Abstract: Disclosed are various embodiments provisioning a data processing unit in a host machine. There can be multiple data processing units within the host machine with varying hardware or software requirements for an installation image that can be utilized to provision the device. Multiple installation images can be generated for different data processing units having varying requirements in a heterogeneous environment.

Abstract: Disclosed are various embodiments for a unified boot image that can be used to install an operating system onto a host machine and a respective operating system onto a data processing units (DPU) installed on a host machine. The unified boot image contains installation files for installing an operating system on the host machine and an installation depot that can be used to create a boot image for installing the same or different operating system on the DPU. During installation of an operating system on a host machine, the installation workflow can also require installation of an additional operating system or other configuration of a DPU installed in a host machine. In response to determining that an operating system is to be installed on the DPU, the installation depot can be obtained and reformatted into a downloadable format that is compatible with the DPU.

Abstract: Disclosed are various embodiments for coordinating the rollback of installed operating systems to an earlier, consistent state. In response to determining that a data processing unit (DPU) installed on a computing device has failed to successfully boot a first time, the computing device can be power cycled for a first time. In response to determining that the DPU has successfully booted a second time, a first version of a host operating system can be booted. A DPU operating system (DPU OS) is then booted from a DPU alternate boot image. In response to determining that the first version of the host operating system fails to match an executing version of the DPU OS, the computing device can be power cycled a second time and the host operating system is then booted from a host alternate boot image.

Abstract: Disclosed are various embodiments for coordinating the rollback of installed operating systems to an earlier, consistent state. In response to determining that a data processing unit (DPU) installed on a computing device has failed to successfully boot a first time, the computing device can be power cycled for a first time. In response to determining that the DPU has successfully booted a second time, a first version of a host operating system can be booted. A DPU operating system (DPU OS) is then booted from a DPU alternate boot image. In response to determining that the first version of the host operating system fails to match an executing version of the DPU OS, the computing device can be power cycled a second time and the host operating system is then booted from a host alternate boot image.

Abstract: Disclosed are various examples of provisioning a data processing unit (DPU) management operating system (OS). A management hypervisor installer executed on a host device launches or causes a server component to provide a management operating system (OS)installer image at a particular URI accessible over a network internal to the host device. A baseboard management controller (BMC) transfers the DPU management OS installer image to the DPU device. A volatile memory based virtual disk is created using the DPU management OS installer image. The DPU device is booted to a DPU management OS installer on the volatile memory based virtual disk. The DPU management OS installer installs a DPU management operating system to a nonvolatile memory of the DPU device on reboot of the DPU device.

Abstract: Disclosed are various examples of provisioning a data processing unit (DPU) management operating system using a capsule. A management hypervisor installer executed on a host device receives a listing DPU device from a baseboard management controller (BMC). A preinstalled DPU management operating system image is identified for a DPU device from the listing, and is wrapped with a capsule that specifies the capsule as a DPU management operating system image capsule. A server component provides the DPU management operating system image capsule at a particular URI, and the URI is transmitted to the BMC.

Abstract: Distributed query processing is often performed by a set of nodes that apply MapReduce to a data set and materialize partial results to storage, which are then aggregated to produce the query result. However, this architecture requires a preconfigured set of database nodes; can only fulfill queries that utilize MapReduce processing; and may be slowed down by materializing partial results to storage. Instead, distributed query processing can be achieved by choosing a node for various portions of the query, and generating customized code for the node that only performs the query portion that is allocated to the node. The node executes the code to perform the query portion, and rather than materializing partial results to storage, streams intermediate query results to a next selected node in the distributed query. Nodes selection may be involve matching the details of the query portion with the characteristics and capabilities of the available nodes.

Abstract: Systems, methods, and computer-readable medium, are disclosed for transforming user-defined-function invocations in a query-based environment. A user-defined-function (UDF) and a query invoking the UDF are received. The UDF is parsed into a plurality of statements. A first expression tree corresponding to the UDF and a second expression tree corresponding the query are constructed, and merged using an operator to generate a transformed expression. The transformed expression is simplified, using transformation rules, if it is determined that is can be simplified.

Abstract: Distributed query processing is often performed by a set of nodes that apply MapReduce to a data set and materialize partial results to storage, which are then aggregated to produce the query result. However, this architecture requires a preconfigured set of database nodes; can only fulfill queries that utilize MapReduce processing; and may be slowed down by materializing partial results to storage. Instead, distributed query processing can be achieved by choosing a node for various portions of the query, and generating customized code for the node that only performs the query portion that is allocated to the node. The node executes the code to perform the query portion, and rather than materializing partial results to storage, streams intermediate query results to a next selected node in the distributed query. Nodes selection may be involve matching the details of the query portion with the characteristics and capabilities of the available nodes.

Abstract: The present disclosure proposes the method for optimizing the performance of data base/web-service backed applications by automatically prefetching query results. The proposed system and methods automatically insert prefetch instructions at the earliest possible points across procedure calls in application source code, in presence of conditional branching and loops. A data flow analysis technique called anticipable expressions analysis is extended, to analyze anticipability of queries. The benefit of prefetching is limited due to the presence of assignment statements and conditional branches that precede the query execution statement. Enhancements such as code motion, chaining and rewriting prefetch requests are devised to increase benefits of prefetching. These techniques perform equivalence preserving program and query transformations.

Abstract: A custom query execution engine can be generated that captures a query. More particularly, the custom query execution engine can be generated based on combination of a query and an execution engine. Subsequent to generation, a custom query execution engine can be submitted to a system configured to execute the custom query execution engine and evaluate the query over a data store.

Abstract: Systems, methods, and computer-readable medium, are disclosed for transforming user-defined-function invocations in a query-based environment. A user-defined-function (UDF) and a query invoking the UDF are received. The UDF is parsed into a plurality of statements. A first expression tree corresponding to the UDF and a second expression tree corresponding the query are constructed, and merged using an operator to generate a transformed expression. The transformed expression is simplified, using transformation rules, if it is determined that is can be simplified.

Abstract: The present disclosure proposes the method for optimizing the performance of data base/web-service backed applications by automatically prefetching query results. The proposed system and methods automatically insert prefetch instructions at the earliest possible points across procedure calls in application source code, in presence of conditional branching and loops. A data flow analysis technique called anticipable expressions analysis is extended, to analyze anticipability of queries. The benefit of prefetching is limited due to the presence of assignment statements and conditional branches that precede the query execution statement. Enhancements such as code motion, chaining and rewriting prefetch requests are devised to increase benefits of prefetching. These techniques perform equivalence preserving program and query transformations.