Solution Overview

Application Scenarios

The explosive growth of internet information presents enterprises and individuals with the challenge of managing and efficiently retrieving massive datasets. While traditional search engines suffice for basic needs, they often fall short when confronted with diverse data types and personalized requirements. This is where DeepSeek emerges. As a Chinese developed AI large language model, DeepSeek has risen to prominence in the AI field, leveraging its high performance, low cost, and multi-modal capabilities to demonstrate significant application potential across various sectors.

This solution enables the rapid deployment of a DeepSeek inference system on Huawei Cloud Flexus X instances (Elastic Cloud Server, ECS). DeepSeek-R1, a high-performance AI inference model specializing in mathematical, code, and natural language reasoning tasks, is deployed via Ollama on the cloud server using its distilled version. This quickly creates your private AI assistant, ideal for the following applications:

• Natural Language Processing (NLP): Understands and generates natural language text, suitable for tasks such as dialogue, translation, and summarization.
• Text Generation: Produces coherent and logically sound text, applicable to content creation and story writing.
• Question Answering System: Answers user queries, ideal for customer service and knowledge base searches.
• Sentiment Analysis: Analyzes the emotional tone of text, useful for market research and public opinion monitoring.
• Text Classification: Categorizes text, applicable to spam filtering and news categorization.
• Information Extraction: Extracts key information from text, suitable for data mining and knowledge graph construction.

Architecture

This solution helps you quickly set up the DeepSeek-R1 distilled models on the Huawei Cloud Flexus X instance (Elastic Cloud Server (ECS)).

Figure 1 Architecture

This solution will:

• Create a FlexusX instance(Elastic Cloud Server (ECS)) to set up the DeepSeek-R1 distilled models.

• Create one Elastic IP (EIP) for internal and external communication.
• Create a security group and configure security group rules to protect Huawei Cloud cloud servers.

Advantages

• High performance

  DeepSeek significantly enhances inference capabilities through reinforcement learning technology, supports multi-step logical inference, and can gradually decompose complex problems and solve them.

• Cost-effectiveness

  Provide high-cost-performance cloud servers, users can customize different specifications of cloud servers according to actual needs.

• Easy deployment

  In just a few clicks, you can easily deploy and complete the quick provisioning of cloud servers, public IP, and other resources, as well as the setup of the DeepSeek-R1 distilled model.

Constraints

• Before deploying this solution, ensure that you have created a Huawei ID with access to the target region and enabled Huawei Cloud services.
• If you select the yearly/monthly billing mode, ensure that your account has sufficient balance. If you do not have sufficient balance, you can go to the Billing Center to manually pay for the order.