I. Background
The end of many mobile products will launch their annual bill, every year the mass media attention. This year the honor involved in the development CreditEase wealth APP annual bill, it will take you to explore the architecture and technology research and development logic behind the bill should believe the wealth of the Year, hoping to inspire some ideas on everyone.
Second, the overall front-end architecture and implementation process
CreditEase wealth Annual Technical bill stack front-end architecture used comprises:
- Front page with H5 production;
- Percentage progress of data loading, and a number of plug-in technology used swiper CSS3 dynamic efficiency;
- Posters used to generate a synthetic canvas picture, the poster background and 2D codes combined.
- In order to complete the track MGM, M1 is embedded in the two-dimensional code information.
Third, data sources and data processing
The Annual Statement data involving customer dimension, a dimension sales, customer labels, which include customer dimension active participation, article, video browsing and other data. This part of the integration of data from the data subject of the data in the table. The following is a schematic diagram of the data in the table:
- ODS: Source layer, extracted from the storage system over the data traffic, data traffic through the system in the original extract, washed, loaded in the transport layer. This layer data close to the original data, the original data is not identical, data loading time was de-emphasis, noise removal, the table name, the field name and a series of operation specifications.
- DW: data warehouse layer, which is the subject of data warehouse, data ODS layer in accordance with the theme data model, is to develop the process for all levels of corporate decision-making, providing all strategic collection types supported by the data, that contains all topics a collection of generic.
- DM: data mart layer, is a business application as a starting point field generated relatively wide table, provide follow-up service for queries, OLAP analysis, data distribution, etc., the main layer and layer data details from the mild summary generated layer data is calculated.
On the table schema data, we have established a "customer-centric" label system. The set of labeling system in accordance with the demographic attributes, value index, geographical index, index and other broad categories of psychological data were stratified management, processing methods mainly from lightweight label DW and DM summary of the data layer or derivatives processing, as well as parts of the model the resulting product forecasting and other labels. This labeling system supports 360-degree portrait and analysis of key customer contact points, providing key contacts based on experience optimization and cross-channel customers to optimize full flow.
The billing data mainly from the source system operational data, user management, etc. These data are stored in a database cluster, and are structured to access the stored data sets, and falls at the timing corresponding to the task relating to real-time data, or area. Billing data processed by data relating to their respective distal end API to access data through the interface.
The bill needs in the "sales evaluation of real-time news push" and "dissemination of messages sent bills" are supported by intelligent operating system, the operating system is a set of activities to create, execute, manage, feedback, iterative as one of the automation platform, through user attributes, labels, plans, operations and other data filtering customer base, achieve the goal of precise touch up and enhance key indicators and operational efficiency.
The following is an intelligent operating system to create a flow chart of the Business Plan:
- Evaluate the message in real-time push sales: This function depends on the real-time internet wormhole database data fall, and configuration data in the intelligent operation system, and ultimately by the message center to push a message Aurora products terminal.
- Bill spread SMS: Screening of eligible customer base in accordance with the business rules configuration SMS templates and other content in the smart operating system, and then send text messages to call notify customers via SMS platform.
Fourth, the technical background
User data from the data itself should believe wealth platform, include: basic information, browse information, participate in activities and many other data, how to ensure that data accurately and efficiently transmitted to the front-end back-end development is necessary for security. Asset platform uses spring + jersery + oracle + redis + jetCache technical architecture, in order to enhance the user experience, faster response times, the data storage The project uses caching of non-relational databases and flexible combination of traditional relational database, the better provide data support.
When docked Annual Statement demand, we also consider the focus on the interface response time. Annual Statement user data including user activity data and operational data of two tables, which is a heavyweight operation data table, in order to reduce IO operation of the database, using the two methods to reduce the IO time:
- The label data set provided to minimize the probability of access to the asset data table;
- Stream java8 using a new feature, the complex SQL code into logic processing.
Stream is not a collection of elements, it is not a data structure of the data is not saved, it is more like an advanced version of the Iterator related algorithms and calculations.
Further Stream technology also provides a parallel, when the internal data set of interest is not sequential, may be used in parallel to speed up the task Stream dismantling process. For example, do the statistics, we need to be aggregated sub-product, or other operations.
If a complex code logic is implemented directly in SQL code will be very long, the efficiency is not high. Logic code is to use a parallel flow Stream, subtotal relevant data according to the type, and according to the service needs of this scenario will be classified into a subclass under another category.
Instead of using a parallel stream Stream SQL logic may accelerate the efficiency, reduce the response time. Interested students Stream if you want additional features, refer to the technical documentation. Stream application logic can make the code more clear and improve speed.
V. Summary
This project is a collaborative effort by multiple teams, this year the bill needs to do a technical sort, because of the time more haste, less detailed content, hoping to bring some development ideas for everyone, we hope users can truly feel to our intentions.
Source: Yixin wealth management team
Author: rice Zhihua, Sun Li Qiang, Li Li, Zhao Quan-chao