Apparatus and method for distributing work to a plurality of compute express link devices
AssigneeSamsung Electronics Co., Ltd.
InventorSoo-Young Ji
Life of this patent
Estimating the term…
Abstract
An apparatus including: a plurality of compute express link (CXL) devices each including a memory and a processor for processing works stored in the memory; and a switch configured to connect the CXL devices to each other, wherein a first CXL device among the plurality of CXL devices selects at least one second CXL device from at least some CXL devices of the plurality of CXL devices to distribute works stored in a memory of the first CXL device based on a usable capacity of a memory of the at least some CXL devices.
Claims15
An apparatus, comprising:
- a plurality of compute express link (CXL) devices each including a memory and a processor for processing works stored in the memory;
- and
- a switch configured to connect the CXL devices to each other,
- wherein a first CXL device among the plurality of CXL devices selects at least one second CXL device from at least some CXL devices of the plurality of CXL devices to distribute works stored in a memory of the first CXL device based on a usable capacity of a memory of the at least some CXL devices,
- wherein the first CXL device determines an amount of the works to distribute to the at least one second CXL device based on a work processing rate of the at least one second CXL device,
- wherein the first CXL device transmits test work to each of the at least one second CXL device, receives a result of the test work from each of the at least one second CXL device and determines a work processing rate of each of the at least one second CXL device based on a time it takes to transmit the test work and receive the test work.
An apparatus, comprising:
- a plurality of first devices each including a memory and a processor for processing works stored in the memory;
- and
- a second device configured to determine at least one first device among the first devices to which works are to be distributed and a work amount to be distributed based on a work processing rate of each of the first devices and a usable capacity of a memory of each of the first devices, and to distribute works based on a work amount corresponding to each of the first devices to the at least one first device to which the works are to be distributed,
- wherein the second device corrects a work processing rate of each of the first devices based on the usable capacity of the memories of the first devices, and determines the work amount corresponding to each of the first devices based on the corrected work processing rate of the first devices,
- wherein the work amount corresponding to each of the first devices increases as a work processing rate of each of the corrected first devices increases.
A method performed by a computing system including a plurality of compute express link (CXL) devices, the method comprising:
- checking usable capacities of the CXL devices;
- selecting at least one first CXL device to which works are to be distributed among the CXL devices based on the usable capacities of the plurality of CXL devices;
- checking a work processing rate of the at least one first CXL device;
- and
- determining a distribution amount of works to be allocated to each of the at least one first CXL device based on the work processing rate of the at least one first CXL device,
- wherein the checking of the work processing rate of the at least one first CXL device includes:
- transmitting test work to each of the at least one first CXL device;
- receiving a result of the test work from each of the at least one first CXL device;
- and
- calculating a work processing rate of the at least one first CXL device based on a time from when the test work was transmitted to each of the at least one first CXL device until receiving a result of the test work from each of the at least one first CXL device.
Dependencies are inferred from the claim text.