ZMO

Baidao Data Helps ZMO Optimize Cloud Architecture

Customer background

ZMO (Perceptual Step) is a brand under ZMO (Shenzhen) Digital Technology Co., Ltd. It is an innovative company focused on AI-generated content (AIGC) technology, committed to providing consumers with a unique digital content creation experience. Through advanced AI technology, everyone can easily create high-quality images and videos without being restricted by the high usage threshold of creation tools. The ZMO team consists of elites from top global technology companies and institutions, including Google, Apple, GrowingIO, ETH, and KUL. The company’s vision is to create an environment that can both inspire personal creativity and promote technological progress, enabling everyone’s creativity to be unleashed and realized through AI technology.

Business requirements and challenges

Customers are very sensitive to network latency because their users are located all over the world. Therefore, they hope to find a cloud platform with better network performance to serve the global market. Secondly, they want to use more Serverless services instead of relying solely on virtual machines for deployment. This can significantly reduce maintenance costs. Finally, they need more technical support services. Since most of the company’s employees are R & D personnel and their operational and maintenance capabilities are relatively weak, they hope that some companies can help relieve part of the operational and maintenance pressure.

Due to meeting the policy of Google’s Startup program, the company successfully applied for the startup support (Startup) program provided by Google in 2022. However, during the process of using Google Cloud services, the company realized that it urgently needed a support partner with profound technical strength and rich experience in terms of service operation and solution integration. For this reason, the company chose Baidao Data, hoping to leverage its outstanding after-sales service and operation and maintenance capabilities in the field of Google Cloud Platform (GCP) infrastructure to seamlessly migrate all the services originally deployed on Alibaba Cloud and Microsoft Azure to the GCP platform. This cooperation not only improves the company’s cloud service efficiency but also lays a solid technical foundation for future business expansion.

Solution

The architecture diagram is as above

Scheme architecture description

1. Client access

• Clients: Users access the system through PCs, mobile devices, or APIs.

2. Security and Load Balancing

• Cloud Armor：Provide security protection against DDoS attacks and other cyber threats.
• Cloud Load Balancing：Responsible for distributing traffic to multiple service instances to ensure the high availability and load balancing of the system; Through anycast IP, ensure a consistent experience for global users.

3. API gateway

• Spring Cloud Gateway：As an API gateway, it is responsible for processing and routing requests from clients to ensure that requests can reach the correct microservice instances.

4. Service cluster

• Task Service：Handle specific task logic, which may involve business processing or data operations.
• User Service：Manage user – related operations, such as user authentication, authorization, and user data management.
• Order Service：Handle operations such as order creation, management, and querying.
• Logging Service：Responsible for recording and managing system logs for monitoring and troubleshooting.

5. Nacos cluster

• Nacos：Used for service discovery and configuration management to ensure the consistency of communication and configuration among microservices.

6. Data storage

• Memory Store (Redis)：As a cache layer, it provides high – speed data access and is suitable for scenarios with frequent reading and writing.
• BigQuery：Used for big data analysis and querying business logs, supporting efficient data processing.
• MongoDB：Used to store unstructured or semi-structured data, suitable for flexible data models.
• Cloud Storage：Used to store a large amount of unstructured data, such as files and objects.
• FileStore：Used for file storage, supporting file upload, download and management.

7. Task Scheduling and Processing

• Job Clusters：Responsible for scheduling and executing various tasks, supporting multiple task types:
  • Stable Diffusion：SD image generation task.
  • Dream Booth：Dream mission.
  • Control Net：Control network tasks.
  • Background Remove：Background removal task.

8. Message Queue

• Queue Services：Used for task queue management to ensure asynchronous processing of tasks and decoupling of system components.

Use the product

– Compute Engine

– Google Kubernetes Engine

– Global Load Balancer

– MemoryStore

– Cloud Storage

– Cloud CDN

– Cloud Armor

– BigQuery

Customer benefits

Compared with other cloud vendors, Google has numerous regions. In terms of the public cloud market, Compute Engine provides an ideal solution regarding throughput, stability, pricing, backup, and security.

Fully managed GKE can automatically handle the creation, configuration, and upgrade of Kubernetes clusters, reducing operational workload. It provides a highly available control plane to ensure the stable operation of the cluster.

Global Load balancer can distribute traffic globally, ensuring that even if an instance in a certain region fails, the traffic can be automatically switched to other healthy instances. Moreover, it regularly conducts health checks on backend instances to ensure that only healthy instances receive traffic, thus improving the reliability of the system.

Cloud CDN reduces user access latency and improves loading speed by caching content globally. Cloud Armor provides powerful DDoS protection capabilities, which can identify and mitigate large-scale distributed denial-of-service attacks to ensure service availability. As a WAF, it can filter malicious traffic and prevent common web attacks such as SQL injection and XSS attacks. This reduces the ineffective consumption of backend resources and improves resource utilization efficiency.

Based on the modern cloud – native system of Google Cloud, it adopts a microservices architecture, containerized deployment, and various storage solutions to ensure the high availability, scalability, and flexibility of the system. Through components such as API gateways, message queues, and task schedulers, the system can efficiently handle complex business logic and data flows.