site-logo

JAWS PANKRATION 2024

Introduction of "Minaden", Which Processes 200TB of Japanese Cell Phone Base Station Data Per Day

Lv300

Lv300

2024/8/25 02:40 (JST)

セッション情報

NTT DOCOMO, Japanese largest mobile network operator, has built a cell phone base station data processing infrastructure for 85 million users on AWS called "Minaden".

We will discuss the details of a large-scale architecture that includes hundreds of EC2 instances, numerous Kinesis shards, and other components.

Also discuss streaming processing functionality using EC2 and S3, as well as ETL processing with Athena. Furthermore, strategies for optimizing AWS costs and raising awareness of cost management.

Ryo  Ozawa

Ryo Ozawa

- AWS Jr. Champions(APN) -

- AWS All Certified Engineers(APN) -



セッションカテゴリ
Analysis
Computing
Database
Machine learning
Storage


関連AWSサービス
ECS(EC2)
Kinesis
ElastiCache
S3
Athena

セッション資料

    セッションアーカイブ

    セッションサマリ(by Amazon Bedrock)
      NTT Docomo, a Japanese mobile network operator, built a home base station data processing infrastructure on AWS for 5 million users. The system processes 200 terabytes of data daily from 66,000 base stations across Japan. The architecture consists of three main functions: 1. Receiver: Converts binary data from base stations to a readable format and filters it. 2. Data Tracking: Analyzes network quality by combining base station and network data for each user. 3. Advanced Analysis: Uses over 300 machine learning models for network analysis. Additional features include: - Athena-based ETL functions for non-real-time processing - Cost reduction strategies: - Migration to Graviton2 instances - Use of EC2 Savings Plans - Development of cost monitoring applications - Regular cost review meetings Future plans involve optimizing data processing using S3, SNS, and SQS to maintain simplicity and scalability while reducing costs associated with Kinesis Firehose. The presentation highlights NTT Docomo's efforts to build an efficient, scalable, and cost-effective network infrastructure using AWS services.

    ©JAWS-UG (AWS User Group - Japan). All rights reserved.