Computing system and method of operation thereof

Abstract

This invention proposes a computing system and its operating method. The computing system includes a master computer and slave computer devices. The slave computer devices are connected to a dedicated Internet Protocol (IP) address of the master computer. The master computer and slave computer devices each establish user accounts for passwordless connection authentication. The slave computer device establishes key data and provides the public key of the key data to the master computer device, so that the master computer device can verify whether the slave computer device has logged in without a password based on the slave computer device's public key. After the slave computer device logs in without a password, the master computer and slave computer devices perform distributed computing to train a large-scale model.

Description

Technical Field

[0001] This invention relates to a system, and more particularly to a computing system and its operating method. Background Technology

[0002] Distributed computing across multiple computers typically requires users to manually configure connections between these devices and manually log in to initiate data transfer, which is time-consuming and costly. Furthermore, conventional connection configurations for multiple computers often lack security, hindering distributed computing. Summary of the Invention

[0003] This invention provides a computing system and its operating method, which can automatically establish connections.

[0004] The computing system of this invention includes a master computer device and a slave computer device. The slave computer device is connected to a dedicated Internet Protocol (IP) address of the master computer device. The master computer device and the slave computer device each establish user accounts for passwordless connection authentication. The slave computer device establishes key data and provides the public key of the key data to the master computer device, so that the master computer device can verify whether the slave computer device has logged in to the master computer device without a password based on the slave computer device's public key. After the slave computer device logs in to the master computer device without a password, the master computer device and the slave computer device perform distributed computing to train a large-scale model.

[0005] The operating method of this invention is applicable to a computing system. The computing system includes a master computer device and slave computer devices. The operating method includes the following steps: connecting the slave computer device to a dedicated Internet Protocol address of the master computer device; establishing user accounts for passwordless connection authentication on both the master and slave computer devices; establishing key data on the slave computer device and providing the public key of the key data to the master computer device; verifying whether the slave computer device has logged into the master computer device without a password using the public key of the slave computer device; and after the slave computer device has logged into the master computer device without a password, performing distributed computing on the master and slave computer devices to train a large-scale model.

[0006] Based on the above, the computing system and its operating method of the present invention enable a computer device to connect to a host computer device without a password in order to perform distributed computing.

[0007] To make the above features and advantages of the present invention more apparent and understandable, specific embodiments are described below in conjunction with the accompanying drawings. Attached Figure Description

[0008] Figure 1This is a schematic diagram of a computing system according to an embodiment of the present invention.

[0009] Figure 2 This is a flowchart of an operation method according to an embodiment of the present invention.

[0010] Figure 3 This is a flowchart illustrating identity verification according to an embodiment of the present invention.

[0011] Figure 4 This is a flowchart of a data transmission operation according to an embodiment of the present invention.

[0012] The reference numerals in the attached figures are explained as follows:

[0013] 100: Computing System

[0014] 110: Main computer device

[0015] 120_1~120_N: From computer devices

[0016] S210~S250, S301~S305, S410~S438: Steps Detailed Implementation

[0017] To make the contents of this invention more readily apparent, the following specific embodiments are provided as examples on which this invention can indeed be implemented. Furthermore, wherever possible, elements / components / steps using the same reference numerals in the drawings and embodiments represent the same or similar parts.

[0018] Figure 1 This is a schematic diagram of a computing system according to an embodiment of the present invention. (See reference) Figure 1 The computing system 100 includes a master computer device 110 and multiple slave computer devices 120_1 to 120_N, where N is a positive integer. In one embodiment, the number of slave computer devices 120_1 to 120_N may also be one. In this embodiment, each of the slave computer devices 120_1 to 120_N can be connected to the master computer device 110 via a Thunderbolt cable. Each of the slave computer devices 120_1 to 120_N can log in to the master computer device 110 without a password.

[0019] In this embodiment, the main computer device 110 and the slave computer devices 120_1 to 120_N may each include a processor and a storage device. The processor may include, for example, a Central Processing Unit (CPU) or other programmable general-purpose or special-purpose microprocessors, digital signal processors (DSPs), programmable controllers, application-specific integrated circuits (ASICs), programmable logic devices (PLDs), other similar processing devices, or combinations thereof. The storage device may include, for example, dynamic random access memory (DRAM), flash memory, or non-volatile random access memory (NVRAM). In this embodiment, the storage device can be used to store algorithms, communication protocols, and applications related to implementing the automatic connection methods, steps, and distributed computing described in the embodiments of the present invention. In addition, the main computer device 110 and the slave computer devices 120_1 to 120_N may each include a communication interface, an input device, and other peripheral functional elements, wherein the communication interface is used to realize the connection between the devices described in the various embodiments of the present invention.

