Instancias de contenedor
Overview
The system and data disks for instances are typically local SSDs,while a small number are cloud disks. Local data disks offer good performance but have no redundant copies; due to the inherent lifespan of local disks, there is a probability of failure. Please regularly back up important data to File Storage or locally. WebCal does not provide any reliability guarantees or services for these types of storage products.
What is a container instance?
Container instances are containers that use Docker technology for resource partitioning and isolation. They offer advantages such as lower performance overhead and higher efficiency compared to virtual machine instances.
Docker is not supported within container instances. If you need to use Docker, please contact customer service to rent a bare-metal server. (Bare-metal servers are available on a monthly subscription basis.)
Allocation of Instance GPU/CPU/Memory
To rent a GPU on WebCal, select a physical host and then create an instance on that host with N GPU cards.Memory and CPU configurations are allocated in multiples based on the number of GPUs. For example, if the CPU and memory allocation rules displayed on the host are 32GB/GPU and 8 cores/GPU, then when renting 1 GPU, the instance configuration is 1卡GPU; 8核CPU; 32GB内存; when renting 2 GPUs, the instance configuration is 2卡GPU; 16核CPU; 64GB内存; and so on.Therefore, when renting an instance, select the appropriate physical host based on your needs, and then choose the number of GPUs you require.
Directories in the instance
To view disk usage on the instance, run the following command in the terminal: source /root/.bashrc
| Name | Path | Size | Performance | Description |
|---|---|---|---|---|
| System drive | Root directory / and all paths beneath it | |||
| (Excluding the following special paths) | 30GB | Local disk | ||
| Fast | Data will not be lost when the instance is shut down. General system dependencies and Python packages are typically installed on the system drive, which can also store small amounts of data such as code; | |||
| Migration occurs when migrating instances, and changes are saved to the image when saving the image. | ||||
| Data Volume | /root/webcal-tmp | Starting at 50GB | ||
| Scalable | Local Storage | |||
| Fast | Data is not lost when the instance is shut down. Suitable for storing data with high read/write I/O requirements. However, it cannot be saved to the image. | |||
| If you need to migrate data from a data disk, please refer to Migrating Instances (Same Region) | ||||
| WebCal File Storage | /root/webcal-fs | 20 GB free | ||
| Overage Billing | Network Drive | |||
| General | Enables file synchronization and sharing between different instances within the same region. | |||
| Public Data | /root/webcal-pub | Read-only | Network drive | |
| Slow | Directory where the platform’s commonly used public data and public models are stored. Read-only; file writing is not supported |
Additional Notes on the System Drive
In the Linux operating system, almost everything exists as a folder or file, and these folders and files share a common root directory: /. If we install the Linux operating system on a disk A (with no other partitions), the root directory will contain the following default system directories:
$ ls /
bin etc lib mnt proc run tmp var boot dev home root srv sys usr
By default, any file written to any directory is ultimately written to Disk A. The system disk in a WebCal instance refers to Disk A, where the operating system is installed; writing a file to any location under the root directory / is equivalent to writing to the system disk.However, there are exceptions. For example, suppose there is a brand-new hard drive B. If you want to use hard drive B under the Linux system, you can mount this drive to a specific directory (assuming the mounted directory is /root/webcal-tmp).in which case, writing a folder or file to the /root/webcal-tmp directory means writing to hard drive B, rather than to hard drive A.At this point, you should understand the distinction between the WebCal system disk and the data disk. For example, the system software installed on the WebCal instance and the Python dependencies installed using the built-in miniconda are all installed on the system disk.
JupyterLab Working Directory
The working directory for JupyterLab is /root. The data disk, cloud storage, and public data directory are all located under the /root directory, so you can see the directories created by default by the platform immediately after logging into JupyterLab. To view the root directory at /, you can open a terminal and use the command line.