ML/AI Service Mesh Made Easy With API Management (DeveloperWeek Global 2020)

ML/AI service mesh made easy with API management: discover how to apply service mesh architecture to machine learning production systems and leverage embeddings for linking entity profiles.

Key takeaways
  • Service mesh architecture is becoming a pattern for machine learning (ML) production systems, allowing for a clean separation of concerns and embracing microservices.
  • ML engineers can focus on designing models, while ops engineers can handle the lifecycle management of the models.
  • Embeddings are a way to represent characteristics of each image, allowing for links between entity profiles across projects, teams, and organizations.
  • API management can solve problems such as contract enforcement and making APIs available for internal and external consumption.
  • Istio service mesh is a pluggable component that can come in to manage the mesh, providing features like circuit breaker, retries, and timeouts.
  • Shared governance and common policies can be applied across multiple proxies in a service mesh.
  • Managed planes can collect telemetry and provide insights into the performance of the system.
  • Embeddings can be used in other applications, such as parallel processing or recommendation systems.
  • Apigee remote service and Envoy proxy can be used to enable easy consumption of APIs.
  • Service mesh can be used to simplify connection between services and provide a high-performing, pluggable proxy.
  • Monolithic models can be broken down into smaller models and made shareable across multiple services.
  • API management can help to modernize legacy applications and provide a nice HTTP REST layer.
  • Embeddings can be used in BigQuery for data warehouse and prediction capabilities can be shared across multiple services.
  • A service mesh can be used to solve problems such as scheduling of containers or instances, and to provide a high-throughput system.
  • Apigee adapter can help enforce policies and make APIs available for internal and external consumption.
  • Shared embeddings can be used for linking entity profiles across projects, teams, and organizations.
  • Embeddings can be used to provide a high-performing, pluggable proxy.
  • Istio service mesh can provide features like circuit breaker, retries, and timeouts.
  • Apigee remote service can provide a high-performing, pluggable proxy.
  • Embeddings can be used to provide a high-performing, pluggable proxy.
  • Service mesh can be used to provide a high-performing, pluggable proxy and to solve problems such as scheduling of containers or instances.
  • Embeddings can be used to provide a high-performing, pluggable proxy.
  • Traditional authentication methods like OAuth can be used to secure the system.
  • Embeddings can be used to provide a high-performing, pluggable proxy.
  • Service mesh can be used to solve problems such as scheduling of containers or instances, and to provide a high-throughput system.
  • API management can help to modernize legacy applications and provide a nice HTTP REST layer.
  • Embeddings can be used in BigQuery for data warehouse and prediction capabilities can be shared across multiple services.
  • A service mesh can be used to solve problems such as scheduling of containers or instances, and to provide a high-throughput system.
  • Embeddings can be used for linking entity profiles across projects, teams, and organizations.
  • Embeddings can be used to provide a high-performing, pluggable proxy.
  • Embeddings can be used to provide a high-performing, pluggable proxy.