High-speed cable arrangement method of supercomputing cluster

Through technologies such as high-speed cable layout and fiber optic communication modules, the problem of long communication distance of computing chips in supercomputing clusters is solved, data exchange capability and computing performance are improved, cable connection and maintenance are simplified, and system efficiency is improved. reliability.

CN118966133APending Publication Date: 2024-11-15SHANGHAI SILANG WANWEI COMPUTING TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410999251.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The communication distance between computing chips in existing supercomputing clusters is long, resulting in large data transmission delays, and traditional cable connection methods are complex, high energy consumption and inconvenient maintenance, which affects computing performance.

Method used

Using high-speed cable arrangement method, the chip is embedded into the PCBA substrate to form a chip array, and the chip is connected through high-speed cable channels to form a cube architecture, shorten the communication distance, increase the data exchange capability, and integrate the optical fiber communication module and the calculation blade. Redundant power supply, set up intelligent fault detection and dynamic load balancing modules.

Benefits of technology

It shortens the communication distance between computing chips, reduces data transmission delay, improves data exchange capabilities, simplifies cable connection, reduces energy consumption and maintenance difficulty, and improves the overall computing performance of supercomputing clusters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118966133A_ABST
    Figure CN118966133A_ABST
Patent Text Reader

Abstract

The invention discloses a high-speed cable arrangement method of a supercomputing cluster, and belongs to the technical field of chip connection arrangement. The method comprises the following steps: S1, embedding a chip into a PCBA (Printed Circuit Board Assembly) substrate of a calculation blade to form a chip array; s2, a high-speed cable channel is arranged on the calculation blade and used for being connected with a communication interface of each calculation chip; wherein each chip is provided with six axial communication interfaces, three of the six axial communication interfaces realize signal connection among the chips through the PCBA substrate, and the other three axial communication interfaces are connected with other computing chips through high-speed cable channels; and S3, deploying the 2n computing blades embedded with the chips in a blade frame system, and connecting the 2n blade frame systems so as to arrange the chips into a cubic framework. Through a novel cable connection mode, the communication distance between the computing chips is shortened, the data transmission time delay is reduced, and the data exchange capacity between the chips is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of chip connection arrangement, and in particular to a high-speed cable arrangement method. Background Art

[0002] Supercomputing cluster, full name Super Computing Cluster (SCC), is a computer cluster based on elastic bare metal servers and connected with high-speed RDMA network. It consists of multiple independent computers that work closely together through high-speed communication technology to complete computing work.

[0003] With the continuous development of science and technology, supercomputer clusters are increasingly used in the field of high-performance computing. For example, in the fields of life sciences, drug development, industrial simulation, etc., large-scale parallel computing is required. Supercomputer clusters are most commonly used in the form of blade computers, which are efficient and scalable computer systems that use blade servers to improve computing power and reduce physical space and energy consumption. This type of computer system is widely used in large enterprises and data centers, and plays an important role in cloud computing, big data processing, and high-performance computing.

[0004] One of the advantages of blade computers is scalability: due to its modular design, it can be expanded as needed to accommodate growing computing needs. In addition, the density and efficiency of blade servers also make super blade computers more efficient in terms of space and energy use. Another advantage of blade computers is flexibility: since it can be configured and modified as needed, it can be used for a variety of different computing tasks. This allows super blade computers to remain efficient in the face of changing computing needs.

[0005] However, the communication distance between computing chips in blade computers used in existing supercomputing clusters is long, resulting in large data transmission delays, which affects the overall computing performance. In addition, traditional cable connection methods also have problems such as complex wiring and inconvenient maintenance. Therefore, how to shorten the communication distance between computing chips, reduce data transmission delays, and improve the data exchange capabilities between chips are issues that need to be urgently addressed in the current field of supercomputing cluster technology.

