Portable real scene three-dimensional modeling algorithm platform

Abstract

The application discloses a kind of portable real scene three-dimensional modeling computing power platform, including software and hardware integrated machine, software and hardware integrated machine includes power module, mainboard module, display card module, fan module, the computing power platform of the present application is built-in real scene three-dimensional modeling engine scheduling platform, according to the priority of task automatically scheduling modeling project, scheduling platform is according to the urgency, complexity and resource occupancy of task, automatically allocate modeling engine to process, reduce the demand of manual intervention, optimize the utilization of computing power resources, improve the processing speed of real scene three-dimensional modeling project;Cloud management platform carries out real-time monitoring to resource pool and virtual machine, can according to monitoring condition to hardware resource scheduling;Software and hardware integrated machine has dustproof, shockproof function, ensure that equipment is stably operated in complex outdoor environment, overall structure is light and portable, so that user easily carries and deploys in different environment to carry out three-dimensional modeling operation.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling computing power platform technology, specifically a portable real-scene 3D modeling computing power platform. Background Technology

[0002] Currently, with the development of information technology, the demand for real-scene 3D modeling is increasing in many fields, such as urban planning, geographic information, architectural design, and emergency disaster relief. In particular, emergency disaster relief scenarios have increasingly higher requirements for the efficiency of real-scene 3D modeling. However, traditional emergency scenario modeling has the following drawbacks: In traditional emergency scenario 3D modeling, it is usually necessary to rely on staff to carry mobile workstations for modeling. It is difficult to complete the overall modeling in a short time. At the same time, there may be problems such as the inconvenience of carrying multiple mobile workstations, insufficient modeling computing power of mobile workstations, inability to uniformly schedule mobile workstations, low modeling efficiency and inability to automatically allocate engines. Traditional mobile workstations often cannot cope well with this scenario. Summary of the Invention

[0003] The purpose of this invention is to provide a portable real-scene 3D modeling computing platform to solve the problems mentioned in the background art. In traditional emergency scenario real-scene 3D modeling, it is usually necessary for staff to carry mobile workstations for modeling, which makes it difficult to complete the overall modeling in a short time. At the same time, there may be problems such as the inconvenience of carrying multiple mobile workstations, insufficient modeling computing power of mobile workstations, inability to uniformly schedule mobile workstations, low modeling efficiency and inability to automatically allocate engines. Traditional mobile workstations often cannot cope well with the problems in this scenario.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a portable real-scene 3D modeling computing platform, comprising an all-in-one hardware and software machine, characterized in that the all-in-one hardware and software machine includes a chassis, a power supply module, a motherboard module, a graphics card module, and a fan module. The power supply module of the all-in-one hardware and software machine is located at the bottom of the chassis, with a total of 4 power supplies, adopting a redundant design. The motherboard module of the all-in-one hardware and software machine is located above the power supply module, and the motherboard module integrates hardware such as CPU, memory, PCIe slots, and I / O interfaces. The graphics card module of the all-in-one hardware and software machine is located above the motherboard module, and the graphics card module is connected to the PCIe slot of the lower motherboard module and the graphics card expansion board of the upper graphics card module through a PCIe splitter card. Eight graphics cards are fixedly installed on the graphics card expansion board. The fan module of the all-in-one hardware and software machine is fixed inside the front panel of the chassis, and the fan module is equipped with two layers of fans, forming two independent air ducts, which respectively cool the motherboard module and the graphics card module.

[0005] As a preferred embodiment of the present invention, the four corners of the top and four corners of the bottom of the all-in-one hardware and software device are fixedly installed with shock-absorbing corner protection mechanisms. The shock-absorbing corner protection mechanisms include shock-absorbing pads and anti-slip strips. The top of the shock-absorbing pads is fixedly connected to the bottom of the anti-slip strips. The surface of the shock-absorbing pads is provided with several grooves. The all-in-one hardware and software device includes a top panel, a front panel, a rear panel, and side panels. The bottom of the top panel is fixedly connected to the top of the front panel and the top of the rear panel, respectively. Side panels are fixedly installed on both sides between the front panel and the rear panel. A power button is fixedly installed on the top of the front panel.

