A leading financial service provider in China established a unified scheduling and monitoring operation and maintenance based on the Beluga scheduling system

file

Guide: A leading domestic wealth management service provider was established in 2019. It is the first batch of joint-stock banks approved for establishment, the first approved to open, and the first established bank wealth management subsidiary. Since launching the first RMB wealth management product in China in 2004, it has led the domestic banking wealth management market through continuous innovation of investment models and continuous improvement of investment management capabilities. The company handles 1W of tasks every day, and has many internal systems. It is urgent to need a unified scheduling and monitoring platform to support the needs of different wealth management business scenarios.

Before 2023, the enterprise's use of the open source version of the big data scheduling platform Apache DolphinScheduler can basically meet the business needs, but with the increase in business volume and the growth of individual needs, after careful research, it was decided to migrate all scheduling tasks to The commercial version is on top of the Beluga Scheduling System.

The adoption of the Beluga scheduling system enabled the company to realize unified scheduling and monitoring operation and maintenance, and the scheduling performance was greatly improved. This article will describe in detail the solutions and related application practices provided by the Beluga Dispatch System for this enterprise.

Business needs

The financial industry of a leading financial service provider in China is a data-intensive industry. With the continuous development of this industry and the continuous advancement of technology, the amount of financial data continues to increase, and the data sources are becoming more and more diverse. Data quality needs to be strictly controlled, and efficient data processing and analysis capabilities are required. In addition, the financial market changes rapidly, and risk control requires real-time response, which puts forward higher requirements for its data processing and analysis capabilities. At the same time, protecting customers' privacy and confidential information, and preventing security issues such as cyber attacks and data leakage all pose important challenges to the enterprise.

Specifically, the production and marketing, investment, risk assessment, operation and investment teams of the company all have their own data marts (data warehouses). ETL process, output statistical data to their respective bazaars, and then share their own bazaar data with others.

Since each team has its own scheduling platform, the maintenance cost of the platform is very high, and the operation and maintenance management is also very inconvenient. On the other hand, it is impossible to centrally count and manage the tasks of all teams.

Therefore, the construction of a unified dispatching system is an urgent task in this enterprise. In the process of building the data system, it needs to build a high-performance, high-reliability, and easy-to-expand scheduling system based on the Xinchuang environment to meet the unified scheduling and monitoring needs of various data processing tasks in the data system construction of the enterprise. Realize unified scheduling and monitoring operation and maintenance, support the realization of individual needs in different scenarios, and ensure the basic construction of the data system.

When receiving business scheduling requirements, the company's scheduling cluster at that time consisted of 3 40C*256G machines, consisting of 2 APIs, 3 Masters, and 3 Workers. 10,000+/day.

Specific to the business line, the company has the following requirements for building a scheduling system:

It is necessary to support each department to use a unified dispatching system, and use permissions to isolate the resources of each department to ensure the isolation and security of system operation.
There must be rich monitoring and statistical functions to support the needs of unified data monitoring and operation and maintenance services;
The scheduling system should support multiple trigger types, such as events, custom calendars, dependent scheduling, etc. to trigger jobs, and support users to flexibly configure jobs and combine workflows to meet scheduling requirements in different scenarios;
It needs to be able to combine the company's own needs to support Xinchuang components, scalable task types, data source types, alarm components and other extended requirements;
Support distributed architecture and resource control methods to realize a scheduling system with high reliability and high processing performance;
Support customized requirements, such as docking user and authority data with the company's own user system, docking with the customer's unified login platform, and other data processing platforms and alarm platforms.

One-stop solution based on Beluga Dispatch System

Beluga Scheduling System The domestic Xinchuang scheduling platform is designed to support the task scheduling needs of customers in different scenarios in the construction of the data system, aiming at the stability, reliability, compliance, high performance, and support of Xinchuang for enterprise-level users. The commercialized version developed for environmental and other requirements is positioned as an enterprise-level unified scheduling platform, which formulates a unified development specification and operation and maintenance method for the batch processing tasks of the enterprise, and performs unified management, scheduling and monitoring of the batch tasks of each system.

The Beluga Scheduling System inherits the powerful functions and ease of use of the big data scheduling platform Apache DolphinScheduler, and at the same time adds a number of enterprise-level functions to meet the needs of actual industry users, allowing enterprises to deploy a scheduling system, reduce maintenance costs, and have a powerful scheduling capabilities.

After clarifying the requirements, Beluga Open Source is confident that the Beluga scheduling system can meet all the needs of customers, and proceeds to implement and deploy the solution step by step and in stages. After a series of tests, Beluga Open Source launched the implementation plan, helped customers solve a series of problems before, during and after deployment, and delivered a complete solution for them.

Overall technical solution

The Beluga Scheduling System mainly uses the design idea of a non-central architecture . Services are discovered through the registration center, and services can back up each other, ensuring the reliability of the system. At the same time, the number of services can also be expanded horizontally, which can well support The number of tasks in tens of millions of enterprises can fully meet the needs of large data processing in financial business scenarios.

