Instagram Twitter Pocket GitHub Linked In

This is a summary of a presentation I put together on how you might go about transitioning a typical enterprise retailer still running their retail site on monolithic application such as Oracle ATG onto a more modern platform.

Target

  • Large Enterprises that are running archaic monolithic stacks such as ATG
  • Enterprises that want to retain control and access to their own stack

Advantages

  • Migration path for existing Commerce stacks to Cloud native solution
  • Ability to run entire Commerce Platform in the Cloud
  • Modern ways of working: TDD, CI/CD
  • Cloud Agnostic
  • Vendor Agnostic

How

Use an API Gateway to provide access to:

  • Serverless Functions
  • Container based Microservices
  • Proxy to third party services
  • Proxy to existing monolithic applications

Ideally you don’t want to get caught up in trying to host and scale a monolithic application such as ATG in the cloud.

API Gateway

The API Gateway provides access to number of logical services.

API Gateway

Microservices

These Micoservices could be containers or serverless functions, proxies to third party services or proxies to existing monolithic applications.

Microservices

Migration Path

  • First implementation would mostly be proxy to existing monolithic application such as ATG
  • Eventually the endpoints would be replaced with custom microservices and third party services
  • Long term goal would be a custom microservice or third party service for every endpoint
  • Going forwards it mean that changes in functionality are as simple as changing the service behind an endpoint

Migration Path

Products vs Services

If done properly there’s an opportunity to build and sell custom services that can be used by other retailers.

There are a couple of ways this can be done and it’s debatable what the best solution would be:

  • Sell subscriptions to services: overhead of client tracking/billing, requires 24/7 support, potentially provides access to useful data (privacy concern)
  • Sell product licenses: harder to enforce licensed use, potential to modify and customise service for client specific requirements

It’s likely a combination of the two could be used where some solutions are sold as a service that customers subscribe to, while for others a license is sold and the microservice is deployed into the customers environment.

For some customers the ability to deploy a product into their own environment rather than rely on subscription service may be a big selling point.

Open Source

  • A way to gain traction and mind share; creates potential to sell other products and services that aren’t open sourced.
  • Benefit from development by other users.

Tooling

  • CI/CD, something to refine and work out during product development
  • Would make the roll out of new services and updates to existing services trivial
  • Should be applicable to any products or services that are used

Machine Learning

  • Using an API Gateway to decouple applications from the services providing the functionality allows you to role out implementations using ML for any endpoint.
  • Having all calls go through the API Gateway not only means that you can easily switch the service provider, it should allow you to record a rich set of data for ML purposes.

What to build?

  • Cloud Template Generator: SPA to generate Cloud Templates for AWS, likely GCP, maybe others?
  • Layout Engine: Helps decouple the frontend applications from the backend services while providing the business with control over content and layout
  • Frontend application: Decoupled frontend application that makes use of the API provided by the API Gateway
  • Site management: Application to manage “site” configuration, an important concept to build in early on
  • Common Commerce API: It may not be practical but, having a common interface that you can switch out the services provided for would be powerful
  • Proxy Services: Build out Cloud Templates to proxy to existing services: ATG 11.3, OCC, Hybris, Demandware, Shopify, Salesforce, etc
  • Custom microservices: Where you want to get to. Having a Common Commerce API would make this easier though isn’t necessary. Potential to explore ML driven content. Management applications to go with them? Management hub? Smart pricing, Smart Recommendations
  • AWS Cloud Templates: Consider building Templates to enable hosting existing services in the cloud.
    • Instance based is most straightforward but least desirable, ideally proxy to existing stack instead
    • Mix of containers and instances
    • Containers and cloud services, eg RDS for Oracle, EFS for publishing files, Elastic Search?

Cloud Template Generator

  • SPA to generate Cloud Templates for AWS, GCP, etc
  • Potential to open source application
  • Open source library of common templates; potentially include our own products and services
  • Build a private library of more client specific templates?

Layout Engine

  • Could be offered as a product or a service
  • Minimal product is API only, i.e. no management application
  • Potential to open source core product (API, Lambda functions); keep other functionality such as the layout management application, url management, form management, e.t.c closed source.
  • These applications can be built out over time, including adding features for versioning/workspaces, ACL, Preview, e.t.c
  • Build in Analytics recording
  • Build service for Multi-Variant testing
  • Build service for ML driven content using analytics
  • Build in reports based on analytics, own and potentially third party
  • Strictly speaking the Layout Engine is entirely optional, however it’s something where I can see us being able to add a lot of demonstrable value

Where to start?

  • DevOps tooling: should come about from developing the applications.
  • Cloud Template Generator: allows pulling together preconfigured Cloud Templates to easily demonstrate rolling out a customised cloud stack.
  • Frontend Application: simple React application to demonstrate functionality, uses minimal Layout Engine to drive content. Demonstrates DevOps tooling to role out application resources.
  • Layout Engine: defines the content to render on the site, provides business configurability. If developed as a product would demonstrate DevOps tooling to role out serverless functions, containers and other cloud resources.
  • Proxy to existing monolithic API: provide access to existing commerce functionality, potential to work on Common Commerce API, potential for other commerce providers, potential to host instead of proxy.
  • ML Services: Pricing? Recommendations? Requires access to data to develop and demo.
  • Other?

Talking Points

  • Cloud Hosting
    • Ideally want to retain control of cloud infrastructure
    • Providing our products as subscription model services versus licensing as products
  • Ideally you want to only provide:
    • proxies to existing products
    • containers
    • serverless functions
  • Ideally you don’t want to get into:
    • supplying tars
    • managing instances
  • Retaining IP:
    • new services created for custom needs,
    • third party integrations
  • Retaining data, especially in regards to future ML purposes