Abstract: Various systems and methods for implementing a service-level agreement (SLA) apparatus receive a request from a requester via a network interface of the gateway, the request comprising an inference model identifier that identifies a handler of the request, and a response time indicator. The response time indicator relates to a time within which the request is to be handled indicates an undefined time within which the request is to be handled. The apparatus determines a network location of a handler that is a platform or an inference model to handle the request consistent with the response time indicator, and routes the request to the handler at the network location.

Abstract: Technologies for providing hardware resources as a service with direct resource addressability are disclosed. According to one embodiment of the present disclosure, a device receives a request to access a destination accelerator device in an edge network, the request specifying a destination address assigned to the destination accelerator device. The device determines, as a function of the destination address, a location of the destination accelerator device and sends the request to the destination accelerator device.

Abstract: Examples are described that relate to waking up or invoking a function such as a processor-executed application or a hardware device. The application or a hardware device can specify which sources can cause wake-ups and which sources are not to cause wake-ups. A device or processor-executed software can monitor reads from or writes to a region of memory and cause the application or a hardware device to wake-up unless the wake-up is specified as inhibited. The updated region of memory can be precisely specified to allow a pinpoint retrieval of updated content instead of scanning a memory range for changes. In some cases, a write to a region of memory can include various parameters that are to be used by the woken-up application or a hardware device. Parameters can include a source of a wake-up, a timer to cap execution time, or any other information.

Abstract: Technologies for providing hardware resources as a service with direct resource addressability are disclosed. According to one embodiment of the present disclosure, a device receives a request to access a destination accelerator device in an edge network, the request specifying a destination address assigned to the destination accelerator device. The device determines, as a function of the destination address, a location of the destination accelerator device and sends the request to the destination accelerator device.

Abstract: Various systems and methods for implementing a service-level agreement (SLA) apparatus receive a request from a requester via a network interface of the gateway, the request comprising an inference model identifier that identifies a handler of the request, and a response time indicator. The response time indicator relates to a time within which the request is to be handled indicates an undefined time within which the request is to be handled. The apparatus determines a network location of a handler that is a platform or an inference model to handle the request consistent with the response time indicator, and routes the request to the handler at the network location.

Abstract: Examples described herein relate to requesting execution of a workload by a next function with data transport overhead tailored based on memory sharing capability with the next function. In some examples, data transport overhead is one or more of: sending a memory address pointer, virtual memory address pointer or sending data to the next function. In some examples, the memory sharing capability with the next function is based on one or more of: whether the next function shares an enclave with a sender function, the next function shares physical memory domain with a sender function, or the next function shares virtual memory domain with a sender function. In some examples, selection of the next function from among multiple instances of the next function based on one or more of: sharing of memory domain, throughput performance, latency, cost, load balancing, or service legal agreement (SLA) requirements.

Abstract: Technologies for collecting metrics associated with a processing resource (e.g., central processing unit (CPU) resources, accelerator device resources, and the like) over a time window are disclosed. According to an embodiment presented herein, a network device receives, in an edge network, a request to provide one or more metrics associated with a processing resource, the request specifying a window indicative of a time period to capture the one or more metrics. The network device obtains the one or more metrics from the processing resource for the specified window and provides the obtained one or more metrics in response to the request.

Abstract: Technologies for providing hardware resources as a service with direct resource addressability are disclosed. According to one embodiment of the present disclosure, a device receives a request to access a destination accelerator device in an edge network, the request specifying a destination address assigned to the destination accelerator device. The device determines, as a function of the destination address, a location of the destination accelerator device and sends the request to the destination accelerator device.

Abstract: Technologies for providing hardware resources as a service with direct resource addressability are disclosed. According to one embodiment of the present disclosure, a device receives a request to access a destination accelerator device in an edge network, the request specifying a destination address assigned to the destination accelerator device. The device determines, as a function of the destination address, a location of the destination accelerator device and sends the request to the destination accelerator device.

Abstract: Examples are described herein that can be used to offload a sequence of work events to one or more accelerators to a work scheduler. An application can issue a universal work descriptor to a work scheduler. The universal work descriptor can specify a policy for scheduling and execution of one or more work events. The universal work descriptor can refer to one or more work events for execution. The work scheduler can, in some cases, perform translation of the universal work descriptor or a work event descriptor for compatibility and execution by an accelerator. The application can receive notice of completion of the sequence of work from the work scheduler or an accelerator.

Abstract: Techniques to facilitate a hardware based table look of a table maintained in or more types of memories or memory domains include examples of receiving a search request forwarded from a queue management device. Examples also include implementing table lookups to obtain a result and sending the result to an output queue of the queue management device for the queue management device to forward the result to a requestor of the search request.