[0006] In the process of implementing the present invention, the inventors found that the prior art has at least the following problems: Traditional supercomputing clusters use standard server nodes and interconnect the nodes through standard network cables or fiber optic jumpers. This approach has many disadvantages, such as bandwidth limitations leading to data processing bottlenecks, limited scalability, high energy consumption, complex wiring, and possible impact on reliability and stability. Due to these limitations, traditional supercomputing clusters may face performance bottlenecks when processing large-scale parallel computing tasks. In order to overcome these challenges and improve the performance of supercomputing clusters, it is necessary to explore and adopt new connection methods and technologies. Summary of the invention

[0007] In order to solve the problems of the prior art, an embodiment of the present invention provides a method for arranging high-speed cables of a supercomputing cluster. The technical solution is as follows:

[0008] The method comprises the following steps:

[0009] S1: Embed the chip into the PCBA substrate of the computing blade to form a chip array; wherein the number of chips embedded in each PCBA substrate is 2 n indivual;

[0010] S2: A high-speed cable channel is set on the computing blade to connect to the communication interface of each computing chip; each chip has a total of 6 axial communication interfaces, namely, X+, X-, Y+, Y-, Z+ and Z-, of which 3 axial communication interfaces realize signal connection between chips through the PCBA substrate, and the other 3 axial communication interfaces are connected to other computing chips through high-speed cable channels;

[0011] S3: 2 n The computing blades embedded with chips are deployed in a blade frame system, and 2 n The blade frame system is connected to arrange the chip into 2 n ×2 n ×2 n Cube architecture; each computing blade has 3×2 n A high-speed interface is connected to the outside.

[0012] Preferably, in step S1, the number of chips in each computing blade is 4 or 8.

[0013] Preferably, in step S3, the chips in the cube architecture are connected in the following manner:

[0014] The chips in the same blade frame system are connected at intervals at the front and rear parts of the computing blade in the X-axis direction, and are connected adjacently at the front part of the computing blade in the Z-axis direction;

[0015] The chips in the same blade frame system are horizontally connected at the front and rear of the computing blade in the X-axis and Y-axis directions.

[0016] Preferably, in step S3, the chips in the cube architecture are also connected in the following manner:

[0017] The chips in the first, second, fifth, sixth, ... computing blades in each blade frame system are vertically connected across the blade frame system at the rear of the computing blade in the Z-axis direction, and are vertically connected across the blade frame system at the front and rear of the computing blade in the Y-axis direction;

[0018] The chips in the 3rd, 4th, 7th, 8th ... computing blades in each blade frame system are vertically connected across the blade frame system at the rear of the computing blade in the Z-axis direction, and are vertically connected across the blade frame system at the front and rear of the computing blade in the Y-axis direction.

[0019] Preferably, in step S2, the number of chips in each computing blade is 8, namely chip M1, chip M2, chip M3, chip M4, chip M5, chip M6, chip M7 and chip M8; in the chip X-axis direction, chip M1 is connected to chip M2, chip M3 is connected to chip M4, chip M5 is connected to chip M6, and chip M7 is connected to chip M8.

[0020] Preferably, in the chip Y-axis direction, chip M1 is connected to chip M3 , chip M2 is connected to chip M4 , chip M5 is connected to chip M7 , and chip M6 is connected to chip M8 .

[0021] Preferably, in the chip Z-axis direction, chip M1 is connected to chip M5 , chip M2 is connected to chip M6 , chip M3 is connected to chip M7 , and chip M4 is connected to chip M8 .

[0022] Preferably, in step S3, the supercomputing cluster is composed of 2 cabinets, and each cabinet is composed of 4 blade frame systems.

[0023] Preferably, the method further comprises:

[0024] S4: Add a heat exchange device on the outside of each computing blade.

[0025] Preferably, the heat exchange device comprises a microchannel radiator and a heat pipe, and performs circulation cooling through a liquid cooling system.

[0026] Preferably, the method further comprises:

[0027] S5: A set of independent redundant power supply modules is set in each blade frame system of the supercomputing cluster.

[0028] Preferably, the redundant power supply module includes an uninterruptible power supply system and an intelligent power management system, which can automatically switch power sources and monitor power status.

[0029] Preferably, the method further comprises:

[0030] S6: Integrate fiber optic communication modules on each computing blade to achieve high-speed fiber optic data transmission, further improving the data transmission rate.

[0031] Preferably, the fiber optic communication module supports multi-mode and single-mode optical fibers.

[0032] Preferably, the method further comprises:

[0033] S7: An intelligent fault detection and diagnosis module is set up in each blade frame system of the supercomputing cluster to monitor and report system faults in real time.

[0034] Preferably, the intelligent fault detection and diagnosis module includes adaptive algorithms and machine learning techniques, which can predict potential faults and provide repair suggestions.