The main services of the Beluga Scheduling System include UI user interface, interface layer API service, five modules of MasterServer, WorkerServer, and AlertServer , among which the task processing service MasterServer and WorkerServer are typical non-central architectures, which automatically discover the startup and shutdown of services through the registration center. It can effectively support scenarios where the number of tasks increases and decreases, while ensuring the scalability and high performance of task processing.

All services of the system are deployed in a distributed manner , supporting high stability, high performance and high fault tolerance, and ensuring sustainable services. Even if the customer's data centers are scattered all over the country, the Beluga dispatching system can easily deal with different requests from customers. Processing needs of departmental data systems.

file

According to the needs of customers, the solution provided by the Beluga Scheduling System allows each team to have its own task execution cluster. The scheduling system is a separate cluster. You can log in to the cluster of each team without secrets. Each team uses shell tasks for scheduling. ssh remote 'shell command' to perform tasks on the respective clusters.

Specific solutions include (due to confidentiality requirements, only part of the information is disclosed):

Multi-calendar management facilitates the management of multiple market trading days, working days, and natural days

Design a variety of calendars for enterprises, support the separation of scheduling calendars and business data calendars, and support functions such as different time zones and calendar import:

Support multiple calendar management: trading days, working days, natural days, etc.
Support calendar Excel template import
Supports scheduling calendar separate from business calendar
Support multiple time zone settings
Support date cut time setting

Powerful business data date function to better meet the needs of the financial industry

It is designed separately for the business data date in customer scheduling to better meet the needs of the financial industry:

Multiple schedules use the same business data date (for example, on the last trading day, if the card does not turn over, the data pointer will stop on this day, for example, if 9.30 does not run 10.1 and starts running on 10.6, it still runs on 9.30)
Powerful business flop mission
Support front and rear custom flops
Support custom scripts to directly generate business date values
Support timing flop and task end flop
Cooperate with the calendar and support a variety of complex functions for data processing
Before and after n days, months (working days, trading days, natural days)
The month and year of the current card (working day, trading day, natural day)

file

Event-driven meets file, Kafka, and database-level flexible trigger scheduling

In addition to date and time triggers, an event trigger mechanism is added to meet the needs of interaction between multiple systems in the enterprise:

file trigger
- Different file interfaces, automatically triggered according to files that meet special definitions
- Scanning frequency and timeout alarm can be set
database trigger
- Automatically trigger tasks after special data definitions are met
Kafka trigger
- Automatically trigger the workflow when the message conforming to the definition is found

Professional permission setting and Token management meet the needs of enterprise-level permissions

Set up users, roles, and permissions for the enterprise to provide different functional permissions for different departments:

Enterprise User Management
- Independent enterprise user account, which can be connected to existing enterprise users
Project Role Management
- Separate administrator roles and developer roles for different projects
- Combining project roles to control resource permissions
Token management can help enterprises set API call permissions

file

The detailed audit log satisfies the enterprise retroactive user behavior audit function

The audit log records the user's detailed operation object and time:
The audit log records the user's operation behavior in detail
You can set the log retention time
Support Excel export and user operation query

file

High scalability supports component expansion

The system is highly reliable and easy to expand
Meet customized needs, such as opening up the user authority system, connecting with other business systems, expanding new task components, data source components, and alarm components, and supporting the connection to blood relationship analysis functions developed by customers.

file

Support Xinchuang Environment

The Beluga scheduling system is fully localized in terms of Xinchuang’s compliance deployment support: it supports domestic servers, domestic CPUs, memory, motherboards, hard drives, etc., such as Kunpeng servers.

Support domestic operating systems, such as Winning Kirin, Galaxy Kirin, and Red Flag Linux.
Support domestic databases, such as Dameng database, GaussDB, TiDB.
Support the transformation of servers, operating systems, middleware, and databases involved in system deployment to meet the needs of system migration and original task scheduling in the environment of Xinchuang.

Unified monitoring, alarm, operation and maintenance

Achieving unified monitoring and operation and maintenance is one of the most important goals of customers' unified dispatching. Beluga dispatching system can provide unified monitoring alarms, unified monitoring indicators, and provide rich API interfaces to meet user customized monitoring needs; unified monitoring alarm platform, The operation and maintenance platform greatly reduces maintenance costs; multiple monitoring indicators, intelligent statistical indicators based on permissions, to meet the monitoring needs of different roles; unified alarm service, provides rich alarm plug-ins, and supports multiple alarm methods to send at the same time; centralized monitoring workflow , task running status and batch operations, unified management of alarm list, event list, etc.

Customer benefits

Currently, the company has built a unified scheduling platform based on the Beluga scheduling system, which helps the management background realize:

Provide customers with a unified scheduling platform, improve the efficiency of data processing, and support the business of the production and sales, investment, risk assessment, and operation and investment teams of the customer's technology department, and will access more businesses in the future;
Reduce the difficulty of operation and maintenance, lower the threshold of non-technical personnel
Improve delivery efficiency
quick dispatch
Flexible Data Analysis

Customer Profile

A top financial service provider in China has provided investors with a wealth of investment products and created rich investment returns for customers.

This article is supported by Beluga Open Source Technology !