Abstract: Technologies for providing efficient data access in an edge infrastructure include a compute device comprising circuitry configured to identify pools of resources that are usable to access data at an edge location. The circuitry is also configured to receive a request to execute a function at an edge location. The request identifies a data access performance target for the function. The circuitry is also configured to map, based on a data access performance of each pool and the data access performance target of the function, the function to a set of the pools to satisfy the data access performance target.

Abstract: Memory controller systems, methods and apparatus for memory access and scheduling are herein disclosed. In some aspects, a memory controller includes a clock, a first interface to be coupled with a first memory device via a common memory channel, and a second interface to be coupled with a second memory device via the common memory channel. The memory controller also includes a register to store data to store data to indicate an access scheme to process access requests to the first memory device according to a first timing scheme and issue access requests to the second memory device according to a second timing scheme. The memory controller further includes logic to cause the access scheme to be implemented in order to issue access requests to the first memory device or to issue access requests to the second memory device via the common memory channel.

Abstract: Technologies for utilizing a split memory pool include a compute sled. The compute sled includes multiple processors communicatively coupled together through a processor communication link. Each processor is to communicate with a different memory sled through a respective memory network dedicated to the corresponding processor and memory sled. The compute sled includes a compute engine to generate a memory access request to access a memory address in far memory. The far memory includes memory located on one of the memory sleds. The compute engine is also to determine, as a function of the memory address and a map of memory address ranges to the memory sleds, the memory sled on which to access the far memory, and send the memory access request to the determined memory sled to access the far memory associated with the memory address.

Abstract: In an example, there is disclosed a host-fabric interface (HFI), including: an interconnect interface to communicatively couple the HFI to an interconnect; a network interface to communicatively couple the HFI to a network; network interface logic to provide communication between the interconnect and the network; a coprocessor configured to provide an offloaded function for the network; a memory; and a caching agent configured to: designate a region of the memory as a shared memory between the HFI and a core communicatively coupled to the HFI via the interconnect; receive a memory operation directed to the shared memory; and issue a memory instruction to the memory according to the memory operation.

Abstract: Embodiments may be generally directed to techniques to cause communication of a registration request between a first end-point and a second end-point of an end-to-end path, the registration request to establish resource load monitoring for one or more resources of the end-to-end path, receive one or more acknowledgements indicating resource loads for each of the one or more resources of the end-to-end path, at least one of the acknowledgements to indicate a resource of the one or more resources is not meeting a threshold requirement for the end-to-end path, and perform an action for communication traffic utilizing the one or more resources based on the acknowledgement.

Abstract: A computing system includes a memory controller having a plurality of bypass parameters set by a software program, a thresholds matrix to store threshold values selectable by the plurality of bypass parameters, and a bypass function to determine whether a first cache line is to be displaced with a second cache line in a first memory or the first cache line remains in the first memory and the second cache line is to be accessed by at least one of a processor core and the cache from a second memory.

Abstract: There is disclosed an example of an artificial intelligence (AI) system, including: a first hardware platform; a fabric interface configured to communicatively couple the first hardware platform to a second hardware platform; a processor hosted on the first hardware platform and programmed to operate on an AI problem; and a first training accelerator, including: an accelerator hardware; a platform inter-chip link (ICL) configured to communicatively couple the first training accelerator to a second training accelerator on the first hardware platform without aid of the processor; a fabric ICL to communicatively couple the first training accelerator to a third training accelerator on a second hardware platform without aid of the processor; and a system decoder configured to operate the fabric ICL and platform ICL to share data of the accelerator hardware between the first training accelerator and second and third training accelerators without aid of the processor.

Abstract: Embodiments may be generally directed to techniques to cause communication of a registration request between a first end-point and a second end-point of an end-to-end path, the registration request to establish resource load monitoring for one or more resources of the end-to-end path, receive one or more acknowledgements indicating resource loads for each of the one or more resources of the end-to-end path, at least one of the acknowledgements to indicate a resource of the one or more resources is not meeting a threshold requirement for the end-to-end path, and perform an action for communication traffic utilizing the one or more resources based on the acknowledgement.

Abstract: Racks and rack pods to support a plurality of sleds are disclosed herein. Switches for use in the rack pods are also disclosed herein. A rack comprises a plurality of sleds and a plurality of electromagnetic waveguides. The plurality of sleds are vertically spaced from one another. The plurality of electromagnetic waveguides communicate data signals between the plurality of sleds.

Abstract: Technology for a data filter device operable to filter training data is described. The data filter device can receive training data from a data provider. The training data can be provided with corresponding metadata that indicates a model stored in a data store that is associated with the training data. The data filter device can identify a filter that is associated with the model stored in the data store. The data filter device can apply the filter to the training data received from the data provider to obtain filtered training data. The data filter device can provide the filtered training data to the model stored in the data store, wherein the filtered training data is used to train the model.