[0035] Preferably, the method further comprises:

[0036] S8: Configure a dynamic load balancing module on each computing blade in the supercomputing cluster to optimize the allocation and utilization of computing resources.

[0037] Preferably, the dynamic load balancing module adopts an artificial intelligence-based algorithm to adjust the distribution of computing tasks according to the real-time workload to maximize resource utilization.

[0038] The technical solution provided by the embodiment of the present invention has the following beneficial effects: by adopting a new cable connection method, the communication distance between computing chips is shortened, the data transmission delay is reduced, the data exchange capability between chips is improved, and a strong guarantee is provided for the overall computing performance of the supercomputing cluster; at the same time, it also has the advantages of shortening the cable connection length, simple and easy operation, convenient maintenance, and beautiful appearance. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 It is a flow chart of the high-speed cable arrangement method of the supercomputing cluster provided by the present invention;

[0041] Figure 2 is a schematic diagram of a cube architecture formed by a high-speed cable arrangement method for a supercomputing cluster according to the present invention;

[0042] Figure 3It is a schematic diagram of the rear wiring of a computing blade in a supercomputing cluster formed according to the high-speed cable arrangement method of the supercomputing cluster of the present invention;

[0043] Figure 4 The diagram is a front wiring diagram of computing blades in a supercomputing cluster formed according to the high-speed cable arrangement method of the supercomputing cluster of the present invention. DETAILED DESCRIPTION

[0044] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, a high-speed cable arrangement method for a supercomputing cluster according to an embodiment of the present invention is mainly used in supercomputing clusters. Supercomputing clusters are mainly composed of a computing chip array supporting three-dimensional communication. Each computing chip has six axes of "X+, X-, Y+, Y-, Z+, Z-" and is connected to adjacent chips in pairs. Among them, three axes of each computing chip have been connected inside the PCBA substrate, and the other three axes need to be connected by high-speed cables to form a cube matrix of chips.

[0046] Embodiment 1:

[0047] A high-speed cable arrangement method for a supercomputing cluster of the present invention comprises the following steps:

[0048] S1: Embed the chip into the PCBA substrate of the computing blade to form a chip array; wherein the number of chips embedded in each PCBA substrate is 2 n (n>=1).

[0049] Taking into account the area and power consumption of the PCBA board, it is preferred that the number of chips per computing blade be 4 or 8.

[0050] S2: A high-speed cable channel is set on the computing blade to connect to the communication interface of each computing chip. Each chip has six axial communication interfaces, namely "X+, X-, Y+, Y-, Z+, and Z-". Three axial communication interfaces realize high-speed signal connection between chips through the PCBA substrate, and the other three axial communication interfaces are connected to the communication interfaces of other computing chips through high-speed cable channels.

[0051] Taking the example of 8 chips on each computing blade, namely M1, M2, M3, M4, M5, M6, M7 and M8, the chips inside the motherboard are connected along the X-axis direction M1 and M2, M3 and M4, M5 and M6, and M7 and M8; the Y-axis direction is connected to M1 and M3, M2 and M4, M5 and M7, and M6 and M8; the Z-axis direction is connected to M1 and M5, M2 and M6, M3 and M7, and M4 and M8; the chips are connected in positive and negative directions, so the axes along which each chip is connected inside the motherboard and to the chips outside the motherboard are different.

[0052] S3: 2 n The computing blades embedded with chips are deployed in a blade frame system, and 2 n The blade frame system is connected to arrange the chip into 2 n ×2 n ×2 n Cube architecture; each computing blade has 3×2 n A high-speed interface is connected to the outside.

[0053] In the present invention, in the cube architecture, taking 8 chips on each computing blade as an example, 8 chips form 1 computing blade, 8 computing blades form 1 blade frame system, 8 blade frame systems form 1 supercomputing cluster, and the chips are connected in the following manner:

[0054] 1. The chips in the same blade frame system are connected at intervals at the front and rear parts of the computing blade in the X-axis direction, and are adjacently connected at the front part of the computing blade in the Z-axis direction;

[0055] 2. The chips in the same blade frame system are horizontally connected at the front and rear of the computing blade in the X-axis and Y-axis directions;

