[EWTT2022] Strategi Implementasi Database dalam Microservice Architecture.pdf
- 1. EWTT2022: Strategi Implementasi Database dalam Microservice
- 2. This document is owned by Equnix Business Solutions Pte LTD. This document contains confidential information, which is protected by law. No part of this publication could be photocopied, reproduced or translated into another language without permitted or the express written consent of Equnix Business Solutions PTE LTD. Data and information regarding the proposal and its offer is for limited use and are not disclosed. The information contained in this document is subject to change at any time without prior notice. All rights reserved, © Copyright 2022 - Equnix Business Solutions Copyright Notice
- 4. GOAL: 1. Breakdown the complexity of monolith business process 2. Easier to develop, easier to deploy 3. Smaller responsibilities for smaller and focused team. 4. Reduce Database Hit. especially for update. Microservice Architecture Recommendation: 1. More cache towards faster inquiry. 2. Avoid real-time data cache, use database internal cache.
- 5. Rest API, RPC, or Binary? Despite of its advantage, Microservice has its tradeoff: more point to communicate, more latency and more things to maintain. However, Microservices advantage are still greater and better. Inter Module communication can use some options: 1. Message Queues: Kafka, RabbitMQ, etc 2. Direct communication, uses: Rest, RPC, JSON invocation, or just plain binary. 3. Thread Pool trigger. 4. On demand thread or routine forks 5. Database access, worst case scenario. Recommendation: 1. REST is great for interoperability between system and entity. REST is the international language for outside world. 2. RPC was invented long time ago, the ideas still fresh. It is always a great idea to be able to utilize other system resources remotely and in flexible manner. 3. Binary representation or serialization especially for sensitive data is high considered.
- 6. Queue, Pool or Trigger? Queue is very good in: 1. Decouple of chained of process 2. Scale out the load per module. 3. Offloading CPU load into flatter hit rate, distribute CPU usage across longer time. Queue is NOT good for: 1. Committing transaction 2. Synchronous things: a. Process/threads b. Communication, Query, etc Pool There are some pools: Memory Pool, Connection Pool, Thread Pool, etc.. Pools provides bunch of ready to use resource, can be combined using queues, OR only use Conditional Variable which is faster. Pool management is tricky, it is better implemented without any locks.
- 8. Example of Microservice Architecture
- 10. Strategy for Comprehensive Redevelopment into Microservice Architecture 1. Application breakdown 2. Business Process/Logic refactoring 3. Data restructuring 4. Transactional level data normalization 5. Analytical level data denormalization
- 11. 1. Each existing application is breaking down into simple modules, such as per feature, or by one independent business process (algorithm). 2. Build the interdependence of each of these modules into a unified business process that going through from accept the input of user (SPOE/WUI) up to the ORM (Object Relational Module). 3. Develop middlewares to couple the communication between modules. (Kafka, Redis, etc) 4. Develop an API Gateway for User Interface, as the single point of entry to ensure better access security. 5. Rebuilding RDBMS data structure to support Microservices. It is required to be assessed more comprehensive and involves all applications Application Breakdown Strategy 6. Develop a data synchronization tools for Old Data Structure and New one. 7. Create a staging database to support local data manipulation per modules before reaching stable transaction state. 8. Develop Data Warehouse System 9. Develop Data Lake for storing huge data from multiple data sources