[0006] As a preferred embodiment of the present invention, a mesh cover is fixedly installed on one side of the all-in-one hardware and software machine, and several heat dissipation slots are opened in the middle of the top of the all-in-one hardware and software machine. The opening of the heat dissipation slots and the mesh cover facilitates the contact between the low-temperature gas in the external environment and the all-in-one hardware and software machine, so as to transfer heat and dissipate heat to the all-in-one hardware and software machine. Handles are fixedly installed on both sides of the top of the all-in-one hardware and software machine, so that the user can carry the all-in-one hardware and software machine by holding it with his / her hands.

[0007] The workflow of a portable real-scene 3D modeling computing platform of the present invention includes the following steps: Step 1: Hardware Resource Setup: Integrate multi-core CPUs, large-capacity memory, multiple GPUs, and storage into a unified hardware resource pool to achieve efficient sharing and dynamic adjustment of computing resources; Step 2: Firecracker Virtualization; The underlying system uses resource virtualization technology to virtualize hardware resources into multiple virtual machine instances. Each virtual machine runs modeling tasks independently. Firecracker virtualization simplifies the virtualization architecture, reduces the resource overhead of traditional virtualization, and combines isolation and high efficiency. It also provides performance advantages by optimizing the startup speed of virtual machines. Step 3: Virtual machine performance optimization based on IO_uring: The IO_uring asynchronous I / O mechanism significantly reduces the overhead of system calls and context switching, making it particularly suitable for high-concurrency and I / O-intensive tasks and improving the performance of I / O operations; Step 4: Unified Management by Cloud Management Platform: The computing power platform has a built-in cloud management platform. Resource pooling technology centrally manages the computing resources of the computing power platform. Through a unified interface, users can call different virtual machine resources through the cloud management platform to complete the allocation and scheduling of modeling tasks. Pooling management makes resource use more efficient and facilitates real-time monitoring to ensure the stability of task execution. Step 5: Real-scene 3D modeling engine scheduling: The computing platform has a built-in real-scene 3D modeling engine scheduling platform, which automatically schedules modeling projects according to task priority. The scheduling platform automatically allocates modeling engines for processing based on the urgency, complexity, and resource consumption of the tasks.

[0008] As a preferred technical solution of the present invention, the hardware resource setup in step one specifically includes: S1: Hardware resource preparation: motherboard, CPU, memory, hard disk, GPU, I / O devices, used to provide necessary computing resources; S2: Hardware assembly: based on the general x86 server architecture, the I / O devices in S1 include a 10 Gigabit network card, providing network services for the portable real-scene 3D modeling computing platform, and the hard disk in S1 uses a high-speed solid-state drive as shared storage for the portable real-scene 3D modeling computing platform, integrating computing power, network, and storage into one device; S3: Appearance design: the chassis is made of alloy material, the overall appearance of the chassis is cuboid, the front of the chassis is equipped with a dust filter, the back is reserved with power, network cable, USB, HDMI, and VGA interfaces, and two alloy handles are installed on the top of the chassis to make the physical machine easy to carry.

[0009] As a preferred technical solution of the present invention, the Firecracker virtualization in step two specifically includes: S1, KVM initialization: Enable KVM, the host machine provides hardware-level virtualization support, allowing each micro-virtual machine to run independently; S2, Firecracker binary startup: Start the Firecracker binary file and initialize the micro-virtual machine management process. This process is responsible for managing the lifecycle of the virtual machine, including creating, starting, pausing, and destroying the micro-virtual machine; S3, Configure the virtual machine: Modify the configuration file to define the resource parameters of the virtual machine, including the number of CPU cores, memory size, and I / O settings; S4, Device emulation selection: By minimizing the emulation of hardware devices, overhead is reduced and startup speed is improved. Supported devices are limited to basic functions, such as virtio storage devices and virtio network devices; S5, virtio driver initialization: Use the virtio driver to configure block storage devices and network interfaces for the virtual machine to ensure that the I / O operations of the micro-virtual machine are completed through an efficient virtualization driver; S6, Memory initialization: When starting the micro-virtual machine, allocate the specified physical memory space for it through Linux memory mapping technology and ensure that the memory of each micro-virtual machine does not interfere with each other; S7, Seccomp BPF security mechanisms: Employs the Seccomp BPF system call filter to limit the scope of system calls made by the micro-virtual machine to the host system, intercepting illegal or unauthorized system calls at runtime to enhance security; S8, cgroup resource limits: Utilizes Linux's cgroup mechanism to limit the CPU and memory resources of each micro-virtual machine, preventing it from consuming resources beyond expectations. Cgroups can also control I / O and network bandwidth usage, ensuring fair resource allocation; S9, RESTful API management: Uses a RESTful API interface to manage the virtual machine lifecycle. The API supports creating, starting, pausing, and destroying virtual machines, and provides resource monitoring and snapshot management functions.