[0056] 3. The chips in the first, second, fifth, sixth, ... computing blades in each blade frame system are vertically connected across the blade frame system at the rear of the computing blade in the Z-axis direction, and are vertically connected across the blade frame system at the front and rear of the computing blade in the Y-axis direction;

[0057] 4. The chips in the third, fourth, seventh, eighth, ... computing blades in each blade frame system are vertically connected across the blade frame system at the rear of the computing blade in the Z-axis direction, and are vertically connected across the blade frame system at the front and rear of the computing blade in the Y-axis direction;

[0058] Compared with traditional wiring methods, the high-speed cables arranged in the above manner shorten the cable length, can intuitively identify the cable direction, facilitate maintenance, and reduce the difficulty of operation and maintenance. In addition, the present invention makes the node layout of the supercomputing cluster more compact, further shortens the cable connection length, simplifies the wiring process, improves the convenience of maintenance, and makes the entire system more beautiful.

[0059] like Figure 2 As shown, taking 8 chips as an example, each computing blade (for example, a computing blade composed of 8 chips numbered 000, 002, ..., 007) has 24 high-speed interfaces for external connection; a computing blade has 8 chips, each chip has 6 axial directions, 3 of which are connected to the chip in the blade, and each chip has 3 axial directions connected to the chips across the blade, 3*8 in total for 24 high-speed interfaces, 12 high-speed connection interfaces are preset at the front and back of the blade frame system, totaling 24, and the chips can be connected into an "8×8×8" cube architecture by connecting them through the high-speed cable of the present invention.

[0060] Figure 2 A connection diagram of a 3D supercomputing cluster formed by interconnecting 512 ("8×8×8" cube architecture cluster) chips is shown in FIG. In the method of the present invention, 8 computing chips are placed on each computing blade in the cluster constructed by 512 chips, and 512 chips are placed on 64 computing blades in total. For ease of management and construction, these 64 computing blades are deployed in 8 blade frame systems, each of which has 8 computing blades.

[0061] The six axes of the chip need to be connected to each other. Three of them have already realized the communication connection between chips on the motherboard of the computing blade, and the other three axes need to be connected through external high-speed cables. Therefore, each computing blade has 24 high-speed interfaces for external connection. The blade frame system has 12 high-speed connection interfaces pre-set at the front and back. Each blade frame system has 8 computing blades, and each blade frame has 24*8=192 interfaces. There are 8 blade frames in total, with a total of 24*8*8, a total of 1536 high-speed connection interfaces. Since they are connected in pairs, the entire cluster needs 24*8*8 / 2=768 high-speed cables for connection, thus forming an "8×8×8" 512-cube supercomputing cluster.

[0062] like Figure 3 and Figure 4 As shown in the figure, the supercomputer cluster consists of 2 cabinets, each cabinet consists of 4 blade frame systems. It can be understood that the supercomputer cluster has 4 layers, and each layer has 2 blade frame systems. Figure 3 and Figure 4 The left half is the two-blade system on the first layer of the supercomputing cluster. The entire cluster has 8 types of cables and a total of 768 cables.

[0063] In order to shorten the communication distance between computing chips, reduce data transmission delay, and improve the data exchange capability between chips, the following 8 specifications of high-speed cables are obtained through the connection method of the high-speed cable of the present invention. The specific length and quantity of each cable are indicated in the following table, including the description of the connection method. The connection method of the physical cable is shown in the following table:

[0064]

[0065] By adopting a new cable connection method, compared with the traditional chip connection, the present invention shortens the communication distance between computing chips, reduces data transmission delay, and improves the data exchange capability between chips, providing a strong guarantee for the overall computing performance of the supercomputing cluster; at the same time, it also has the advantages of shortening the cable connection length, simple and easy operation, convenient maintenance, and beautiful appearance.

[0066] Embodiment 2:

[0067] On the basis of Example 1, the purpose of this Example 2 is to further optimize the heat dissipation and power management of the supercomputing cluster. Therefore, on the basis of Example 1, a high-speed cable arrangement method for a supercomputing cluster of the present invention further includes the following steps:

[0068] S4: Application of heat exchange device:

[0069] A heat exchange device is added to the outside of each computing blade. The heat exchange device uses a microchannel radiator and a heat pipe and performs circulating cooling through a liquid cooling system.

