What is the minimal cluster deployment configuration to handle the following business scenarios?

Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: 需要应对以下业务场景最小的集群部署方案是如何配置

| username: TiDBer_E3pRgGAy

[TiDB Usage Environment] Production Environment
[TiDB Version] 6.5.4
[Business Scenario] The business generates four to five million data entries daily, and in extreme cases, it can reach tens of millions of data entries per day; there are relatively high concurrent read and write operations, and sometimes a single query retrieves a large amount of data.
[Question] If I want to deploy a TiDB cluster to handle the above two scenarios, I am not sure what the minimum configuration plan would be reasonable.

| username: dockerfile | Original post link

Cloud or on-premises?

| username: TiDBer_E3pRgGAy | Original post link

Move to the cloud.

| username: dockerfile | Original post link

Are you referring to Alibaba Cloud?

Check out the performance-limited models; they have the lowest overall cost.
I’m currently running several terabyte clusters without any issues.

TiDB, PD, TiCDC, and DM are mixed on a 16C16G performance-limited model.
TiKV is separate on an 8C32G performance-limited model.

| username: TiDBer_E3pRgGAy | Original post link

I am the high-performance Alibaba Cloud.

| username: dockerfile | Original post link

Then you can just play around with it. You can run TiKV separately and mix other components.

| username: Kongdom | Original post link

Is this data volume for a single table or the entire database?

| username: TiDBer_E3pRgGAy | Original post link

Single table

| username: TiDBer_E3pRgGAy | Original post link

I have studied this before. The main issue is that the update concurrency is too high, and once it gets very high, it will cause a significant increase in TiKV’s CPU usage.

| username: zhanggame1 | Original post link

TiKV performs scheduled garbage collection (GC), so the load will be significantly higher during updates when GC is happening.

| username: tidb菜鸟一只 | Original post link

Mixing PD and TiDB on at least one lower-performance machine, configuring at least three high-performance TiKV nodes to handle high concurrent read and write operations, and adding at least one TiFlash node to handle large data queries in a single request.

| username: Kongdom | Original post link

The amount of data in a single table is a bit too much, so you need to get a high read-write server as a TiKV node and try to optimize the table creation to avoid data hotspots.

| username: redgame | Original post link

TiKV standalone, others mixed.

| username: system | Original post link

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.