[0010] As a preferred technical solution of the present invention, the virtual machine performance optimization based on IO_uring in step three specifically includes: S1. Initializing the IO_uring structure: Initializing an IO_uring instance, allocating a commit queue and a completion queue, the buffer is shared, and the user space and kernel implement asynchronous I / O transfer through this structure. The depth of the commit queue and the completion queue is defined to ensure appropriate memory allocation and concurrent processing capabilities; S2. Operation of the commit queue: Obtaining a new queue entry from the commit queue to describe the I / O operation to be committed. The operation type includes read / write and file operations. The type of I / O operation is set as needed, and relevant parameters such as file descriptors and data buffers are defined for each operation; S3. Submitting I / O requests: After filling all I / O operations into the commit queue, submitting them to the kernel at once, improving efficiency by reducing system calls for each request; S4. Processing the completion queue: Checking the I / O operations processed by the kernel through the completion queue. The user process is blocked and waits for the operation, or periodically checks the completion queue for new completion events, reads the operation result from the completion queue, checks whether it is successful, and if it is successful, executes the subsequent operation in S5; otherwise, it continues to wait for I / O to complete. S5, resource cleanup: After all I / O operations are completed, resources are released and the io_uring instance and its associated queues and buffers are destroyed.

[0011] As a preferred technical solution of the present invention, the unified management of the cloud management platform in step four specifically includes: S1, Virtual Machine Resource Scheduling: The cloud management platform obtains an authentication token by sending a POST request to the Firecracker server. This token's permissions are consistent with the management user permissions in the Firecracker service, ensuring that the platform has the authority to create, destroy, start, and pause virtual machines; S2, Resource Pooling: The CPU, memory, GPU, and hard disk hardware resources of the computing platform are divided into different resource pools. Each resource pool is allocated to different departments according to business needs. The resource pool further allocates the allocated resources to virtual machines for computing tasks. The resources are hierarchically divided into: First Resource: the overall hardware resources allocated to the resource pool; Second Resource: the hardware resources allocated to a specific virtual machine within the First Resource; S3, Resource Monitoring: The platform obtains real-time data from the Firecracker service to monitor the usage of resource pools and virtual machine resources, i.e., the usage of the First Resource and Second Resource. The cloud management platform can adjust the First Resource and Second Resource as needed to maximize resource utilization.

[0012] As a preferred technical solution of the present invention, the scheduling of the real-scene 3D modeling engine in step five is specifically as follows: S1, Information input: Input the real-scene 3D production project and corresponding engineering information, set the processing priority of the real-scene 3D production project and engineering, and submit the tasks to be processed for each engineering; S2, Task calculation: Determine whether the task needs to be calculated by the resource engine. If it needs to be calculated by the resource engine, then execute step S53; if it does not need to be calculated by the resource engine, then execute step S4; S3, Task processing: Automatically process the task according to the priority, generate the task completion result, and execute step S5; S4, Anomaly handling: Remind manual personnel to handle task nodes or anomalies; S5, Task judgment: Determine whether the task completion result is qualified. If the task completion result is qualified, then continue to execute the next stage of the task; if the task completion result is unqualified, then execute step S4.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. The computing platform has a built-in real-scene 3D modeling engine scheduling platform, which automatically schedules modeling projects according to task priority. The scheduling platform automatically allocates modeling engines for processing based on the urgency, complexity and resource consumption of the tasks, reducing the need for manual intervention. This not only optimizes the utilization of computing resources, but also greatly improves the processing speed of real-scene 3D modeling projects. 2. The hardware and software all-in-one machine adopts an independent appearance design and has dustproof and shockproof functions to ensure stable operation of the equipment in complex outdoor environments. The overall structure is lightweight and portable, allowing users to easily carry and deploy it in different environments for 3D modeling work. Attached Figure Description

[0014] Figure 1 This is a perspective view of the hardware and software integrated machine of the present invention; Figure 2 This is a flowchart of the process of the present invention; Figure 3 A flowchart illustrating the hardware resource setup for this invention; Figure 4 This is a flowchart of the Firecracker virtualization process of the present invention; Figure 5 This is a flowchart illustrating the virtual machine performance optimization based on IO_uring according to the present invention; Figure 6 This is a flowchart illustrating the unified management of the virtual machine platform according to the present invention; Figure 7 This is a flowchart of the real-scene 3D modeling engine scheduling process of the present invention; Figure 8 This is a flowchart of the scheduling process for the real-scene 3D modeling engine of this invention; Figure 9 This is a side view of the hardware and software integrated machine of the present invention; Figure 10 This is a cross-sectional view of the hardware and software integrated machine of the present invention.