Abstract: There is disclosed an example of an artificial intelligence (AI) system, including: a first hardware platform; a fabric interface configured to communicatively couple the first hardware platform to a second hardware platform; a processor hosted on the first hardware platform and programmed to operate on an AI problem; and a first training accelerator, including: an accelerator hardware; a platform inter-chip link (ICL) configured to communicatively couple the first training accelerator to a second training accelerator on the first hardware platform without aid of the processor; a fabric ICL to communicatively couple the first training accelerator to a third training accelerator on a second hardware platform without aid of the processor; and a system decoder configured to operate the fabric ICL and platform ICL to share data of the accelerator hardware between the first training accelerator and second and third training accelerators without aid of the processor.

Abstract: There is disclosed an example of an artificial intelligence (AI) system, including: a first hardware platform; a fabric interface configured to communicatively couple the first hardware platform to a second hardware platform; a processor hosted on the first hardware platform and programmed to operate on an AI problem; and a first training accelerator, including: an accelerator hardware; a platform inter-chip link (ICL) configured to communicatively couple the first training accelerator to a second training accelerator on the first hardware platform without aid of the processor; a fabric ICL to communicatively couple the first training accelerator to a third training accelerator on a second hardware platform without aid of the processor; and a system decoder configured to operate the fabric ICL and platform ICL to share data of the accelerator hardware between the first training accelerator and second and third training accelerators without aid of the processor.

Abstract: There is disclosed in one example an application-specific integrated circuit (ASIC), including: an artificial intelligence (Al) circuit; and circuitry to: identify a flow, the flow including traffic diverted from a core cloud service of a network to be serviced by an edge node closer to an edge of the network than to the core of the network; receive telemetry related to the flow, the telemetry including fine-grained and flow-level network monitoring data for the flow; operate the Al circuit to predict, from the telemetry, a future service-level demand for the edge node; and cause a service parameter of the edge node to be tuned according to the prediction.

Abstract: An embodiment of an electronic apparatus may comprise one or more substrates, and a controller coupled to the one or more substrates, the controller to allocate a first secure portion of a pooled memory to a first instantiation of an application on a first node, and circuitry coupled to the one or more substrates and the controller, the circuitry to provide a failover interface for a second instantiation of the application on a second node to access the first secure portion of the pooled memory in the event of a failure of the first node. Other embodiments are disclosed and claimed.

Abstract: Aspects of the disclosure are directed to systems, methods, and devices that include an application processor. The application processor includes an interface logic to interface with a communication module using a bidirectional interconnect link compliant with a peripheral component interconnect express (PCIe) protocol. The interface logic to receive a data packet from across the link, the data packet comprises a header and data payload; determine a hint bit set in the header of the data packet; determine a steering tag value in the data packet header based on the hint bit set; and transmit the data payload to non-volatile memory based on the steering tag set in the header.

Abstract: Systems, apparatuses, and methods provide for memory management where an infrastructure processing unit bypasses a central processing unit. Such an infrastructure processing unit determines if incoming packets of memory traffic trigger memory rules stored by the infrastructure processing unit. The incoming packets are routed to the central processing unit in a default mode when the incoming packets do not trigger the memory rules. Conversely, the incoming packets are routed to the infrastructure processing unit and bypass the central processing unit in an inline mode when the incoming packets trigger the memory rules. A memory architecture communicatively coupled to the central processing unit receives a set of atomic transactions from the infrastructure processing unit that bypasses the central processing unit and performs the set of atomic transactions from the infrastructure processing unit.

Abstract: Racks and rack pods to support a plurality of sleds are disclosed herein. Switches for use in the rack pods are also disclosed herein. A rack comprises a plurality of sleds and a plurality of electromagnetic waveguides. The plurality of sleds are vertically spaced from one another. The plurality of electromagnetic waveguides communicate data signals between the plurality of sleds.

Abstract: Data management for edge architected computing systems extends current storage and memory schemes of edge resources to expose interfaces to allow a device, such as an endpoint or client device, or another edge resource, to specify criteria for managing data originating from the device and stored in an edge resource, and extends the storage and memory controllers to manage data in accordance with the criteria, including removing stored data that no longer satisfies the criteria. The criteria includes a temporal hint to specify a time after which the data can be removed, a physical hint to specify a list of edge resources outside of which the data can be removed, an event-based hint to specify an event after which the data can be removed, and a quality of service condition to modify the time specified in the temporal hint based on a condition, such as memory and storage capacity of the edge resource in which the data is managed.

Abstract: An embodiment of a semiconductor apparatus may include technology to receive data with a unique identifier, and bypass encryption logic of a media controller based on the unique identifier. Other embodiments are disclosed and claimed.

Abstract: A host fabric interface (HFI) apparatus, including: an HFI to communicatively couple to a fabric; and a remote hardware acceleration (RHA) engine to: query an orchestrator via the fabric to identify a remote resource having an accelerator; and send a remote accelerator request to the remote resource via the fabric.