Multitenancy

As explained in Designing Deep Learning Systems: A guide for software engineers

An application that serves multiple tenants while keeping their data, configuration, or usage contexts logically separated.

Multitenancy lets one application serve multiple tenants while keeping their accounts, data, configurations, or usage contexts logically separated. In model-serving systems, tenants share the software while receiving predictions from models trained for their own data. section 6.3.2

A multitenant HPO service must isolate resources so different users or groups remain within their boundaries. This requirement complements fault tolerance, which allows sequential trials to resume after failures, and portability, which decouples execution from infrastructure. section 5.3.1 section 5.3.1

Tenant-specific chatbot models

A chatbot serves schools, retail stores, or other organizations through the same application. Each tenant has a separate account and segregated data, trains an intent-classification model with its own dataset, and receives predictions from that tenant's model. A model cache keeps model graphs and dependent data in memory so one prediction service can serve multiple models; a model file server supplies the model files. section 6.3.2 section 6.3.2

Figure 6.6 A prediction service with model caching for multitenant applications

Use the architecture diagram to inspect how model caching and model-file retrieval extend a prediction service for multiple tenant models. section 6.3.2

How Designing Deep Learning Systems: A guide for software engineers covers it

How deep it goes

Explains resource isolation and a model-cache architecture, stopping short of detailed tenancy boundaries or isolation mechanisms.

Examples

Uses a chatbot serving schools or retailers, tenant-trained intent models, and a cached multi-model prediction service.

What it assumes

Introduces tenants, accounts, data segregation, and tenant-specific models through the chatbot scenario rather than assuming them.

What you can do after

Apply it by isolating tenant activities and routing prediction requests to the corresponding tenant-specific model.

How it connects

Connects multitenancy with hyperparameter optimization, resource isolation, model serving, model caching, and intent classification.

liveBookmanning.com

Read Designing Deep Learning Systems: A guide for software engineers on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Google Anthos in ActionMultitenancy is an architecture in which multiple teams or tenants share a platform while retaining logical separation of their workloads and resources.
• 

  liveBookmanning.com
  Platform Engineering on KubernetesMulti-tenancy is an architecture in which multiple tenants share underlying infrastructure while maintaining controlled separation of resources and access.

Compare how 3 books explain multitenancy →

Related concepts

• database tableuses · A database table is a relational database structure that organizes records into rows and defined columns.
• openid connectuses · OpenID Connect is an identity layer built on OAuth 2.0 that enables clients to authenticate end users and obtain standardized identity claims.
• namespaceproduced by · A namespace is a naming context or identifier space in which names are defined and resolved, distinguishing them from names in other contexts.

This explainer was generated by Manning's knowledge graph from the text of Designing Deep Learning Systems: A guide for software engineers. The section links point to the parts of the book it draws on.