[0015] In the diagram: 1. All-in-one hardware and software PC; 101. Top panel; 102. Front panel; 103. Side panel; 104. Rear panel; 2. Shock-absorbing corner protector; 21. Shock-absorbing pad; 22. Anti-slip strip; 23. Groove; 3. Handle; 4. Mesh cover; 5. Heat sink; 6. Power button; 7. Fan module; 8. Graphics card module; 9. Motherboard module; 10. Power supply module; 11. Chassis; 12. Graphics card expansion board. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1-10 This invention provides a portable real-scene 3D modeling computing platform, including an all-in-one hardware and software machine 1. The all-in-one hardware and software machine 1 comprises a chassis 11, a power supply module 10, a motherboard module 9, a graphics card module 8, and a fan module 7. The power supply module 10 is located at the bottom of the chassis 11, with a total of four power supplies in a redundant design. The motherboard module 9 is located above the power supply module 10 and integrates hardware such as a CPU, memory, PCIe slots, and I / O interfaces. The graphics card module 8 is located above the motherboard module 9 and is connected to the PCIe slot of the lower motherboard module 9 and the graphics card expansion board 12 of the upper graphics card module 8 via a PCIe splitter card. Eight graphics cards are fixedly installed on the graphics card expansion board 12. The fan module 7 is fixed inside the front panel of the chassis 11 and has two layers of fans forming two independent air ducts for cooling the motherboard module 9 and the graphics card module 8, respectively.

[0018] The four corners of the top and four corners of the bottom of the all-in-one hardware and software machine 1 are fixedly installed with shock-absorbing corner protection mechanisms 2. The shock-absorbing corner protection mechanism 2 includes shock-absorbing pads 21 and anti-slip strips 22. The top of the shock-absorbing pads 21 is fixedly connected to the bottom of the anti-slip strips 22. The surface of the shock-absorbing pads 21 is provided with several grooves 23. The all-in-one hardware and software machine 1 includes a top panel 101, a front panel 102, a rear panel 104 and a side panel 103. The bottom of the top panel 101 is fixedly connected to the top of the front panel 102 and the rear panel 104 respectively. The side panels 103 are fixedly installed on both sides between the front panel 102 and the rear panel 104. The power button 6 is fixedly installed on the top of the front panel 102.

[0019] A mesh cover 4 is fixedly installed on one side of the hardware and software all-in-one machine 1. Several heat dissipation slots 5 are opened in the middle of the top of the hardware and software all-in-one machine 1. The opening of the heat dissipation slots 5 and the mesh cover 4 facilitates the contact between the low temperature gas in the external environment and the hardware and software all-in-one machine 1, so as to transfer heat and dissipate heat to cool down the hardware and software all-in-one machine 1. Handles 3 are fixedly installed on both sides of the top of the hardware and software all-in-one machine 1 for easy carrying when working outdoors.

[0020] The workflow of a portable real-scene 3D modeling computing platform of the present invention includes the following steps: Step 1: Hardware Resource Setup: Integrate multi-core CPUs, large-capacity memory, multiple GPUs, and storage into a unified hardware resource pool to achieve efficient sharing and dynamic adjustment of computing resources; Step 2: Firecracker Virtualization; The underlying system uses resource virtualization technology to virtualize hardware resources into multiple virtual machine instances. Each virtual machine runs modeling tasks independently. Firecracker virtualization simplifies the virtualization architecture, reduces the resource overhead of traditional virtualization, and combines isolation and high efficiency. It also provides performance advantages by optimizing the startup speed of virtual machines. Step 3: Virtual machine performance optimization based on IO_uring: The IO_uring asynchronous I / O mechanism significantly reduces the overhead of system calls and context switching, making it particularly suitable for high-concurrency and I / O-intensive tasks and improving the performance of I / O operations; Step 4: Unified Management by Cloud Management Platform: The computing power platform has a built-in cloud management platform. Resource pooling technology centrally manages the computing resources of the computing power platform. Through a unified interface, users can call different virtual machine resources through the cloud management platform to complete the allocation and scheduling of modeling tasks. Pooling management makes resource use more efficient and facilitates real-time monitoring to ensure the stability of task execution. Step 5: Real-scene 3D modeling engine scheduling: The computing platform has a built-in real-scene 3D modeling engine scheduling platform, which automatically schedules modeling projects according to task priority. The scheduling platform automatically allocates modeling engines for processing based on the urgency, complexity, and resource consumption of the tasks.