[0070] The working principle of the microchannel radiator mainly relies on the flow and heat conduction properties of the fluid in the tiny channel. Specifically, the microchannel heat dissipation system consists of microchannels, coolant and heat source. The rapid flow of coolant in the microchannel can take away the heat from the heat source and achieve efficient heat dissipation. The microchannel radiator can efficiently absorb the heat generated by the chip, and the heat pipe transfers the heat to the liquid cooling system, which takes away the heat through the circulating coolant to achieve efficient heat dissipation. This design can significantly reduce the temperature of the supercomputing cluster and ensure that the system remains stable during high-performance computing.

[0071] S5: Setting of redundant power supply module:

[0072] A set of independent redundant power supply modules is arranged in each knife frame system, including an uninterruptible power supply (UPS) system and an intelligent power management system.

[0073] This setting is designed to provide a highly reliable and stable power supply guarantee for the knife frame system. The uninterruptible power supply (UPS) system can quickly and seamlessly take over the power supply when the main power fails or the voltage fluctuates, and quickly switch to the backup power supply to ensure that the knife frame system will not stop running due to a momentary power outage, effectively preventing data loss and system failure. The intelligent power management system can monitor the power status in real time, including current, voltage, power and other parameters, and intelligently allocate and adjust the power supply according to the load requirements of the system, so as to achieve efficient use of power, automatically adjust power distribution, optimize power use efficiency, and further enhance the reliability and stability of the entire power supply.

[0074] S6: Integrate optical fiber communication modules on each computing blade to achieve high-speed optical fiber data transmission, further improving the data transmission rate. The optical fiber communication module supports multi-mode and single-mode optical fibers, which is suitable for high-speed data transmission requirements at different distances.

[0075] Fiber optic communication has extremely high bandwidth and extremely fast data transmission rate. By integrating the fiber optic communication module directly into the computing blade, data can be quickly transmitted between devices at the speed of light, greatly reducing the delay of data transmission.

[0076] In addition, high-speed fiber optic data transmission also helps to achieve efficient collaboration between computing blades. In large-scale parallel computing, each computing blade needs to exchange data frequently. Fast data transmission can improve the parallel processing efficiency of computing tasks, shorten computing time, and thus improve the performance of the entire supercomputing cluster.

[0077] The multimode optical fiber used in the fiber optic communication module has a larger core diameter and can support the simultaneous transmission of multiple optical modes. This makes it perform well in data transmission over shorter distances, such as within the same computer room or between adjacent cabinets. Multimode optical fiber is relatively low in cost and can provide higher bandwidth over short distances, meeting the needs of short-distance high-speed data exchange within supercomputing clusters.

[0078] The single-mode optical fiber used in the optical fiber communication module has a smaller core diameter and only allows single optical mode transmission. Although its cost is relatively high, it has lower attenuation and higher transmission performance during long-distance transmission. This makes single-mode optical fiber suitable for long-distance high-speed data transmission between supercomputing clusters and remote data centers, scientific research institutions or other supercomputing facilities, ensuring that data can remain stable and high-speed during long-distance transmission.

[0079] By supporting both multi-mode and single-mode optical fibers, the optical fiber communication module can be flexibly selected and configured according to the needs of high-speed data transmission at different distances. Whether it is a short-distance connection within the computer room or a long-distance communication with external facilities, it can provide the best transmission performance and reliability, providing a strong guarantee for the efficient operation of the supercomputing cluster.

[0080] Embodiment 3:

[0081] On the basis of Embodiment 1 and Embodiment 2, the purpose of Embodiment 3 is to further enhance the intelligent monitoring and resource optimization capabilities of the supercomputing cluster. Therefore, on the basis of Embodiment 1 and Embodiment 2, a high-speed cable arrangement method for a supercomputing cluster of the present invention further includes the following steps:

[0082] S7: Intelligent fault detection and diagnostics:

[0083] An intelligent fault detection and diagnosis module is set up in each tool frame system, and the system operation status is monitored in real time using adaptive algorithms and machine learning technology.

[0084] The adaptive algorithm can automatically adjust monitoring parameters and thresholds according to the dynamic changes of system operation to adapt to different workloads and environmental conditions. For example, when the system load is high, the monitoring sensitivity of key component temperature, current and other parameters can be increased accordingly. Machine learning technology enables the module to learn and identify normal and abnormal patterns from a large amount of historical operation data. By continuously accumulating and analyzing new data, its diagnostic capabilities are continuously optimized and improved.

