Shenma novels: Using opensearch to create a high-performance search service

Shenma novels ---

Use opensearch to create a high-performance search service

 

[Background]

Shenma novels were the first users to use opensearch and grew up with opensearch. Currently, Shenma novels search PV for 0.2 billion million users every day. The product form is divided into three parts: the entire network SC, the vertical search, and the app.

Opensearch is stable and reliable, with an average latency of 40 ms per day and an average QPS of 2500.

[Usage process]

The Shenma fiction team has a senior search background. When the novel product was just getting started, it was faced with the problem of building its own engine or choosing the basic direction of opensearch. The main reason why opensearch was selected at that time:

1. Trust the opensearch team. The past experiences of the opensearch team show the outstanding qualities of determination and reliable work;

2. opensearch is easy to use. You can create an application on the web page, Configure index fields, and enter a few documents using the data interface. Then the engine takes effect, common engineers can familiarize themselves with the use of opensearch in just a few days;

3. Engine deployment, upgrade, and expansion are very troublesome for system O & M. Using opensearch can save a lot of trouble and focus more on the business.

4. opensearch provides a perfect solution for text fuzzy search. It can customize multi-dimensional word segmentation, Rich Text relevance feature, low latency, and high QPS to meet our needs.

Facts prove that our selection was correct.

Of course, opensearch is also used. For example, when the index field type can be changed online, the index becomes invalid and there is no search result. In addition, early searches occasionally have no results, which is a bug in opensearch.

It is inevitable to step on the pitfall, just as there must be bugs in the program. What is important is the subsequent processing speed and solution after the pitfall. The performance of opensearch is reassuring. After a problem is posted in the group, the opensearch team quickly answers the question and then actively participates in problem tracking until the problem is solved. Even when an emergency occurs at night, no help is found.

 

[Requirements and suggestions]

Let's talk about the functional problems encountered when using opensearch.

1. opensearch does not support full reconstruction. It is inconvenient to pull out expired data and delete it one by one.

Opensearch answer: currently, all API functions are being planned.

2. opensearch is expected to support application aliases. The user uses the application alias A. A actually points to A1. When switching an application, the user directs a to A2, which is transparent to the user.

Opensearch: the phased release function of the application is being planned.

 

[Summary]

The recently revised front-end page is very useful. The investment in data centers in Hangzhou and Shanghai makes the service more stable and reliable. The upgrade of offline processes and the enhancement of refined sorting functions give users a better user experience.

I wish opensearch better and better, and bring more benefits to everyone.

 

Shenma novels: http://m.sm.cn/novel/

Apply for opensearch Free Public Beta now

Weibo Interaction

 

Shenma novels: Using opensearch to create a high-performance search service