[0021] Step 1 involves the following hardware resource setup: S1: Hardware Resource Preparation: Motherboard, CPU, memory, hard drive, GPU, and I / O devices to provide necessary computing resources; S2: Hardware Assembly: Based on a general x86 server architecture, the I / O devices in S1 include a 10 Gigabit Ethernet card to provide network services for the portable 3D modeling computing platform. The hard drive in S1 uses a high-speed solid-state drive as shared storage for the portable 3D modeling computing platform, integrating computing power, network, and storage into a single device; S3: Appearance Design: The chassis is constructed from alloy materials and has an overall rectangular shape. A dust filter is installed on the front of the chassis, while the back has reserved interfaces for power, network cable, USB, HDMI, and VGA. Two alloy handles are installed on the top of the chassis for easy portability.

[0022] Step two of the Firecracker virtualization process specifically includes: S1, KVM initialization: Enabling KVM, the host machine provides hardware-level virtualization support, allowing each micro-virtual machine to run independently; S2, Firecracker binary startup: Starting the Firecracker binary file and initializing the micro-virtual machine management process, which is responsible for managing the lifecycle of the virtual machines, including creating, starting, pausing, and destroying them; S3, configuring the virtual machine: Modifying the configuration file to define the virtual machine's resource parameters, including the number of CPU cores, memory size, and I / O settings; S4, device emulation selection: Minimizing the emulation of hardware devices reduces overhead and improves startup speed. Supported devices are limited to basic functions, such as virtio storage devices and virtio network devices; S5, virtio driver initialization: Using the virtio driver to configure block storage devices and network interfaces for the virtual machine, ensuring that the micro-virtual machine's I / O operations are completed through an efficient virtualization driver; S6, memory initialization: When starting the micro-virtual machine, allocating the specified physical memory space for it using Linux's memory mapping technology, and ensuring that the memory of each micro-virtual machine does not interfere with each other; S7, Seccomp BPF security mechanism: Employing Seccomp... BPF system call filter restricts the scope of system calls made by the microvirtual machine to the host system, intercepting illegal or unauthorized system calls at runtime to enhance security; S8, cgroup resource limitation: utilizes Linux's cgroup mechanism to limit the CPU and memory resources of each microvirtual machine, preventing the microvirtual machine from consuming resources beyond expectations. cgroup can also control I / O and network bandwidth usage to ensure fair resource allocation; S9, RESTful API management: uses a RESTful API interface to manage the lifecycle of virtual machines. The API supports creating, starting, pausing, and destroying virtual machines, and provides resource monitoring and snapshot management functions.

[0023] Step 3, the virtual machine performance optimization based on IO_uring, specifically involves: S1. Initializing the IO_uring structure: Initializing an IO_uring instance, allocating a commit queue and a completion queue. The buffer is shared, and user space and the kernel use this structure to implement asynchronous I / O transfer. The depth of the commit queue and completion queue is defined to ensure appropriate memory allocation and concurrent processing capabilities. S2. Commit queue operations: Obtaining a new queue entry from the commit queue to describe the I / O operation to be committed. Operation types include read / write and file operations. The type of I / O operation is set as needed, and relevant parameters such as file descriptors and data buffers are defined for each operation. S3. Submitting I / O requests: After filling all I / O operations into the commit queue, submitting them to the kernel all at once, improving efficiency by reducing system calls for each request. S4. Processing the completion queue: Checking the kernel's completed I / O operations through the completion queue. The user process is blocked and waits for the operation, or periodically checks the completion queue for new completion events, reads the operation result from the completion queue, checks whether it is successful, and if it is successful, executes the subsequent operation in S5; otherwise, it continues to wait for I / O to complete. S5, resource cleanup: After all I / O operations are completed, resources are released and the io_uring instance and its associated queues and buffers are destroyed.