[0020] In this embodiment, the main computer device 110 and the slave computer devices 120_1 to 120_N may each include one or more video RAM (VRAM), dynamic random-access memory (DRAM), and / or non-volatile memory express solid-state drives (NVMe SSDs). In this embodiment, the main computer device 110 can split the large model to perform distributed training or related clustering operations on the model 220. The computing system 100 can split the large model into multiple sub-models and allocate them to the VRAM, DRAM, and / or non-volatile memory express solid-state drives of the main computer device 110 and the slave computer devices 120_1 to 120_N to selectively use the computing resources of the main computer device 110 and the slave computer devices 120_1 to 120_N to effectively distribute the training of the large model. In this embodiment, the large model may be, for example, a large language model (LLM) or a large multimodal model (LMM).

[0021] Figure 2 This is a flowchart of an operation method according to an embodiment of the present invention. (See reference) Figure 1 as well as Figure 2Taking the connection operation between the main computer device 110 and the slave computer device 120_1 as an example, the computing system 100 can perform the following steps S210 to S250. In this embodiment, the user can connect to the main computer device 110 and the slave computer device 120_1 via a network (Internet) connection through a terminal computer device, and can assign dedicated Internet Protocol (IP) addresses to the main computer device 110 and the slave computer device 120_1. In step S210, the slave computer device 120_1 can connect to the dedicated Internet Protocol address of the main computer device 110. In step S220, the user can create user accounts for passwordless connection authentication on the main computer device 110 and the slave computer device 120_1 respectively. In step S230, the slave computer device 120_1 can create key data, and the slave computer device 120_1 provides the public key of the key data to the main computer device 110. In step S240, the master computer device 110 can verify whether the slave computer device 120_1 has logged into the master computer device 110 without a password based on the public key of the slave computer device 120_1. In step S250, after the slave computer device 120_1 logs into the master computer device 110 without a password, the master computer device 110 and the slave computer device 120_1 can perform distributed computing to train a large model.

[0022] In this embodiment, the main computer device 110 and the slave computer devices 120_1 to 120_N can establish a passwordless login connection based on the Secure Shell Protocol (SSH), but the present invention is not limited thereto. Therefore, the computing system 100 of this embodiment can automatically and quickly establish a secure communication connection between the main computer device 110 and the slave computer device 120_1 to facilitate data transmission during distributed computing. In addition, the slave computer devices 120_2 to 120_N can also establish a passwordless login connection with the main computer device 110 by performing the operation method described above.

[0023] Figure 3 This is a flowchart illustrating authentication according to an embodiment of the present invention. (See reference) Figure 1 and Figure 3Taking the connection operation between the master computer device 110 and the slave computer device 120_1 as an example, the computing system 100 can perform authentication as follows: S301 to S305. In step S301, the master computer device 110 and the slave computer device 120_1 exchange keys and determine a session identifier (Session ID). In step S302, the slave computer device 120_1 signs the session identifier and sends an authentication request to the master computer device 110. In step S303, the master computer device 110 verifies the correctness of the session identifier based on the authentication request. In step S304, if the session identifier is correct, the master computer device 110 notifies the slave computer device 120_1 that authentication is successful. In step S305, the slave computer device 120_1 provides its public key and / or credential data to the master computer device 110 to complete the establishment of a communication channel. It is worth noting that when the master computer device 110 and the slave computer device 120_1 authenticate to establish a connection for the first time, the aforementioned generated session identifier can be saved. Therefore, even if the master computer device 110 and the slave computer device 120_1 re-exchange keys during subsequent communications, the same session identifier will still be maintained between them. Furthermore, slave computer devices 120_2 to 120_N can also perform the same authentication process with the master computer device 110.

[0024] Figure 4 This is a flowchart of a data transmission operation according to an embodiment of the present invention. (See reference) Figure 1 as well as Figure 4 The specific implementation details of the above steps S301 to S304 can be further described in... Figure 4 This is achieved through steps S410 to S438. In step S410, key exchange can occur between the master computer device 110 and the slave computer device 120_1. In step S411, the slave computer device 120_1 can issue an initial key exchange request. The slave computer device 120_1 can send a communication command (using a secure shell protocol as an example, such as "SSH_MSG_KEXINIT") to the master computer device 110. In step S412, the master computer device 110 can respond accordingly with a corresponding communication command (such as "SSH_MSG_KEXINIT") to the slave computer device 120_1.