Abstract: Technologies for providing a multi-tenant local breakout switching and dynamic load balancing include a network device to receive network traffic that includes a packet associated with a tenant. Upon a determination that the packet is encrypted, a secret key associated with the tenant is retrieved. The network device decrypts a payload from the packet using the secret key. The payload is indicative of one or more characteristics associated with network traffic. The network device evaluates the characteristics and determines whether the network traffic is associated with a workload requesting compute from a service hosted by a network platform. If so, the network device forwards the network traffic to the service.

Abstract: A central processing unit can offload table lookup or tree traversal to an offload engine. The offload engine can provide hardware accelerated operations such as instruction queueing, bit masking, hashing functions, data comparisons, a results queue, and a progress tracking. The offload engine can be associated with a last level cache. In the case of a hash table lookup, the offload engine can apply a hashing function to a key to generate a signature, apply a comparator to compare signatures against the generated signature, retrieve a key associated with the signature, and apply the comparator to compare the key against the retrieved key. Accordingly, a data pointer associated with the key can be provided in the result queue. Acceleration of operations in tree traversal and tuple search can also occur.

Abstract: An apparatus for training artificial intelligence (AI) models is presented. In embodiments, the apparatus may include an input interface to receive in real time model training data from one or more sources to train one or more artificial neural networks (ANNs) associated with the one or more sources, each of the one or more sources associated with at least one of the ANNs; a load distributor coupled to the input interface to distribute in real time the model training data for the one or more ANNs to one or more AI appliances; and a resource manager coupled to the load distributor to dynamically assign one or more computing resources on ones of the AI appliances to each of the ANNs in view of amounts of the training data received in real time from the one or more sources for their associated ANNs.

Abstract: Various systems and methods for implementing a service-level agreement (SLA) apparatus receive a request from a requester via a network interface of the gateway, the request comprising an inference model identifier that identifies a handler of the request, and a response time indicator. The response time indicator relates to a time within which the request is to be handled indicates an undefined time within which the request is to be handled. The apparatus determines a network location of a handler that is a platform or an inference model to handle the request consistent with the response time indicator, and routes the request to the handler at the network location.

Abstract: A method of operating a first networking device includes receiving, from a media server, a cable length request message. The cable length request message includes data indicative of a length of a first networking cable arranged between the media server and the first networking device. The method includes determining a cumulative cable length based on the length of the first networking cable and a length of a second networking cable arranged between the first networking device and a media client. The method includes, in response to presence of a second networking device between the first networking device and the media client, sending a further cable length request message to the second networking device indicating the cumulative cable length. The method includes, in response to absence of the second networking device, sending a cable length response message to the media server indicating the cumulative cable length.

Abstract: Aspects of the disclosure provide a device transmitting a sequence of packets. The device may include a first queue configured to queue the sequence of packets for transmission by an egress port to a receiving device, and a queue control module configured to transmit the sequence of packets in time slots of the egress port. The sequence of packets can be uniformly distributed across the time slots to satisfy an agreed data rate between the device and the receiving device.

Abstract: A transmit portion of a network device including a medium access control (MAC) module configured to receive a frame of data to be transmitted from the network device in accordance with a MAC security (MACsec) protocol. In response to the frame of data being a precise time protocol (PTP) frame, the MAC module is configured to encrypt the PTP frame in accordance with the MACsec protocol, and associate an identifier with the encrypted PTP frame. A physical layer module includes a transmit module configured to transmit the encrypted PTP frame from the network device at a particular time. A PTP module configured to, based on the identifier associated with the encrypted PTP frame, generate a time stamp indicating the particular time that the transmit module transmits the encrypted PTP frame from the network device. The time stamp is transmitted from the network device along with the encrypted PTP frame.

Abstract: A network device configured to i) control a rate of traffic while operating in a rate-limiting mode, and ii) mirror traffic while operating a sampling mode. The network device includes an ingress circuit configured to receive packets, and a rate limiter in communication with the ingress circuit. The rate limiter is configured to, while the network device is operating in the rate-limiting mode, perform a rate-limiting scheme to control a rate of the packets received through the ingress circuit. The rate limiter is further configured to, while the network device is operating in the sampling mode, perform a modified form of the rate-limiting scheme to mirror the packets received through the ingress circuit.

Abstract: A network apparatus including a central processing unit, first physical layer devices, a second physical layer device, and a network switch. The network switch includes first ports communicating with the first physical layer devices; a second port communicating with the second physical layer device; and third and fourth ports communicating with the central processing unit over first and second media independent interfaces, respectively. The network switch includes a switch core module configured to route data packets between connected ports, including the first and third ports. The network switch includes a bypass switch configured to (i) in response to a first mode being selected, directly connect the second port and the fourth port, bypassing the switch core module, and (ii) in response to a second mode being selected, disconnect the second port from the fourth port and connect the second and fourth ports to the switch core module.