[0024] Step four, the unified management of the virtual machine platform, specifically involves: S1, Virtual Machine Resource Scheduling: The platform obtains an authentication token by sending a POST request to the Firecracker server. This token's permissions are consistent with the management user permissions in the Firecracker service, ensuring that the platform has the authority to create, destroy, start, and pause virtual machines; S2, Resource Pooling: The computing platform's CPU, memory, GPU, and hard disk hardware resources are divided into different resource pools. Each resource pool is allocated to different departments according to business needs. The resource pool further allocates the allocated resources to virtual machines for computing tasks. These are hierarchically divided into: First Resource: the overall hardware resources allocated to the resource pool; Second Resource: the hardware resources allocated to a specific virtual machine within the First Resource; S3, Resource Monitoring: The platform obtains real-time data from the Firecracker service to monitor the usage of resource pools and virtual machine resources, i.e., the usage of the First and Second Resources. The platform can adjust the First and Second Resources as needed on the cloud management platform to maximize resource utilization.

[0025] Step 5, the scheduling of the real-scene 3D modeling engine, is as follows: S1, Information Input: Input the real-scene 3D production project and corresponding engineering information, set the processing priority of the real-scene 3D production project and engineering, and submit the tasks to be processed for each engineering; S2, Task Calculation: Determine whether the task requires calculation by the resource engine. If it does, proceed to step S53; otherwise, proceed to step S4; S3, Task Processing: Automatically process tasks according to priority, generate task completion results, and proceed to step S5; S4, Exception Handling: Remind manual personnel to handle task nodes or exceptions; S5, Task Judgment: Determine whether the task completion result is qualified. If the task completion result is qualified, continue to the next stage of the task; if the task completion result is unqualified, proceed to step S4.

[0026] In this invention, the computing platform adopts an independent exterior design with dustproof and shockproof functions, ensuring stable operation of the device in complex outdoor environments. Its lightweight and portable structure allows users to easily carry and deploy it for 3D modeling tasks in various environments. Secondly, this invention integrates a multi-core CPU, large-capacity memory, and multiple GPU cards to form a unified hardware resource pool. The underlying system uses resource virtualization technology to virtualize hardware resources into multiple virtual machine instances. Each virtual machine independently runs modeling tasks, simplifying the virtualization architecture, reducing the resource overhead of traditional virtualization, and combining isolation and high efficiency. Furthermore, performance advantages are provided through optimization of virtual machine startup speed. Finally, the use of the IO_uring asynchronous I / O mechanism significantly reduces the overhead of system calls and context switching. This technology is particularly suitable for high-concurrency and I / O-intensive tasks, improving I / O operation performance. Then, resource pooling technology centrally manages the computing resources of the computing platform. Through a unified interface, users can access different virtual machine resources via the cloud management platform to allocate and schedule modeling tasks. Pooling management makes resource utilization more efficient and facilitates real-time monitoring, ensuring task execution stability. Finally, the computing platform has a built-in real-scene 3D modeling engine scheduling platform that automatically schedules modeling projects according to task priority. The scheduling platform automatically allocates modeling engines for processing based on the urgency, complexity, and resource usage of the tasks, reducing the need for manual intervention. This not only optimizes the utilization of computing resources but also greatly improves the processing speed of real-scene 3D modeling projects. Hardware parameters are shown in the table below:

