In this talk delivered by Ilake Chang at RubyConf MY 2017, the main focus is on a significant memory reduction achievement in a production server using Postgres schemas and the Apartment gem. The video outlines the circumstances surrounding a severe memory issue faced by the speaker's organization, which was related to the use of AWS EC2 for server management and resulted in excessive memory usage. Notably, they achieved a remarkable 500x reduction in memory usage through strategic restructuring of their database schema and improving their architecture. The session covers the following key points:

• Background and Context: The speaker introduces their company and context for the talk, explaining how their e-commerce platform was affected by memory consumption issues.

• Initial Findings: Early development testing indicated memory usage dramatically reduced from 400MB to 800KB when simulating 10,000 schemas, leading to interest in a scalable solution.

• Root Cause Analysis: The speaker explains how they traced memory usage to a large database in production, which significantly differed from the staging environment. Issues were linked to the operational needs of a multitenancy structure using Postgres schemas.

• Key Tools and Strategies: Various tools such as Medusa and New Relic were used to monitor memory consumption, and strategies focused on optimizing garbage collection and request handling. Techniques to test memory structure duplicated real production requests in staging environments.

• Development Approaches: The speaker rationalized the choice of the Apartment gem for multi-tenancy and indicated various adjustments in the source code of Rails and PostgreSQL settings were necessary to handle the influx of records without causing memory bloat.

• Lessons Learned: The speaker emphasizes the necessity of returning to foundational analysis rather than simply relying on existing tools when addressing serious technical issues. The problems of migration in multi-schema structures and the need for potential refactoring for future scalability were also highlighted.

• Conclusions: Chang stresses the importance of recognizing architectural choices and their implications on scalability and performance, advising against overly complex schemas that could impede future growth. The experiences shared serve as a cautionary tale for effectively managing data architecture in production environments.

Ultimately, the session combines theoretical insights with practical applications to highlight significant memory management techniques in a multi-tenancy environment using Rails and Postgres, encouraging developers to think critically about their database structures.