How Yahoo Cost Optimizes Their in Memory Workloads with Aws Dat321

Title

AWS re:Invent 2022 - How Yahoo cost optimizes their in-memory workloads with AWS (DAT321)

Summary

  • Yahoo moved their ad platform to AWS and optimized costs using Amazon ElastiCache.
  • Maulik Shah from Yahoo and Itai Meos from AWS presented the session.
  • Yahoo's ad platform generates 320 billion events per day, requiring efficient data processing.
  • Yahoo initially considered DynamoDB and HBase but settled on ElastiCache due to its performance and cost-effectiveness.
  • Data tiering in ElastiCache allowed Yahoo to reduce cluster size and costs by 50%.
  • Yahoo's architecture includes AWS Kinesis, S3, EMR, and Glue, with data querying via Athena or direct S3 access.
  • Yahoo learned to manage connections efficiently and implement exponential backoff with retries.
  • Yahoo plans to further optimize costs and migrate on-prem workloads to AWS.
  • AWS introduced data tiering, allowing storage expansion by moving data from memory to SSDs, saving up to 60% in costs.
  • Data tiering is transparent to applications and requires no code changes.
  • AWS also discussed MemoryDB for Redis, a durable in-memory database service.
  • New ElastiCache features include an improved management console, Memcached 1.6.12, encryption in transit, IPv6 support, Redis log delivery, native JSON support, AWS PrivateLink, Redis 7, and IAM authentication.
  • ElastiCache is available in all AWS regions and will be included in new regions.

Insights

  • Yahoo's use case demonstrates the scalability and cost-effectiveness of AWS services for handling large-scale data workloads.
  • Data tiering is a significant innovation that allows for cost savings by intelligently managing data storage between memory and SSDs.
  • The session highlighted the importance of monitoring and managing connections to maintain performance and reliability.
  • AWS's commitment to customer feedback is evident in the development of data tiering and other new features.
  • MemoryDB for Redis offers durability and high performance, expanding the options for customers needing in-memory databases.
  • The introduction of new features and services like MemoryDB for Redis and data tiering indicates AWS's focus on continuous innovation and addressing diverse customer needs.
  • The global availability of ElastiCache ensures that customers can rely on consistent performance and features regardless of their geographic location.