[0027] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A portable real-scene 3D modeling computing platform, comprising an integrated hardware and software machine (1), characterized in that, The all-in-one hardware and software machine (1) includes a chassis (11), a power supply module (10), a motherboard module (9), a graphics card module (8), and a fan module (7). The power supply module (10) of the all-in-one hardware and software machine (1) is located at the bottom of the chassis (11), with a total of 4 power supplies, adopting a redundant design. The motherboard module (9) of the all-in-one hardware and software machine (1) is located above the power supply module (10). The motherboard module (9) integrates the CPU, memory, PCIe slots, and IO interface hardware. The graphics card module (8) is located above the motherboard module (9). The graphics card module (8) is connected to the PCIe slot of the motherboard module (9) below and the graphics card expansion board (12) of the graphics card module (8) above via a PCIe split card. Eight graphics cards are fixedly installed on the graphics card expansion board (12). The fan module (7) of the hardware and software all-in-one machine is fixed inside the front panel of the chassis (11). The fan module (7) is equipped with two layers of fans, forming two independent air ducts to dissipate heat from the motherboard module (9) and the graphics card module (8) respectively. The hardware and software all-in-one machine (1) has four corners at the top and four corners at the bottom fixedly installed with shock-absorbing corner protection mechanisms (2). The shock-absorbing corner protection mechanism (2) includes a shock-absorbing pad (21) and an anti-slip strip (22). The top of the shock-absorbing pad (21) is fixedly connected to the bottom of the anti-slip strip (22). The surface of the shock-absorbing pad (21) is provided with several grooves (23). The hardware and software all-in-one machine (1) includes a top panel (101), a front panel (102), a rear panel (104), and a side panel (103). The bottom of the top panel (101) is fixedly connected to the top of the front panel (102) and the rear panel (104) respectively. The side panels (103) are fixedly installed on both sides between the front panel (102) and the rear panel (104). The top of the front panel (102) is fixedly installed with a power button (6). A mesh cover (4) is fixedly installed on one side of the hardware and software integrated machine (1), and several heat dissipation slots (5) are opened in the middle of the top of the hardware and software integrated machine (1). Handles (3) are fixedly installed on both sides of the top of the hardware and software integrated machine (1). The workflow of the portable real-scene 3D modeling computing platform includes the following steps: Step 1: Hardware Resource Setup: Integrate multi-core CPUs, large-capacity memory, multiple GPUs, and storage into a unified hardware resource pool to achieve efficient sharing and dynamic adjustment of computing resources; Step 2: Firecracker Virtualization; The underlying system uses resource virtualization technology to virtualize hardware resources into multiple virtual machine instances. Each virtual machine runs modeling tasks independently. Firecracker virtualization simplifies the virtualization architecture, reduces the resource overhead of traditional virtualization, and combines isolation and high efficiency. It also provides performance advantages by optimizing the startup speed of virtual machines. Step 3: Virtual machine performance optimization based on IO_uring: The IO_uring asynchronous I / O mechanism significantly reduces the overhead of system calls and context switching, making it suitable for high-concurrency and I / O-intensive tasks and improving the performance of I / O operations; Step 4: Unified Management by Cloud Management Platform: The computing power platform has a built-in cloud management platform. Resource pooling technology centrally manages the computing resources of the computing power platform. Through a unified interface, users can call different virtual machine resources through the cloud management platform to complete the allocation and scheduling of modeling tasks. Pooling management makes resource use more efficient and facilitates real-time monitoring to ensure the stability of task execution. Step 5: Real-scene 3D modeling engine scheduling: The computing platform has a built-in real-scene 3D modeling engine scheduling platform, which automatically schedules modeling projects according to task priority. The scheduling platform automatically allocates modeling engines for processing based on the urgency, complexity, and resource consumption of the tasks. The hardware resource setup in step one is specifically as follows: S1: Hardware resource preparation: motherboard, CPU, memory, hard disk, GPU, I / O devices, used to provide computing resources; S2: Hardware assembly: based on the general x86 server architecture, the I / O devices in S1 include a 10 Gigabit network card, providing network services for the portable real-scene 3D modeling computing platform, and the hard disk in S1 uses a high-speed solid-state drive as shared storage for the portable real-scene 3D modeling computing platform, integrating computing power, network, and storage into one device; S3: Appearance design: the chassis is made of alloy material, with an overall rectangular shape. A dust filter is installed on the front of the chassis, and power, network cable, USB, HDMI, and VGA interfaces are reserved on the back. Two alloy handles are installed on the top of the chassis to make the physical machine easy to carry. Step two of the Firecracker virtualization process specifically includes: S1. KVM Initialization: Enable KVM, providing hardware-level virtualization support to allow each micro-virtual machine to run independently; S2. Firecracker Binary Boot: Start the Firecracker binary file and initialize the micro-virtual machine management process, which is responsible for managing the lifecycle of the virtual machines, including creating, starting, pausing, and destroying them; S3. Configure Virtual Machines: Modify the configuration file to define the virtual machine's resource parameters, including the number of CPU cores, memory size, and I / O settings; S4. Device Emulation Selection: Minimize the emulation of hardware devices to reduce overhead and improve boot speed. Supported devices are limited to basic functions, including virtio storage devices and virtio network devices; S5. Virtio Driver Initialization: Configure block storage devices and network interfaces for the virtual machine using the virtio driver to ensure that the micro-virtual machine's I / O operations are completed through an efficient virtualization driver; S6. Memory Initialization: When starting the micro-virtual machine, allocate the specified physical memory space for it using Linux's memory mapping technology, ensuring that the memory of each micro-virtual machine does not interfere with each other; S7. Seccomp BPF Security Mechanism: Employ Seccomp... BPF system call filter restricts the scope of system calls made by the microvirtual machine to the host system, intercepting illegal or unauthorized system calls at runtime to enhance security; S8, cgroup resource limitation: utilizes Linux's cgroup mechanism to limit the CPU and memory resources of each microvirtual machine, preventing the microvirtual machine from consuming resources beyond expectations. cgroup can also control I / O and network bandwidth usage to ensure fair resource allocation; S9, RESTful API management: uses a RESTful API interface to manage the lifecycle of virtual machines. The API supports creating, starting, pausing, and destroying virtual machines, and provides resource monitoring and snapshot management functions.

