Shulou(Shulou.com)06/03 Report--

Preface

The company changes from page games to mobile games, and the company's data analysis needs to be designed for mobile games, so the original data analysis framework for page games is not very appropriate. on the one hand, it is different in some business logic between mobile games and page games, on the other hand, there is a change in the level of data, as well as the connection between channels and regional services, as well as cross-query of the channel BI system of mobile games. So that the original set of 4399 games for page games from the framework appears to be a bit of a struggle. What is analyzed here is the adjustment work done in view of big data's analysis in the process of page swimming to mobile games, as well as the analysis cases between them.

Page game cooperation scene

In the page game stage, if a page game is ready to be pushed to a certain platform, then you need to dock with the platform and take an examination on the platform to achieve a certain score before the platform can provide resources for online expansion. So if you want to go to another platform, you need to go through the same process as another platform. At present, most of the cooperation between the page game company and the platform is like this. In fact, games, platforms and district services all correspond one to one. One platform can open multiple district services, and one district service only belongs to one platform (different from mobile games).

The process of data acquisition and analysis of page games

With the launch of the game, the company needs to check the number analysis of the game, such as retention loss, payment, active users, crash, and so on. This is a subsystem of many BI systems (data analysis system: for user behavior analysis, product quality monitoring, activity participation statistics, etc.). So the process is roughly like the following figure:

Analysis:

1. The data between platforms is isolated.

two。 When using ETL to deal with project team data, maintain the original game platform-sub-server library, if for some special tables, you can pull summary tables, such as login tables, payment tables.

3. Write a statistical script, put it into the Linux scheduled task, and put the statistical results into the result database, which is divided into databases according to the platform.

4. Analyze the statistical results from the result database according to the BI system data, and show the charts of the relevant index data.

5. In this process, there will be a small amount of real-time data analysis requirements, which need to directly connect the number of project groups in the BI system or pull to the result database through a short-time scheduled task.

The process of data acquisition and analysis of mobile games

In the mobile game stage, if a game needs to be launched through a certain channel, you need to cooperate with channel vendors. IOS and Android are the platforms, and game data from different channels under the same platform can be written into the same regional server, that is to say, there is a many-to-many relationship between channel and regional service. Then compared with the page game situation is different, the data in a server may come from multiple channels, and the data from one channel is divided into multiple regional servers. The general analysis process is as follows:

Analysis:

1. There is a direct many-to-many relationship between channel and service.

When 2.etl pulls and cleans, it still divides the warehouse according to the channel-area service.

3. Write a statistical script, save the statistical results to the result database, and there is only one result library for a game (including all channels, all district servers).

4. Because there is only one result bank, you need to consider sub-tables to avoid too large a single table and slow to query. The analysis is as follows:

Assuming that there are 50 channels, 100 districts and no more than 5 dimensions of a single table, and the increment of [10W-15W] / day is estimated according to experience, then the table is divided according to the quarter, and the amount of data in each sub-table is between 900W and 1350W. In addition, for some statistical tables with relatively active records, monthly breakdown or more detailed tables may be required. All in all, it is to reduce the size of a single table as much as possible, but this will bring a problem, that is, how to query and paging in the BI system, but this problem has corresponding components to help us complete, such as: Ali's mycat, Dangdang's Sharding-JDBC, I plan to use Sharding-JDBC, because mycat for query, paging I feel a little heavy. If you need to use similar features, please choose according to your own needs.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.