[0085] By monitoring the system's operating status in real time, potential signs of failure can be discovered in a timely manner, such as abnormal heating of components, current fluctuations, data transmission errors, etc. Once an abnormality is detected, the module will quickly analyze and diagnose the type, location and possible cause of the failure.

[0086] With the above settings, the fault detection and diagnosis module can detect and report hardware and software anomalies, predict potential failures, and provide repair suggestions. For example, when an abnormal temperature rise is detected on a computing blade, the system can suggest checking the cooling system or replacing the faulty component.

[0087] This step S7 helps to take maintenance measures in advance, reduce system downtime, improve the reliability and availability of the cutter frame system, reduce maintenance costs, and ensure the stable operation of the entire system.

[0088] S8: Dynamic load balancing:

[0089] A dynamic load balancing module is configured on each computing blade, using an artificial intelligence-based algorithm to optimize the allocation and utilization of computing resources according to real-time workload conditions.

[0090] The AI-based algorithm enables the module to intelligently perceive and analyze real-time workload conditions. By collecting and processing a large amount of data about computing task requirements, computing resource usage, and system performance, the algorithm can accurately evaluate the current load distribution. For example, when the workload is unbalanced, such as some computing blades being overly busy while others are relatively idle, the dynamic load balancing module will respond quickly and optimize the allocation and utilization of computing resources based on pre-trained models and strategies. This may include migrating some tasks from busy blades to idle blades, or adjusting the allocation ratio of computing resources to ensure that each computing blade can get a reasonable workload, thereby giving full play to the computing power of the entire system.

[0091] Through the above settings, the dynamic load balancing module monitors the working status of each chip and dynamically adjusts the distribution of tasks to avoid overloading or idleness of some chips. For example, when a computing task requires a large amount of computing resources, the system can automatically allocate more chips for processing to improve the overall computing efficiency.

[0092] Embodiment 4:

[0093] On the basis of Embodiment 1, Embodiment 2 and Embodiment 3, the purpose of Embodiment 4 is to further improve the flexibility and security of the supercomputing cluster. Therefore, on the basis of Embodiment 1, Embodiment 2 and Embodiment 3, a high-speed cable arrangement method for a supercomputing cluster of the present invention further includes the following steps:

[0094] S9: Modular Design:

[0095] The supercomputing cluster is designed as a modular structure, so that computing blades, blade frame systems and cabinets can be easily replaced and upgraded.

[0096] Through this modular design, each key component, such as computing blades, blade frame system and cabinet, is designed as a relatively independent and pluggable module. For computing blades, when computing performance needs to be improved or a faulty blade needs to be replaced, the old blade can be directly pulled out and a new blade with higher performance or repaired can be inserted without making large-scale changes to the entire system.

[0097] Through this modular design, users can flexibly adjust the scale and configuration of the supercomputing cluster according to their needs. For example, when computing power needs to be increased, new computing blades or blade frame systems can be simply added without replacing the entire cabinet.

[0098] This convenient replacement and upgrade capability not only reduces the difficulty and cost of maintenance and upgrades, but also greatly shortens system downtime, improves the availability and flexibility of supercomputing clusters, and can better adapt to changing computing needs and technological developments.

[0099] S10: Safety precautions:

[0100] A security protection module is set up in each computing blade and blade frame system of the supercomputing cluster, including hardware encryption, identity authentication, access control and other measures.

[0101] The hardware encryption function provides underlying protection for data. By performing encryption operations at the hardware level, the confidentiality and integrity of data during storage and transmission are ensured. Even in extreme cases, such as device theft or illegal access to data, unauthorized personnel cannot decipher the encrypted data content.

[0102] Identity authentication measures ensure that only legitimate users and devices can access the system. They may include the use of passwords, biometric technologies (such as fingerprints, iris recognition, etc.), digital certificates, and other authentication methods to strictly authenticate users or devices attempting to access the system to prevent unauthorized access.

[0103] Access control further refines the user's operating rights on system resources. Users of different levels are granted different access rights, such as read-only, read-write, modify, etc. At the same time, access time, location, network environment, etc. can be restricted to ensure that users can only access their authorized resources under legal conditions, thereby minimizing internal and external security risks.