2. The portable real-scene 3D modeling computing platform according to claim 1, characterized in that: Step three, the virtual machine performance optimization based on IO_uring, specifically involves: S1. Initializing the IO_uring structure: Initializing an IO_uring instance, allocating a commit queue and a completion queue. The buffer is shared, and user space and the kernel use this structure to implement asynchronous I / O transfer. The depth of the commit queue and the completion queue are defined to ensure appropriate memory allocation and concurrent processing capabilities. S2. Operations on the commit queue: Obtaining a new queue entry from the commit queue to describe the I / O operation to be committed. Operation types include read / write and file operations. The type of I / O operation is set as needed, and relevant parameters are defined for each operation, including file descriptors and data buffers. S3. Submitting I / O requests: After filling all I / O operations into the commit queue, submitting them to the kernel all at once. This improves efficiency by reducing system calls for each request. S4. Completion Queue Processing: Check the I / O operations completed by the kernel through the completion queue. The user process is blocked and waits, or the completion queue is checked periodically for new completion events. The operation result is read from the completion queue and checked for success. If successful, the subsequent operations in S5 are executed; otherwise, the process continues to wait for I / O completion. S5. Resource Cleanup: After all I / O operations are completed, resources are released and the io_uring instance and its associated queues and buffers are destroyed.

3. The portable real-scene 3D modeling computing platform according to claim 1, characterized in that: In step four, the unified management by the cloud management platform specifically refers to S1, virtual machine resource scheduling: The cloud management platform obtains an authentication token by sending a POST request to the Firecracker server. The permissions of this token are consistent with the permissions of the management user in the Firecracker service, ensuring that the platform has the authority to create, destroy, start, and pause virtual machines. S2. Resource Pooling: The computing platform's CPU, memory, GPU, and hard disk hardware resources are divided into different resource pools. Each resource pool is allocated to different departments according to business needs. The resource pool further allocates the allocated resources to virtual machines for computing tasks. It is hierarchically divided into: First Resource: the overall hardware resources allocated to the resource pool; Second Resource: the hardware resources allocated to a specific virtual machine within the First Resource. S3. Resource Monitoring: The platform obtains real-time data from the Firecracker service to monitor the usage of resource pools and virtual machine resources, i.e., the usage of the First Resource and the Second Resource. The First Resource and the Second Resource can be adjusted as needed on the cloud management platform to maximize resource utilization.

4. The portable real-scene 3D modeling computing platform according to claim 1, characterized in that: The scheduling of the real-scene 3D modeling engine in step five is specifically as follows: S1, Information entry: Enter the information of the real-scene 3D production project and the corresponding engineering project, set the processing priority of the real-scene 3D production project and the engineering project, and submit the tasks to be processed for each engineering project. S2. Task Calculation: Determine if the task requires calculation by the resource engine. If it does, proceed to step S3; otherwise, proceed to step S4. S3. Task Processing: Automatically process tasks according to priority, generate task completion results, and proceed to step S5. S4. Exception Handling: Remind manual intervention for task nodes or exceptions. S5. Task Judgment: Determine if the task completion result is satisfactory. If the result is satisfactory, continue to the next stage of the task; otherwise, proceed to step S4.

Images