Build a Serverless Order Storage System with Alibaba Cloud Tablestore

Background and Challenges

Over the past decade, rapid internet growth has created complex business scenarios that demand high‑performance data systems. While MySQL remains the most widely used relational database due to its strong query and transaction capabilities, it struggles in Serverless environments because of limited elasticity, high operational complexity, and costly over‑provisioning.

Desired Serverless Database Characteristics

Full elasticity : automatic scaling up and down according to workload, enabling pay‑per‑use billing and smoother user experience.

Pay‑per‑use pricing : charges only for actual compute and storage consumption.

Zero operations : no manual capacity planning, software upgrades, or kernel tuning required.

Tablestore Overview

Alibaba Cloud’s Tablestore is a multi‑model, Serverless table storage service launched in 2009. It supports wide‑column, timeline (message), and time‑series models, offering elastic storage from GB to PB, ten‑thousands of TPS, and integrated search and analytics capabilities. It powers large‑scale e‑commerce platforms such as Century Lianhua Group, which uses Function Compute, API Gateway, and Tablestore for major sales events.

Step‑by‑Step Experience

1. Preparation

Create an Alibaba Cloud account and obtain an AccessKey (AK) for API authentication.

Download and start the Tablestore CLI tool.

Configure the CLI and enable the service:

config --id accessKeyID --key accessKeySecret

enable_service

2. Create an Instance

An instance in Tablestore is analogous to a MySQL database. Create one with:

create_instance -d "order storage" -n serverless-db -r cn-hangzhou

The instance automatically handles capacity, and you only pay for actual read/write and storage usage.

3. Create a Wide‑Column Table

Wide‑column tables are schema‑free; only the primary key needs definition. Example order table creation: create_table -t order -p "id" The table starts with a single data partition that will split automatically as data grows.

4. Data Import

Generate 1 million sample orders and import them: import -i orderDataFile -l 1000000 Import speeds reach tens of thousands of rows per second, scaling further as partitions expand.

5. Querying

Retrieve a single order by primary key:

get --pk '["0000005be2b43dd134eae18ebe079774"]'

+----------------------------------+-------+--------+---------+-------------+---------------+--------+--------+----------+--------+---------+-------+-------+--------+------------+

The command only supports row‑key lookups.

6. Indexing and Analytics

Tablestore provides multi‑dimensional indexes similar to Elasticsearch. Create a search index on the order table:

create_search_index -t order -n order_index
{
    "FieldSchemas": [
        {"FieldName": "id", "FieldType": "KEYWORD", "Index": true, "EnableSortAndAgg": true, "Store": true},
        {"FieldName": "pName", "FieldType": "TEXT", "Index": true, "EnableSortAndAgg": false, "Store": true},
        {"FieldName": "totalPrice", "FieldType": "DOUBLE", "Index": true, "EnableSortAndAgg": true, "Store": true}
        // ... other fields
    ]
}

Tablestore also supports SQL queries compatible with MySQL syntax, automatically leveraging indexes for millisecond‑level latency even on billions of rows.

select * from `order` where sName = "售周五" and pBrand = "小米" and pName like "红米%" limit 3;

select pBrand, count(*) from `order` group by pBrand;

Conclusion

Tablestore delivers a cost‑effective, fully elastic, zero‑maintenance Serverless database. In the demonstrated order‑processing scenario (100 million orders, ~2000 TPS), the monthly cost is under 400 CNY, while providing seamless scaling from GB to PB and supporting rich query and analytics capabilities.