Hardware requirements
Necessary condition
Currently, private deployments require K8S based deployments, so the hardware requirements are somewhat higher than before.
Hardware requirements vary due to factors such as demand module, request volume and data volume, resulting in inconsistent hardware configuration requirements.
tip
If multiple products are deployed together and the disk space is not superimposed, the maximum value can be taken
Disk space depends on:
- Data Retention Policy (Data Retention Period)
- Log Retention Policy
- Data volume size
- It is not recommended to use network mounted disks such as NAS or NFS unless you can ensure stable network transmission between the server and the network disk
- It is strongly recommended to use SSDs or HDDs with better performance for disks
The disk size resources can be appropriately increased or reduced according to the actual business volume and demand module. Data reference is as follows:
MinIO: One worker generates 60-70MB screen recording files every 2 minutes with a resolution of 1920 * 1080
MySQL: One worker generates approximately 1GB per year. 100 Task per day, 50 records per Task, 0.5KB per record
Hardware performance requirements
- Disk performance: The read speed is not less than 200M/s, and the write speed is not less than 200M/s Inspection method
- Network performance: Intranet transmission speed not less than 512k/s
- CPU instruction set: If the product you purchased includes OCR recognition capability (CPU version), the instruction set needs to include fma, avx, and avx2 instructions
caution
If the product you purchased includes OCR recognition capability (GPU version), please ensure that your server has a GPU graphics card and a driver is installed.
The recommended graphics card model is Nvidia [v100, p100, p40, t4]
Single node deployment
You can refer to the following table to count the hardware resources you need:
| Product | CPU Cores | RAM(GB) | HDD(GB) | GPU(GB) | remarks |
|---|---|---|---|---|---|
| Laiye Automation Commander | 8 | 16 | 500 | - | - |
| Dialogue AI platform | 16 | 32 | 300 | - | - |
| The Hub | 4 | 8 | - | - | - |
| data service | 4 | 8 | 100 | - | - |
caution
Because there are too many modules in the Laiye Intelligent Document Processing, it is not possible to display All in document. You can accessIDP Configuration Calculation ToolTo calculate the hardware configuration
Multinode deployment
caution
For a multi node deployment, three or more nodes are required. You need to create a load balancing (lb) as an access portal and process multiple nodes as back-end servers.
If you cannot provide a load balancing (lb) address, you can provide us with a free IP address, and we will use software to meet your demand.
(However, as far as I know, third-party cloud server providers such as Aliyuns do not support this approach.)
Three nodes
The following is the configuration information for each node
| Product | CPU Cores | RAM(GB) | HDD(GB) | GPU(GB) | remarks |
|---|---|---|---|---|---|
| Laiye Automation Commander | 8 | 16 | 500 | - | - |
| Dialogue AI platform | 16 | 32 | 300 | - | - |
| The Hub | 4 | 8 | - | - | - |
| data service | 4 | 8 | - | - | - |
caution
Because there are too many modules in the Laiye Intelligent Document Processing, it is not possible to display All in document. You can accessIDP Configuration Calculation ToolTo calculate the hardware configuration
More than three nodes
For installations with more than three nodes, the master and node methods will be used to plan the hardware configuration
master:Deploy basic K8S services, such as:apiserver, Kube controller, kube scheduler, and middleware servicesnode:The node that actually runs the service module
It can be roughly divided into the following two situations:
3 < nodes < 6: Master=3, node=number of nodes (master is also a node)
6 =< nodes: Master=3, node=number of nodes - number of masters (master is only a master)
Assuming that the number of nodes provided is 4, the number of master roles is 3, and the number of node roles is 4.