Skip to content

add resource configuration recommendations in table ainode deployment #1141

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
leto-b wants to merge 1 commit into
base: main
Choose a base branch
from
Open

add resource configuration recommendations in table ainode deployment #1141

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/.vuepress/sidebar/V2.0.x/en-Table.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ export const enSidebar = {
children: [
{ text: 'Deployment form', link: 'Deployment-form_apache' },
{ text: 'Obtain IoTDB', link: 'IoTDB-Package_apache' },
{ text: 'Database Resources', link: 'Database-Resources' },
{ text: 'Database Resources', link: 'Database-Resources_apache' },
{ text: 'Environment Requirements', link: 'Environment-Requirements' },
],
},
Expand Down
2 changes: 1 addition & 1 deletion src/.vuepress/sidebar_timecho/V2.0.x/en-Table.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ export const enSidebar = {
children: [
{ text: 'Deployment form', link: 'Deployment-form_timecho' },
{ text: 'Obtain TimechoDB', link: 'IoTDB-Package_timecho' },
{ text: 'Database Resources', link: 'Database-Resources' },
{ text: 'Database Resources', link: 'Database-Resources_timecho' },
{ text: 'Environment Requirements', link: 'Environment-Requirements' },
],
},
Expand Down
25 changes: 24 additions & 1 deletion ...de/Master/Table/Deployment-and-Maintenance/AINode_Deployment_Upgrade_timecho.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,7 +87,30 @@ sha512sum timechodb-2.0.6.1-ainode-bin.zip
### 2.3 Environment Requirements

- Recommended operating environment: Linux, macOS.
- TimechoDB Version: >= V2.0.8-beta.
- TimechoDB Version: >= V2.0.8.

#### 2.3.1 Resource Configuration Recommendations

> Note: The resource configuration recommendations in this section apply only to **model inference tasks**. Guidelines for model training tasks will be provided in subsequent releases.

The following are baseline resource configurations for model inference running on a single NVIDIA RTX 4090 (24 GB VRAM). For model inference on AINode, overall throughput can be improved by horizontally scaling the number of GPUs. It is generally recommended to deploy servers with 1, 2, 4 or 8 GPUs.

Specifications of inference tasks used in benchmark tests:
- **Univariate inference**: Historical sequence length: 2880, prediction length: 720
- **Covariate inference**: Historical sequence length: 2880, prediction length: 720, with 20 known covariates

| Number of GPUs (NVIDIA 4090, 24 GB VRAM) | Recommended CPU Cores | Recommended Memory (GB) | Supported QPS for Univariate Inference | Supported QPS for Covariate Inference |
|------------------------------------------|-----------------------|-------------------------|-----------------------------------------|---------------------------------------|
| 1 GPU | 16 cores | 24 GB | 100 | 10 |
| 2 GPUs | 32 cores | 48 GB | 200 | 20 |
| 4 GPUs | 64 cores | 96 GB | 400 | 40 |
| 8 GPUs | 128 cores | 192 GB | 800 | 80 |

**Notes**:
- The CPU and memory configurations above follow this general rule: allocate 16 CPU cores per GPU, and set system memory equal to GPU VRAM at a ratio of 1:1.
- The throughput figures are benchmark references. Actual performance may vary depending on model type, data complexity and deployment environment.
- The throughput of univariate and covariate inference shall be evaluated separately as required, and the two values cannot be summed directly.


## 3. Installation, Deployment, and Usage

Expand Down
204 changes: 3 additions & 201 deletions src/UserGuide/Master/Table/Deployment-and-Maintenance/Database-Resources.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,6 @@
---
redirectTo: Database-Resources_apache.html
---
<!--

Licensed to the Apache Software Foundation (ASF) under one
Expand All @@ -18,204 +21,3 @@
under the License.