[0025] In this embodiment, the master computer device 110 and the slave computer device 120_1 can each generate a temporary session key pair. In step S413, the slave computer device 120_1 can provide its public key to the master computer device 110 (e.g., via the communication command "SSH_MSG_KEXDH_INIT"). In this embodiment, the slave computer device 120_1 can authenticate the master computer device 110 based on its host public key. The slave computer device 120_1 can obtain shared secret data based on the host public key and obtain hash data based on the hash signature. The slave computer device 120_1 can generate a symmetric key and a session identifier for the data channel between the master computer device 110 and the slave computer device 120 based on the shared secret data and the hash data. In step S414, the master computer device 110 may send back the public key of the slave computer device 120_1, the host public key of the master computer device 110, and the hash signature to the slave computer device 120_1 (e.g., via the communication command "SSH_MSG_KEXDH_REPLY").

[0026] In step S415, the slave computer device 120_1 may notify the master computer device 110 (e.g., via the communication command "SSH_MSG_NEWKEYS") that the key exchange has been completed and a new session key has been generated and is ready for subsequent encrypted communication. In step S415, the master computer device 110 may also respond to the slave computer device 120_1 (e.g., via the communication command "SSH_MSG_NEWKEYS"), indicating that subsequent encrypted communication is possible.

[0027] In step S420, the slave computer device 120_1 can perform authentication. In step S421, the slave computer device 120_1 can send an authentication request to the master computer device 110 (e.g., via the communication command "SSH_MSG_SERVICE_REQUEST"). In step S422, the master computer device 110 can respond to the slave computer device 120_1 (e.g., via the communication command "SSH_MSG_SERVICE_ACCEPT") to acknowledge the authentication request. In step S423, the slave computer device 120_1 can sign the session identifier according to its private key and send an authentication request to the master computer device 110 (e.g., via the communication command "SSH_MSG_USERAUTH_REQUEST").

[0028] In this embodiment, the public key of the key data is copied from the slave computer device 120_1 to the authorized public key list of the master computer device 110 beforehand, so that the master computer device 110 can determine whether to allow authentication based on the authorized public key list. When the master computer device 110 allows authentication, it can verify the session identifier based on the public key of the slave computer device 120_1 to establish a data channel for training large models. In step S424, the master computer device 110 can notify the slave computer device 120_1 of successful authentication (e.g., via the communication command "SSH_MSG_USERAUTH_SUCCESS").

[0029] In step S430, the slave computer device 120_1 can perform remote operations (login) and execute commands. In step S431, the slave computer device 120_1 can request remote session service from the master computer device 110 (e.g., via the communication command "SSH_MSG_SERVICE_REQUEST"). In step S432, the master computer device 110 can respond to the slave computer device 120_1 with an agreement request (e.g., via the communication command "SSH_MSG_SERVICE_ACCEPT"). In step S433, the slave computer device 120_1 can request remote session service from the master computer device 110 (e.g., via the communication command "SSH_MSG_CHANNEL_OPEN"). In step S434, the master computer device 110 can verify the identity of the slave computer device 120_1 and respond to the slave computer device 120_1 with a channel opening confirmation (e.g., via the communication command "SSH_MSG_CHANNEL_OPEN_CONFIRMATION").

[0030] In step S435, the slave computer device 120_1 may specify a request type to the master computer device 110 (e.g., via the communication command "SSH_MSG_CHANNEL_REQUEST"). In step S436, the master computer device 110 may respond to the slave computer device 120_1 with a successful setup (e.g., via the communication command "SSH_MSG_CHANNEL_SUCCESS"). In step S437, the slave computer device 120_1 may send a data request for remote session service to the master computer device 110 (e.g., via the communication command "SSH_MSG_CHANNEL_DATA"). In step S438, when the data transmission is complete, the master computer device 110 may notify the slave computer device 120_1 to close the channel (e.g., via the communication command "SSH_MSG_CHANNEL_CLOSE").

[0031] Therefore, the computing system 100 of this embodiment can automatically and quickly establish a secure communication connection between the host computer device 110 and the slave computer device 120_1, to achieve, for example, SSH passwordless login and data transmission operations. In addition, the slave computer devices 120_2 to 120_N can also perform the above-described process with the host computer device 110, enabling the host computer device 110 to establish connections with multiple slave computer devices and facilitating the allocation of computational data required for training large models.

