Implementing the Reactive Manifesto with Azure and AWS
- by Elton Stoneman
Originally posted on: http://geekswithblogs.net/EltonStoneman/archive/2013/10/31/implementing-the-reactive-manifesto-with-azure-and-aws.aspxMy latest Pluralsight course, Implementing the Reactive Manifesto with Azure and AWS has just been published! I’d planned to do a course on dual-running a messaging-based solution in Azure and AWS for super-high availability and scale, and the Reactive Manifesto encapsulates exactly what I wanted to do. A “reactive” application describes an architecture which is inherently resilient and scalable, being event-driven at the core, and using asynchronous communication between components. In the course, I compare that architecture to a classic n-tier approach, and go on to build out an app which exhibits all the reactive traits: responsive, event-driven, scalable and resilient. I use a suite of technologies which are enablers for all those traits: ASP.NET SignalR for presentation, with server push notifications to the user Messaging in the middle layer for asynchronous communication between presentation and compute Azure Service Bus Queues and Topics AWS Simple Queue Service AWS Simple Notification Service MongoDB at the storage layer for easy HA and scale, with minimal locking under load. Starting with a couple of console apps to demonstrate message sending, I build the solution up over 7 modules, deploying to Azure and AWS and running the app across both clouds concurrently for the whole stack - web servers, messaging infrastructure, message handlers and database servers. I demonstrating failover by killing off bits of infrastructure, and show how a reactive app deployed across two clouds can survive machine failure, data centre failure and even whole cloud failure. The course finishes by configuring auto-scaling in AWS and Azure for the compute and presentation layers, and running a load test with blitz.io. The test pushes masses of load into the app, which is deployed across four data centres in Azure and AWS, and the infrastructure scales up seamlessly to meet the load – the blitz report is pretty impressive: That’s 99.9% success rate for hits to the website, with the potential to serve over 36,000,000 hits per day – all from a few hours’ build time, and a fairly limited set of auto-scale configurations. When the load stops, the infrastructure scales back down again to a minimal set of servers for high availability, so the app doesn’t cost much to host unless it’s getting a lot of traffic. This is my third course for Pluralsight, with Nginx and PHP Fundamentals and Caching in the .NET Stack: Inside-Out released earlier this year. Now that it’s out, I’m starting on the fourth one, which is focused on C#, and should be out by the end of the year.