[0104] The security protection module can protect the confidentiality and integrity of data and prevent unauthorized access and operation. For example, when performing remote access, the system can require multi-factor authentication to ensure that only authorized users can access sensitive data and resources.

[0105] Embodiment 5:

[0106] On the basis of Embodiment 1, Embodiment 2, Embodiment 3 and Embodiment 4, the purpose of this Embodiment 5 is to further improve the management and monitoring capabilities of the supercomputing cluster. Therefore, on the basis of Embodiment 1, Embodiment 2, Embodiment 3 and Embodiment 4, a high-speed cable arrangement method for a supercomputing cluster of the present invention further includes the following steps:

[0107] S11: Centralized Management Platform:

[0108] Establish a centralized management platform to monitor and manage the entire supercomputing cluster.

[0109] The centralized management platform can collect a large amount of data from various computing blades, blade frame systems and related infrastructure in real time, including but not limited to hardware status (such as CPU temperature, memory usage, hard disk health, etc.), system performance indicators (such as computing speed, network bandwidth utilization, etc.), task execution progress and resource allocation, etc. By integrating and analyzing these data, managers can intuitively understand the operation status of the entire supercomputing cluster on a unified interface. No matter which part has an abnormality or potential problem, it can be detected and warned in time, so that corresponding measures can be taken to intervene and repair, avoid the expansion of the problem, and ensure the stability and reliability of the system.

[0110] Through this centralized management platform, the operating data of each computing blade and blade frame system can be collected and analyzed, and a unified interface can be provided for system configuration and maintenance. For example, administrators can view the operating status of the supercomputing cluster in real time through the centralized management platform, and perform remote diagnosis and troubleshooting.

[0111] S12: Environmental Monitoring System:

[0112] An environmental monitoring system is set up in the cabinet of the supercomputing cluster, including monitoring sensors such as temperature, humidity, air flow rate and power consumption.

[0113] Temperature sensors can sense temperature changes inside the cabinet in real time. Since supercomputing clusters generate a lot of heat when running, if the temperature is too high and cannot be regulated in time, it may cause hardware performance degradation, failure or even damage. Temperature sensors can allow managers to understand the temperature status in time so that they can take effective heat dissipation measures, such as adjusting fan speed and starting refrigeration equipment.

[0114] Humidity sensors are used to monitor humidity levels in cabinets. Too high or too low humidity may have adverse effects on electronic equipment, such as causing short circuits, corrosion, or static electricity accumulation. Through the data feedback of the humidity sensor, corresponding dehumidification or humidification measures can be taken to maintain a suitable humidity environment.

[0115] The air velocity sensor can monitor the airflow in the cabinet. This helps to evaluate the working effect of the cooling system and determine whether there is a problem of local airflow obstruction, so as to optimize the air duct design and fan layout in the cabinet and improve the cooling efficiency.

[0116] Power consumption sensors can monitor the power consumption of each cabinet in real time. This not only helps to understand the energy consumption distribution of the supercomputing cluster and provide a basis for energy-saving optimization, but also can promptly detect abnormal power consumption and warn of possible power failures or equipment leakage.

[0117] Through the above settings, the environmental monitoring system can monitor the environmental conditions in the cabinet in real time to ensure that the computing equipment operates in the best state. For example, when the temperature is detected to be too high, the system can automatically adjust the working state of the cooling system to prevent the equipment from overheating.

[0118] The newly added intelligent fault detection, dynamic load balancing, modular design, safety protection measures, centralized management platform and environmental monitoring system have further improved the intelligence, flexibility and security of the supercomputing cluster, enabling it to cope with complex computing tasks more efficiently while ensuring the reliability and maintainability of the system.

[0119] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A high-speed cable arrangement method for a supercomputing cluster, characterized in that: The method comprises the following steps: S1: Embed the chip into the PCBA substrate of the computing blade to form a chip array; wherein the number of chips embedded in each PCBA substrate is 2 n indivual; S2: A high-speed cable channel is set on the computing blade to connect to the communication interface of each computing chip; each chip has a total of 6 axial communication interfaces, namely, X+, X-, Y+, Y-, Z+ and Z-, of which 3 axial communication interfaces realize signal connection between chips through the PCBA substrate, and the other 3 axial communication interfaces are connected to other computing chips through high-speed cable channels; S3: 2 n The computing blades embedded with chips are deployed in a blade frame system, and 2 n The two blade frame systems are connected to arrange the chips into 2 n ×2 n ×2 n Cube architecture; each computing blade has 3×2 n A high-speed interface is connected to the outside.

2. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: In step S1, the number of chips in each computing blade is 4 or 8.

3. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: In step S3, the chips in the cube architecture are connected in the following manner: The chips in the same blade frame system are connected at intervals at the front and rear parts of the computing blade in the X-axis direction, and are connected adjacently at the front part of the computing blade in the Z-axis direction; The chips in the same blade frame system are horizontally connected at the front and rear of the computing blade in the X-axis and Y-axis directions.

4. The high-speed cable arrangement method for a supercomputing cluster according to claim 3, characterized in that: In step S3, the chips in the cube architecture are also connected in the following manner: The chips in the first, second, fifth, sixth, ... computing blades in each blade frame system are vertically connected across the blade frame system at the rear of the computing blade in the Z-axis direction, and are vertically connected across the blade frame system at the front and rear of the computing blade in the Y-axis direction; The chips in the 3rd, 4th, 7th, 8th ... computing blades in each blade frame system are vertically connected across the blade frame system at the rear of the computing blade in the Z-axis direction, and are vertically connected across the blade frame system at the front and rear of the computing blade in the Y-axis direction.

5. The high-speed cable arrangement method for a supercomputing cluster according to claim 2, characterized in that: In step S2, each computing blade has 8 chips, namely chip M1, chip M2, chip M3, chip M4, chip M5, chip M6, chip M7 and chip M8; in the chip X-axis direction, chip M1 is connected to chip M2, chip M3 is connected to chip M4, chip M5 is connected to chip M6, and chip M7 is connected to chip M8.

6. The high-speed cable arrangement method for a supercomputing cluster according to claim 5, characterized in that: In the chip Y-axis direction, chip M1 is connected to chip M3, chip M2 is connected to chip M4, chip M5 is connected to chip M7, and chip M6 is connected to chip M8.

7. The high-speed cable arrangement method for a supercomputing cluster according to claim 6, characterized in that: In the chip Z-axis direction, chip M1 is connected to chip M5, chip M2 is connected to chip M6, chip M3 is connected to chip M7, and chip M4 is connected to chip M8.

8. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: In step S3, the supercomputing cluster consists of 2 cabinets, and each cabinet consists of 4 blade frame systems.

9. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: The method further comprises: S4: Add a heat exchange device on the outside of each computing blade.

10. The high-speed cable arrangement method for a supercomputing cluster according to claim 9, characterized in that: The heat exchange device includes a microchannel radiator and a heat pipe, and performs circulation cooling through a liquid cooling system.

11. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: The method further comprises: S5: A set of independent redundant power supply modules is set in each blade frame system of the supercomputing cluster.

12. The high-speed cable arrangement method for a supercomputing cluster according to claim 11, characterized in that: The redundant power supply module includes an uninterruptible power supply system and an intelligent power management system, which can automatically switch power sources and monitor power status.

13. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: The method further comprises: S6: Integrate fiber optic communication modules on each computing blade to achieve high-speed fiber optic data transmission, further improving the data transmission rate.

14. The high-speed cable arrangement method for a supercomputing cluster according to claim 13, characterized in that: The fiber optic communication module supports multi-mode and single-mode optical fibers.

15. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: The method further comprises: S7: An intelligent fault detection and diagnosis module is set up in each blade frame system of the supercomputing cluster to monitor and report system faults in real time.

16. The high-speed cable arrangement method for a supercomputing cluster according to claim 15, characterized in that: The intelligent fault detection and diagnosis module includes adaptive algorithms and machine learning techniques to predict potential faults and provide repair suggestions.

17. The high-speed cable arrangement method for a supercomputing cluster according to claim 1, characterized in that: The method further comprises: S8: Configure a dynamic load balancing module on each computing blade in the supercomputing cluster to optimize the allocation and utilization of computing resources.

18. The high-speed cable arrangement method for a supercomputing cluster according to claim 17, characterized in that: The dynamic load balancing module adopts an artificial intelligence-based algorithm to adjust the distribution of computing tasks according to the real-time workload to maximize resource utilization.