VMware

An example computing system includes: a hardware platform having a central processing unit (CPU), a memory, a first network interface controller (NIC) coupled to a first network, and a second NIC coupled to a second network; software executing on the hardware platform, the software including a hypervisor and a virtual machine (VM) managed by the hypervisor, the software including first software executing in the VM and configured to send a packet to a device, the software including a virtual switch executing in the hypervisor; wherein the virtual switch is configured to receive the packet from the first software, insert a sequence number in the packet, duplicate the packet into a first packet and a second packet, send the first packet to the device over the first network through the first NIC, and send the second packet to the device over the second network through the second NIC.

An example computer includes: a hardware platform having a central processing unit (CPU), a memory, and a hardware clock; a hypervisor executing on the CPU and managing the hardware platform; a first virtual machine (VM) supported by the hypervisor and having a passthrough connection to the hardware clock; first guest software executing in the first VM, the first guest software including a first system clock synchronized by communication with the hardware clock via the passthrough connection; a second VM supported by the hypervisor; and second guest software executing in the second VM, the second guest software including a second system clock synchronized with the first system clock via a first portion of the memory shared between the first VM and the second VM, the first portion of the memory coordinated by the hypervisor.

An example apparatus includes a hardware platform having arithmetic circuits and a memory, the memory configured to store, at a first precision, first compressed parameters of a machine learning (ML) model; a network interface controller; and a controller, supported by the hardware platform, configured to: decompress, from the memory through an increase in precision to a second precision, the first compressed parameters to obtain decompressed parameters; control the arithmetic circuits to train, using arithmetic operations, the ML model and update the decompressed parameters; compress, using quantization and reduction in precision to the first precision, the decompressed parameters as updated to obtain second compressed parameters; send, using the network interface controller, the second compressed parameters over a network to a server; and update the first compressed parameters in the memory in response to data received, through the network interface controller, from the server over the network.

A computer system including a first computer and a second computer, wherein the first computer is configured to access data stored in a memory system remotely from the first and second computers, by performing the steps of: determining a first key associated with first data that is stored in the memory system; determining, based on the first key, to transmit to the second computer, a request for a first reference to the first data, and then transmitting the request for the first reference to the second computer; receiving the first reference from the second computer, which the second computer determines by using the first key to locate a first key-reference pair in the memory system and then reading the first reference therefrom; and reading the first data directly from the memory system using the first reference received from the second computer to locate a memory address of the first data.

Example methods and systems for congestion control in a distributed training environment are described. In one example, a computer system may obtain model information associated with a model that is being trained by multiple worker nodes. Based on the model information, the computer system may generate a first payload portion that is non-trimmable, and a second payload portion that is trimmable. The computer system may generate a trimmable payload information that includes the first payload portion and the second payload portion. The trimmable payload information may be forwarded towards a destination. In response to determination that congestion control is required, an intermediate network device may generate and forward trimmed payload information towards the destination. The trimmed payload information may include the first payload portion but excludes at least some of the second payload portion.

A computer is configured to execute network file system (NFS) client software, and issue NFS commands to an NFS server along with checksums that are used by the NFS server to verify the integrity of remote procedure call (RPC) packets that include the NFS commands, by performing the steps of: computing a first checksum of a first RPC packet by applying a checksum algorithm to the first RPC packet, wherein the first RPC packet includes a first NFS packet, the first NFS packet including an NFS command to be performed by the NFS server; inserting the first checksum into the first RPC packet; creating a first transmission control protocol (TCP) packet that includes the first RPC packet having the first checksum inserted therein; and transmitting the first TCP packet to the NFS server over a network using a network interface controller (NIC).

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …