US-12619511-B2 - Method, device, and product for determining fluid dynamics for server

US12619511B2US 12619511 B2US12619511 B2US 12619511B2US-12619511-B2

Abstract

Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for determining computational fluid dynamics (CFD) for a server. The method includes determining, by a trained machine learning model, benchmark CFD for a benchmark server. The method further includes determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server, and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD. Embodiments of the present disclosure can realize real-time thermal simulations and enable real-time thermal models with specific server configurations to be adapted to servers with different components.

Inventors

Pedro Fernandez Orellana
Qiang Chen

Assignees

DELL PRODUCTS L.P.

Dates

Publication Date: 20260505
Application Date: 20240515
Priority Date: 20240423

Claims (20)

1 . A method, comprising: determining, by a trained machine learning model, benchmark computational fluid dynamics (CFD) for a benchmark server; determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server; and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD; wherein determining the difference between the target server configuration and the benchmark server configuration comprises determining (i) geometric similarity between one or more components in the target server configuration and one or more components in the benchmark server configuration and (ii) geometric absence of one or more additional components in the target server configuration from the benchmark server configuration; and wherein the trained machine learning model determines the target CFD for the target server based on at least the determined geometric similarity between the target server configuration and benchmark server configuration and the determined geometric absence of the one or more additional components.
2 . The method according to claim 1 , wherein determining the difference between the target server configuration and the benchmark server configuration comprises: classifying the target server configuration based on the benchmark server configuration, wherein: classifying one or more components in the target server configuration as geometrically similar components in response to the one or more components in the target server configuration and one or more components in the benchmark server configuration having a geometric similarity within a predetermined threshold or having the same number of components; and classifying the one or more components in the target server configuration as geometrically absent components in response to the one or more components in the target server configuration and the one or more components in the benchmark server configuration having a geometric similarity outside the predetermined threshold or having different numbers of components.
3 . The method according to claim 2 , wherein determining, by the trained machine learning model, the target CFD for the target server based on the difference and the benchmark CFD comprises: determining the target CFD for the target server by scaling the benchmark CFD for the benchmark server in response to the one or more components in the target server configuration being the geometrically similar components.
4 . The method according to claim 2 , wherein determining, by the trained machine learning model, the target CFD for the target server based on the difference and the benchmark CFD further comprises: in response to the one or more components in the target server configuration being the geometrically absent components, performing one or more of: retraining the trained machine learning model based on data associated with the geometrically absent components; using physical placeholders to replace the geometrically absent components; and ignoring the geometrically absent components.
5 . The method according to claim 4 , further comprising: setting a table of unique geometrically significant configurations for the target server, wherein components in the table of unique geometrically significant configurations are geometrically dissimilar relative to components in the benchmark server and do not have the physical placeholders.
6 . The method according to claim 4 , wherein training the machine learning model comprises: collecting load data and cooling data associated with operation of the benchmark server; and training the machine learning model based on the load data, the cooling data, and a user predetermined condition.
7 . The method according to claim 6 , wherein the machine learning model is based on a deep learning model, and has a normalization layer and is adjusted based on the target server configuration.
8 . The method according to claim 1 , wherein the benchmark server configuration comprises component information associated with the benchmark server, and user-defined geometry and component preferences.
9 . The method according to claim 1 , wherein determining, by the trained machine learning model, the target CFD for the target server based on the difference and the benchmark CFD further comprises: increasing a degree of generalization of the CFD for the target server by introducing an additional error.
10 . An electronic device, comprising: at least one processor; and memory coupled to the at least one processor and having instructions stored therein which, when executed by the at least one processor, cause the electronic device to perform actions comprising: determining, by a trained machine learning model, benchmark computational fluid dynamics (CFD) for a benchmark server; determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server; and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD; wherein determining the difference between the target server configuration and the benchmark server configuration comprises determining (i) geometric similarity between one or more components in the target server configuration and one or more components in the benchmark server configuration and (ii) geometric absence of one or more additional components in the target server configuration from the benchmark server configuration; and wherein the trained machine learning model determines the target CFD for the target server based on at least the determined geometric similarity between the target server configuration and benchmark server configuration and the determined geometric absence of the one or more additional components.
11 . The electronic device according to claim 10 , wherein determining the difference between the target server configuration and the benchmark server configuration comprises: classifying the target server configuration based on the benchmark server configuration, wherein: classifying one or more components in the target server configuration as geometrically similar components in response to the one or more components in the target server configuration and one or more components in the benchmark server configuration having a geometric similarity within a predetermined threshold or having the same number of components; and classifying the one or more components in the target server configuration as geometrically absent components in response to the one or more components in the target server configuration and the one or more components in the benchmark server configuration having a geometric similarity outside the predetermined threshold or having different numbers of components.
12 . The electronic device according to claim 11 , wherein determining, by the trained machine learning model, the target CFD for the target server based on the difference and the benchmark CFD comprises: determining the target CFD for the target server by scaling the benchmark CFD for the benchmark server in response to the one or more components in the target server configuration being the geometrically similar components.
13 . The electronic device according to claim 11 , wherein determining, by the trained machine learning model, the target CFD for the target server based on the difference and the benchmark CFD further comprises: in response to the one or more components in the target server configuration being the geometrically absent components, performing one or more of: retraining the trained machine learning model based on data associated with the geometrically absent components; using physical placeholders to replace the geometrically absent components; and ignoring the geometrically absent components.
14 . The electronic device according to claim 13 , further comprising: setting a table of unique geometrically significant configurations for the target server, wherein components in the table of unique geometrically significant configurations are geometrically dissimilar relative to components in the benchmark server and do not have the physical placeholders.
15 . The electronic device according to claim 13 , wherein training the machine learning model comprises: collecting load data and cooling data associated with operation of the benchmark server; and training the machine learning model based on the load data, the cooling data, and a user predetermined condition.
16 . The electronic device according to claim 15 , wherein the machine learning model is based on a deep learning model, and has a normalization layer and is adjusted based on the target server configuration.
17 . The electronic device according to claim 10 , wherein the benchmark server configuration comprises component information associated with the benchmark server, and user-defined geometry and component preferences.
18 . The electronic device according to claim 10 , wherein determining, by the trained machine learning model, the target CFD for the target server based on the difference and the benchmark CFD further comprises: increasing a degree of generalization of the CFD for the target server by introducing an additional error.
19 . A computer program product tangibly stored on a non-transitory computer-readable storage medium and comprising machine-executable instructions, wherein the machine-executable instructions, when executed by a machine, cause the machine to perform actions comprising: determining, by a trained machine learning model, benchmark computational fluid dynamics (CFD) for a benchmark server; determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server; and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD; wherein determining the difference between the target server configuration and the benchmark server configuration comprises determining (i) geometric similarity between one or more components in the target server configuration and one or more components in the benchmark server configuration and (ii) geometric absence of one or more additional components in the target server configuration from the benchmark server configuration; and wherein the trained machine learning model determines the target CFD for the target server based on at least the determined geometric similarity between the target server configuration and benchmark server configuration and the determined geometric absence of the one or more additional components.
20 . The computer program product according to claim 19 , wherein determining the difference between the target server configuration and the benchmark server configuration comprises: classifying the target server configuration based on the benchmark server configuration, wherein: classifying one or more components in the target server configuration as geometrically similar components in response to the one or more components in the target server configuration and one or more components in the benchmark server configuration having a geometric similarity within a predetermined threshold or having the same number of components; and classifying the one or more components in the target server configuration as geometrically absent components in response to the one or more components in the target server configuration and the one or more components in the benchmark server configuration having a geometric similarity outside the predetermined threshold or having different numbers of components.

Description

RELATED APPLICATION The present application claims priority to Chinese Patent Application No. 202410494204.9, filed Apr. 23, 2024, and entitled “Method, Device, and Product for Determining Fluid Dynamics for Server,” which is incorporated by reference herein in its entirety. FIELD Embodiments of the present disclosure relate to the field of computers and, more particularly, to a method, an electronic device, and a computer program product for determining fluid dynamics for a server. BACKGROUND As the performance of certain types of electronic devices, especially servers, which may be implemented in the form of high-performance computers and/or other related electronic devices, is continuously improved, the amount of heat they generate also increases, and effective thermal management solutions are thus needed to ensure stable operation and extended service life of such servers. Thermal emulation or thermal simulation is a function widely used in engineering and designing servers. It is critical to know the thermal distribution of a server chassis with different components and configurations to ensure the overall reliability and efficiency of the server. By means of thermal emulation, it is possible to consider heat dissipation schemes at an early design stage, so as to select the appropriate heat dissipation techniques and to design better heat dissipation structures, such as heat sinks, fan layouts, or coolant passages, thus reducing costs and development time. SUMMARY Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for determining computational fluid dynamics (CFD) for a server. According to a first aspect of the present disclosure, a method for determining CFD for a server is provided. The method includes determining, by a trained machine learning model, benchmark CFD for a benchmark server. The method further includes determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server, and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD. According to a second aspect of the present disclosure, an electronic device for determining CFD for a server is provided. The device includes at least one processor, and a memory coupled to the at least one processor and having instructions stored therein, wherein the instructions, when executed by the at least one processor, cause the electronic device to perform actions including: determining, by a trained machine learning model, benchmark CFD for a benchmark server. The actions further include determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server, and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD. According to a third aspect of the present disclosure, a computer program product is provided. The computer program product is tangibly stored on a non-transitory computer-readable medium and includes machine-executable instructions, wherein the machine-executable instructions, when executed by a machine, cause the machine to perform actions including: determining, by a trained machine learning model, benchmark CFD for a benchmark server. The actions further include determining a difference between a target server configuration for a target server and a benchmark server configuration for the benchmark server, and determining, by the trained machine learning model, target CFD for the target server based on the difference and the benchmark CFD. BRIEF DESCRIPTION OF THE DRAWINGS By description of example embodiments of the present disclosure, provided in more detail herein with reference to the accompanying drawings, the above and other objectives, features, and advantages of the present disclosure will become more apparent. In the example embodiments of the present disclosure, the same reference numerals generally represent the same elements, and in which: FIG. 1A is a schematic diagram of an environment in which embodiments of the present disclosure can be implemented; FIG. 1B is a schematic diagram of another environment in which embodiments of the present disclosure can be implemented; FIG. 2 is a flowchart for determining CFD for a server; FIG. 3 is a schematic diagram of a process for training a machine learning model according to an embodiment of the present disclosure; FIG. 4 is a schematic diagram of another process for training a deep learning model in a machine learning model according to an embodiment of the present disclosure; FIG. 5 is an architectural diagram of a deep learning model for real-time thermal simulation according to an embodiment of the present disclosure; FIG. 6 is a schematic diagram of a process for another real-time thermal simulation method according to a