Scenarios

CSS can be used to build search boxes for websites and apps to improve user experience. You can also build a log analytics platform with it, facilitating data-driven O&M and business operations. CSS vector search can help you quickly build intelligent applications, such as AI-enabled image search, recommendations, and semantic search.

Site Search

CSS can be used to search for website content by keyword as well as search for and recommend products on e-commerce sites.

• Real-time search: When site content is updated, you can find the updated content in your search within minutes, or even just seconds.
• Categorized statistics: You can apply search filters to sort products by category.
• Custom highlight style: You can define how the search results are highlighted.

Figure 1 Site search

All-Scenario Log Analytics

Analyze the logs of Elastic Load Balance (ELB), servers, containers, and applications. In CSS, the Kafka message buffer queue is used to balance loads in peak and off-peak hours. Logstash is used for data extract, transform and load (ETL). Elasticsearch retrieves and analyzes data. The analysis results are visualized by Kibana and presented to you.

• High cost-effectiveness: CSS uses the Kunpeng computing power, separates cold and hot storage, and decouples computing and storage resources, achieving high performance and reducing costs by over 30%.
• Ease of use: Perform queries in a GUI editor. Easily create reports using drag-and-drop components.
• Powerful processing capability: CSS can import hundreds of terabytes of data per day, and can process petabytes of data.

Figure 2 All-scenario log analytics

Database Query Acceleration

CSS can be used to accelerate database queries. E-commerce and logistics companies have to respond to a huge number of concurrent order queries within a short period of time. Relational databases, although having good transaction atomicity, are weak in transaction processing, and can rely on CSS to enhance OLTP and OLAP capabilities.

• High performance: Retrieve data from hundreds of millions of records within milliseconds. Text, time, numeric, and spatial data types are supported.
• High scalability: CSS can be scaled to have over 200 data nodes and over 1000 columns.
• Zero service interruption: The rolling restart and dual-copy mechanisms can avoid service interruption in case of specifications change or configuration update.

Figure 3 Database query acceleration

Vector Search

When you search for unstructured data, such as images, videos, and corpuses, the nearest neighbors or approximate nearest neighbors are searched based on feature vectors. This has the following advantages:

• Efficiency and reliability: The Huawei Cloud vector search engine provides ultimate search performance and distributed disaster recovery capabilities.
• Abundant indexes: Multiple indexing algorithms and similarity measurement methods are available and can meet diverse needs.
• Easy learning: CSS is fully compatible with the open-source Elasticsearch ecosystem.

Figure 4 Vector search

OpenSearch Observability Solution

CSS OpenSearch clusters support the full-stack observability solution. A unified platform integrates logs, metrics, and traces to enable end-to-end monitoring, from infrastructure to application services.

• Unified full-stack data analysis: Seamlessly integrates application logs, performance metrics, and distributed traces via the standard OpenTelemetry protocol, breaking data silos and enabling unified, context-aware observability across services.
• Intelligent root cause locating: The built-in Trace Analytics displays the complete call chain for each microservice in an intuitive manner. Used with a PPL query engine, it can quickly identify performance bottlenecks, significantly accelerating troubleshooting.
• Real-time association analysis: Supports real-time association query and analysis of logs, metrics, and traces. A single platform handles the end-to-end troubleshooting process, from detection to root cause locating, significantly enhancing O&M efficiency.

Figure 5 Observability solution overview