[0032] In summary, the computing system and its operating method of the present invention enable a secure communication connection to be quickly established between a master computer device and multiple slave computer devices, and the slave computer devices can connect to the master computer device without a password, so that the computing system can perform distributed computing through the memory computing resources of the master computer device and multiple slave computer devices.

[0033] Although the present invention has been disclosed above with reference to embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make some modifications and refinements without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A computing system, comprising: One main computer device; as well as A dedicated Internet Protocol address is connected from the computer device to the main computer device. The main computer device and the slave computer device each establish a user account for passwordless connection authentication. The slave computer device establishes a key data set, and provides a public key of this key data set to the master computer device. This allows the master computer device to verify whether the slave computer device has logged into the master computer device without a password, based on the slave computer device's public key. After the slave computer logs into the master computer without a password, the master computer and the slave computer perform distributed computing to train a large model.

2. The computing system of claim 1, wherein the master computer device and the slave computer device exchange keys and determine a session identifier. The slave computer device signs the session identifier and sends an authentication request to the master computer device. The host computer device verifies whether the session identifier is correct based on the authentication request. When the session identifier is correct, the master computer device notifies the slave computer device that authentication was successful.

3. The computing system of claim 2, wherein the master computer device and the slave computer device each generate a temporary session key pair, and the slave computer device provides its public key to the master computer device so that the master computer device sends back the slave computer device's public key, a host public key of the master computer device, and a hash signature to the slave computer device.

4. The computing system of claim 3, wherein the slave computer device performs authentication of the master computer device based on the host public key of the master computer device. The slave computer device obtains a shared secret data based on the host public key and a hash data based on the hash signature. The slave computer device generates a symmetric key and a session identifier for a data channel between the master computer device and the slave computer device based on the shared secret data and the hash data.

5. The computing system of claim 4, wherein the slave computer device signs the session identifier according to a private key.

6. The computing system of claim 4, wherein the slave computer device pre-copies the public key of the key data to an authorized public key list of the master computer device, so that the master computer device determines whether to allow authentication based on the authorized public key list.

7. The computing system of claim 6, wherein the master computer device verifies the session identifier as correct based on the public key of the slave computer device to establish the data channel for training the large model.

8. The computing system of claim 1, wherein the slave computer device is connected to the host computer device via a Thunderbolt cable.

9. A method of operating a computing system, wherein the computing system includes a master computer device and a slave computer device, wherein the method of operating the system includes: Through a dedicated Internet Protocol address connecting the computer device to the main computer device; On the main computer device and the slave computer device, a user account is created for passwordless connection authentication. A key data is created from the computer device, and a public key of the key data is provided to the host computer device. The main computer device uses the public key of the slave computer device to verify whether the slave computer device has logged into the main computer device without a password. as well as After the slave computer logs into the master computer without a password, distributed computing is performed through the master computer and the slave computer to train a large model.

10. The operating method as described in claim 9, wherein the step of confirming whether the slave computer device has logged into the master computer device without a password includes: The master computer and the slave computer exchange keys and determine a session identifier. The slave computer device signs the session identifier and sends an authentication request to the master computer device. The host computer device verifies the session identifier's correctness based on the authentication request; and When the session identifier is correct, the master computer device notifies the slave computer device that authentication was successful.

11. The operation method as described in claim 10, wherein the step of exchanging keys includes: The master computer device and the slave computer device each generate a temporary session key pair. as well as The slave computer device provides its public key to the master computer device, so that the master computer device sends back the slave computer device's public key, a host public key of the master computer device, and a hash signature to the slave computer device.

12. The method of operation as described in claim 11, wherein the step of exchanging keys further includes: The slave computer device authenticates the identity of the master computer device based on the host public key of the master computer device. The computer device obtains shared secret data based on the host's public key and obtains hash data based on the hash signature; and The slave computer device generates a symmetric key and a session identifier for a data channel between the master computer device and the slave computer device based on the shared secret data and the hash data.

13. The method of operation as described in claim 12, wherein the step of signing the session identifier includes: The session identifier is signed by the computer device using a private key.

14. The operating method as described in claim 12, further comprising: The public key of the gold key data is copied from the computer device to an authorized public key list on the main computer device in advance; as well as The host computer device determines whether to allow authentication based on the list of authorized public keys.

15. The method of operation as described in claim 14, wherein the step of verifying whether the session identifier is correct includes: The master computer device verifies the session identifier based on the public key of the slave computer device to establish the data channel used in the training of the large model.

Images