Using Financial and Business Thinking as a DBA: Why is TiDB More Expensive than OceanBase? We Can Save More Money

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

Original topic: 用业财的思维做dba,tidb为啥比oceanbase费钱?不我们可以更省钱

| username: tidb狂热爱好者

This is a rough draft. Could you all take a look and see if it can be written, and if so, how to optimize it?

Using financial and business thinking to manage DBA tasks, why is TiDB more expensive than OceanBase? Actually, we can save more money.

OceanBase often criticizes TiDB for being expensive and not suitable for business.

Unity after long separation, separation after long unity.

Today, let’s discuss the cost analysis of big data business.

Recently, the company needs to implement a batch of archiving tasks.

These data cannot be lost. They need to be searchable, quickly accessible, widely used, and low-cost. This problem has stumped me.

Everyone knows that cloud machine costs are very high. If you spend 1 million, you can buy machines for a lifetime offline, but on the cloud, it might only last a year.

Recently, there has been a focus on energy saving and emission reduction. We previously used AWS RDS, which costs tens of millions a year.

Cost savings can be achieved in several ways:

  1. Storage

  2. Number of machines

  3. Single machine price

  4. Machine type

  5. Machine architecture

  6. Cross-network exchange costs

  7. Exploiting capitalist resources

  8. PITR

  9. Storage: Storage on AWS is priced. EBS includes disk costs and IO costs.

  10. Number of machines: We can’t just deploy distributed business because TiDB is a distributed database. We also need to consider unit price issues. Moreover, distributed systems have network exchange, which affects performance. If we place everything on the same machine, network loss is saved.

  11. Single machine price: Is it better to buy a batch of 4-core machines or one 64-core machine? Which has stronger computing power?

  12. Machine architecture: AWS machines are divided into ARM and x86 machines. ARM machines are about 70% the price of x86 machines. TiDB supports ARM machines, which is a significant cost-saving.

  13. Cross-network exchange costs.

  14. How to save the most on S3 costs. How to use TiDB on the cloud to exploit resources.

S3 costs are very cheap.

Moreover, AWS’s network performance is very good. We can use S3 costs to do local computing tasks.
Regular archiving:

BACKUP TABLE `test`.`sbtest01` TO 'local:///mnt/backup/sbtest01/';

To calculate data:

RESTORE DATABASE `test` FROM 'local:///mnt/backup/2020/04/';

I even wrote a program to save records of each backup.
Because I increased the backup concurrency, the execution is very fast. Large tables take minutes, and bandwidth is fully utilized at 50G.

  1. PITR: Using PITR to ensure multiple copies is definitely a bad idea.

Because OceanBase has its unique 2-copy guarantee for high data availability. TiDB originally needed 3 copies to ensure no data loss, but now we use S3 for storage.

Combining these features is like a magic attack. The cost is definitely lower than OB.

| username: Jack-li | Original post link

Fight magic with magic :+1:

| username: TiDBer_JUi6UvZm | Original post link

Start cloud computing, beginning with Elon Musk.

| username: TiDBer_JUi6UvZm | Original post link

I don’t know how many banks or stock exchanges will move to the cloud.

| username: shigp_TIDBER | Original post link

In the short term, using the cloud saves money, but in the long term, having your own servers saves money…

| username: 随缘天空 | Original post link

Open source and reduce expenditure

| username: zhaokede | Original post link

Respect to the expert.

| username: YuchongXU | Original post link

None of them save money.

| username: mono | Original post link

If the business data volume is small, moving to the cloud might save money overall. However, if the business data volume is large, the storage costs alone can be quite high, and when you add in computing and network costs, the cloud can become very expensive. Hiring professional engineers and building your own infrastructure is the way to go.

| username: xiaoqiao | Original post link

The OB is very heavy.

| username: DBAER | Original post link

Finance is all about private clouds; regulators don’t allow otherwise.

| username: 呢莫不爱吃鱼 | Original post link

:+1: :+1: :+1:

| username: forever | Original post link

Download it.

| username: 柴米油酱 | Original post link

Buying machines offline doesn’t seem to save money either. Construction cycle, maintenance, hardware repairs, business interruption impacts, etc. The higher the usage on the cloud, the more you can negotiate discounts with cloud providers.

| username: 数据库真NB | Original post link

I don’t understand, what are they doing? Are they calculating the costs, advantages, and disadvantages of public cloud and physical machines?