University of Illinois

Skhynix Inc.

Jian Huang

Hyung Jin Lim

Alexander Gerhard Schwing

Myeong Joon Kang

Deming Chen

Jong Ryool Kim

Luyi Kang

Xiaohao Wang

Youjie Li

Chang Hwan Youn

An example embodiment may involve a storage device comprising one or more processors, memory, a host interface connectable to a host device, a network interface connectable to a backup device, and program instructions, stored in the memory, that upon execution by the one or more processors cause the storage device to perform operations comprising: causing data received, by way of the host interface, to be stored in a page of the memory; after storing the data, marking the page as invalid due to input/output operations of the memory; determining a predicted idle time period within which the memory is not expected to be engaged in further input/output operations; determining that a duration of the predicted idle time period exceeds a threshold length; and during the predicted idle time period, transmitting, by way of the network interface, the data to the backup device.

Ransomware-aware solid-state drive

A data storage device includes a nonvolatile memory device, a volatile memory device, a data encryption circuit configured to encrypt data outputted from the nonvolatile memory device, a data decryption circuit configured to decrypt encrypted data output from the data encryption circuit and configured to provide the decrypted data to the volatile memory device, and a processor configured to perform a first process that controls installation of a first in-storage program in the data storage device, a second process configured to manage a mapping table storing a relation between a logical address and a physical address of the nonvolatile memory device, and a third process configured to execute the first in-storage program.

Data storage device performing in-storage processing

A programmable switch includes an input arbiter to analyze packet headers of incoming packets and determine which of the incoming packets are part of gradient vectors received from worker computing devices that are performing reinforcement learning. The programmable switch also includes an accelerator coupled to the input arbiter, the accelerator to: receive the incoming packets from the input arbiter; asynchronously aggregate gradient values of the incoming packets, as the gradient values are received, to generate an aggregated data packet associated with a gradient segment of the gradient vectors; and transfer the aggregated data packet to the input arbiter to be transmitted to the worker computing devices, which are to update local weights based on the aggregated data packet.

Accelerating distributed reinforcement learning with in-switch computing

Growing community of inventors

Champaign, IL, United States of America

Jian Huang