioritizes regional NRFs having NFs that satisfy the discovery parameters of the discovery request and match the preferred locality indicated in the discovery request.

Abstract: Techniques for concurrent lazy reference tracking in an old garbage collection generation are disclosed, including: encountering, by a mutator thread during a garbage collection epoch, a first instruction to write a first value to a field; responsive to encountering the first instruction to write the first value to the field: entering a slow-path write barrier; performing, by the slow-path write barrier, a first one or more reference counting operations with respect to the field; encountering, by the mutator thread during the first garbage collection epoch and subsequent to encountering the first instruction to write the first value to the field, a second instruction to write a second value to the field; responsive to encountering the second instruction to write the second value to the field: entering a fast-path write barrier; wherein the fast-path write barrier does not perform any reference counting operations with respect to the field.

Abstract: A distributed computing system is described that leverages a nearline storage layer to minimize the downtime required for bootstrapping a new computing cluster in the distributed computing system. The system executes a computing cluster comprising a set of computing nodes and determines a set of one or more data segments to be written to a nearline storage system. The system writes the data segments to the nearline storage system. In certain examples, the system receives a request to create a second computing cluster and responsive to the request, bootstraps the second computing cluster using the set of data segments stored on the nearline storage system. The system additionally leverages the nearline storage layer to accelerate query processing by the computing nodes of a computing cluster.

Abstract: Techniques are disclosed herein for focused training of language models and end-to-end hypertuning of the framework. In one aspect, a method is provided that includes obtaining a machine learning model pre-trained for language modeling, and post-training the machine learning model for various tasks to generate a focused machine learning model. The post-training includes: (i) training the machine learning model on an unlabeled set of training data pertaining to a task that the machine learning model was pre-trained for as part of the language modeling, and the unlabeled set of training data is obtained with respect to a target domain, a target task, or a target language, and (ii) training the machine learning model on a labeled set of training data that pertains to another task that is an auxiliary task related to a downstream task to be performed using the machine learning model or output from the machine learning model.

Abstract: Techniques for processing a large dataset to extract information from the dataset, where the processing is performed by multiple systems and where the entire dataset, due to its size, cannot be loaded into the memory of any one of the multiple systems. In certain implementations, the information extracted from the dataset is in the form of a set of one or more statistical metrics computed for the dataset. For example, the dataset may include datapoints related to a machine-learning (ML) model, and a metric value can be computed for the ML model based upon the dataset datapoints. The metric value may, for example, be a metric that measures the performance of the ML model.

Abstract: Techniques are disclosed for efficient generation and maintenance of interface components in an infrastructure-as-a-service (IaaS) environment. A component facilitation system may receive a first request to display a component dashboard and generate declarative metadata for component generation, at least a portion of the declarative metadata corresponding to one or more visual components to be displayed on an interface. The component facilitation system may transmit the declarative metadata used to determine the one or more visual components. The component facilitation system may receive a second request for service data to be used to complete the one or more visual components and identify one or more services configured to execute at a service provider system and extract metric data. The component facilitation system may transmit the service data to be used to generate one or more sets of rendering data.

Abstract: A smart network interface card (smartNIC) may receive first traffic for a first process configured with a first bandwidth limit. The smartNIC may receive second traffic for a second process configured with a second bandwidth limit, the second bandwidth limit corresponding to a larger value between a second transmit limit and a second receive limit associated with the second process. The smartNIC may queue the received traffic associated with the first process and the second process in a scheduler, the scheduler having a first set of queues configured to store traffic from the first process, and a second set of queues configured to store traffic from the second process. The smartNIC may forward queued traffic from the first set of queues or the second set of queues, a maximum amount of forwarded first process traffic corresponding to the first bandwidth limit minus an amount of forwarded second process traffic.

Abstract: Systems that determine relationships between network components of a cluster using packet filters are disclosed. A system can identify objects that implement services of a cluster and network connections associated with respective pairs of the objects. The system can also filter out network connections from the identified network connections. The filtering can remove connections between source objects and destination objects based on the destination objects lacking any components that implement a service in cluster. The filtering can also retain network connections between source objects and destination objects based on the source objects including components that implement at least one service, and based on the second destination object including components that implement at least one service. Additionally, the system can generate relationship maps and network topologies using the determined relationships.

Abstract: Techniques for sales orchestration using iterative machine learning are disclosed, including: obtaining, by a customer relationship management (CRM) system, opportunity-specific CRM data; generating, by the CRM system, one or more suggested sales actions applicable to the opportunity-specific CRM data, at least by applying a machine learning model to the opportunity-specific CRM data, wherein the machine learning model is configured to generate suggested sales actions based on CRM data; presenting, in a graphical user interface (GUI) of the CRM system, the one or more suggested sales actions; receiving, via the GUI of the CRM system, user input that indicates a chosen sales action associated with the opportunity-specific CRM data; obtaining, by the CRM system, data that indicates an outcome of the chosen sales action; updating the machine learning model based on the outcome of the chosen sales action.

Abstract: The present disclosure relates to adaptively overlapping redo writes. A log writer, while operating in a thin mode, may assign a first log writer group of a plurality of log writer groups to write one or more first redo log records to an online redo log in response to determining that a pipelining parameter is satisfied. The thin mode may be associated with one or more target sizes that are less than one or more target sizes associated with a thick mode. The log writer may determine to operate the thick mode based at least in part on at least a portion of the plurality of log writer groups being unavailable to write one or more second redo log records to the online redo log.