-->
# Database Resources
## 1. CPU
<table style="text-align: center;">
<tbody>
<tr>
<th rowspan="2">Number of timeseries (frequency&lt;=1HZ)</th>
<th rowspan="2">CPU</th>
<th colspan="3">Number of nodes</th>
</tr>
<tr>
<th>standalone</th>
<th>Dual-Active</th>
<th>Distributed</th>
</tr>
<tr>
<td>Within 100000</td>
<td>2-4 cores</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 300000</td>
<td>4-8 cores</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 500000</td>
<td>8-16 cores</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 1000000</td>
<td>16-32 cores</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 2000000</td>
<td>32-48 cores</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 10000000</td>
<td>48core</td>
<td>1</td>
<td>2</td>
<td>Please contact Timecho Business for consultation</td>
</tr>
<tr>
<td>Over 10000000</td>
<td colspan="4">Please contact Timecho Business for consultation</td>
</tr>
</tbody>
</table>

> Supported CPU models: Kunpeng, Phytium, Sunway, Hygon, Zhaoxin, Loongson

## 2. Memory
<table style="text-align: center;">
<tbody>
<tr>
<th rowspan="2">Number of timeseries (frequency&lt;=1HZ)</th>
<th rowspan="2">Memory</th>
<th colspan="3">Number of nodes</th>
</tr>
<tr>
<th>standalone</th>
<th>Dual-Active</th>
<th>Distributed</th>
</tr>
<tr>
<td>Within 100000</td>
<td>2-4G</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 300000</td>
<td>6-12G</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 500000</td>
<td>12-24G</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 1000000</td>
<td>24-48G</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 2000000</td>
<td>48-96G</td>
<td>1</td>
<td>2</td>
<td>3</td>
</tr>
<tr>
<td>Within 10000000</td>
<td>128G</td>
<td>1</td>
<td>2</td>
<td>Please contact Timecho Business for consultation</td>
</tr>
<tr>
<td>Over 10000000</td>
<td colspan="4">Please contact Timecho Business for consultation</td>
</tr>
</tbody>
</table>

> Flexible memory configuration options are provided. Users can adjust them in the datanode-env file. For details and configuration guidelines, please refer to [datanode-env](../Reference/System-Config-Manual.md#_3-2-datanode-env-sh-bat)


## 3. Storage (Disk)
### 3.1 Storage space
Calculation Formula:

```Plain
Storage Space = Number of Measurement Points * Sampling Frequency (Hz) * Size of Each Data Point (Bytes, see the table below) * Storage Duration * Replication Factor / Compression Ratio
```

Data Point Size Calculation Table:

<table style="text-align: center;">
<tbody>
<tr>
<th>Data Type</th>
<th>Timestamp (Bytes)</th>
<th> Value (Bytes)</th>
<th> Total Data Point Size (Bytes)
</th>
</tr>
<tr>
<td>Boolean</td>
<td>8</td>
<td>1</td>
<td>9</td>
</tr>
<tr>
<td> INT32 / FLOAT (Single Precision)</td>
<td>8</td>
<td>4</td>
<td>12</td>
</tr>
<tr>
<td>INT64 / DOUBLE (Double Precision)</td>
<td>8</td>
<td>8</td>
<td>16</td>
</tr>
<tr>
<td>TEXT (String)</td>
<td>8</td>
<td>Average = a</td>
<td>8+a</td>
</tr>
</tbody>
</table>
Example:

- Scenario: 1,000 devices, 100 measurement points per device, i.e. 100,000 sequences in total. Data type is INT32. Sampling frequency is 1Hz (once per second). Storage duration is 1 year. Replication factor is 3.
- Full Calculation:
```Plain
1,000 devices * 100 measurement points * 12 bytes per data point * 86,400 seconds per day * 365 days per year * 3 replicas / 10 compression ratio = 11 TB
```
- Simplified Calculation:
```Plain
1,000 * 100 * 12 * 86,400 * 365 * 3 / 10 = 11 TB
```
### 3.2 Storage Configuration

- For systems with > 10 million measurement points or high query loads, SSD is recommended.

## 4. Network (NIC)
When the write throughput does not exceed 10 million points per second, a gigabit network card is required. When the write throughput exceeds 10 million points per second, a 10-gigabit network card is required.

| **Write** **Throughput** **(Data Points/Second)** | **NIC** **Speed** |
| ------------------------------------------------- | -------------------- |
| < 10 million | 1 Gbps (Gigabit) |
| ≥ 10 million | 10 Gbps (10 Gigabit) |

## 5. Additional Notes

- IoTDB supports second-level cluster scaling . Data migration is not required when adding new nodes, so there is no need to worry about limited cluster capacity based on current data estimates. You can add new nodes to the cluster when scaling is needed in the future.
Loading
Loading