Do you remember what a LOB (Line of Business) application used to look like 15 years ago? One or multiple virtual machines, .NET 3.5, web and desktop applications and a SQL Server database.
Cloud was one of the catalysts that helped large organisations to change how a LOB application is built and designed. In this article, we do a radiography of a modern LOB application.
Nowadays, we build IT solutions using container-based technologies. What we will use in 10 years, we don't know. Technology is changing so fast to understand how serverless will impact the ways of designing and building a new generation of software. But for now, microservices, together with Kubernetes and managed cloud services, change how we build, run and manage our software.
Front Office
The front office runs on top of a Kubernetes cluster and exposes an API for internal and external users and systems. An Azure Load Balancer is configured on top of the cluster with Azure API Management, with a preconfigured and powerful WAF (Azure Web Application Firewall).
The web content is provided through App Services, which accesses the LOB API through the Azure API Management. The dynamic web content is served over Azure Front Door, offering a global entry point for all requests. Static content is also handled by Azure Front Door, together with Azure CDN and a public Azure Blob Storage.
A part of API security concerns is moved outside the computation units and microservices to Azure WAF, Front Door and API Management.
Azure Redis Cache and Azure Cosmos DB or Azure SQL are successfully used to optimise the Front Office. Most of the monitoring is done over Azure App Insights powered by Azure Monitor.
Business Logic
The backend system runs in the same way as Front Office inside a Kubernetes cluster in a microservice approach. The communication with Front Office is done indirectly using Azure Event Grid, together with Azure Event Hub and Azure Service Bus. Direct API communication between the two layers is based on events and messages to provide scalability capabilities and improve latency.
Like the Front Office, Azure App Insights and Azure Monitor are used to log and monitor. In addition to Azure CosmosDB and Azure SQL, Azure Data Lake can be used to store a large amount of data, that it is consumed by Azure Synapse Analytics.
Back Office
The more complex system might require a back-office system to communicate with external systems. This can be achieved easily by a Kubernetes cluster that hosts the adapters for each external system.
This layer might require Azure CosmosDB, Redis Cache. Azure SQL for storing the content. Storage is needed, especially when you have an async communication and the external system is not available to communicate with.
As for the rest of the layers, monitoring and logging can be done successfully with App Insights and Azure Monitoring.
Wait?! What?... Do I really need 3 different AKS instances for each layer?
NO, you don't. You can use only one instance or Kubernetes cluster when a service mesh like Istio. Depending on compliance reqs, complexity and your specific need, you can end up with the above approach.
In financial services or banking, where PCI-DSS is required or in healthcare (HIPPA), there are big chances to end up with the above approach.
What about other services?
Some additional services are required from a security and automation perspective. App Config, together with Key Vault, needs to be on our list, and for IAM, Azure AD and Azure Policies are almost mandatory.
For backup solutions, Azure Backup Service needs to go hand-in-hand with Azure Container Registry and Azure DevOps. It's excellent for your pipelines and development & release strategy.
Comments
Post a Comment