Azure public multi-access edge compute (MEC) is a great platform for hosting applications at the edge and can make them more responsive, but it doesn't currently support high availability features. This article describes how to deploy workloads in active/standby mode to achieve high availability and disaster recovery.
Apache®, Apache Ignite, Ignite, and the flame logo are either registered trademarks or trademarks of the Apache Software Foundation in the United States and/or other countries. No endorsement by The Apache Software Foundation is implied by the use of these marks.
Download a Visio file of this architecture.
Azure Traffic Manager. Configure Traffic Manager to use priority routing. Set the load balancer IP address in Azure public MEC (primary) to Priority 1. Set the one in the secondary region to Priority 2. This configuration sends all traffic in the non-failover case to the Azure public MEC.
Traffic Manager for Azure public MEC doesn't currently support performance routing, which could dynamically determine the previously described routing based on the lowest latency to the endpoint.
In this architecture, failback is automatically achieved after the virtual machines (VMs) and/or standard load balancer is back online. Traffic Manager determines that the workloads are up and reroutes traffic back to the primary Azure public MEC region.
Public load balancer. This load balancer fronts the application tier and balances traffic to the pool of VMs in the virtual machine scale set.
Internal load balancer. This load balancer is used to access the database layer. Depending on the type of database you use for your application, you might not need a load balancer here, assuming other platform as a service (PaaS) services have their own load balancer.
Azure Virtual Machine Scale Sets. Most production deployments use Virtual Machine Scale Sets to dynamically scale their workloads based on traffic load. Azure public MEC also supports Azure Kubernetes Service for cloud-native and container-based applications.
Database tier. Azure public MEC doesn't currently support SQL database PaaS services like SQL Server on Azure Virtual Machines and Azure SQL Managed Instance. It also doesn't currently support NoSQL PaaS services like Azure Cosmos DB and Azure Managed Instance for Apache Cassandra. You can deploy third-party solutions that support SQL or NoSQL services and replication of data across their geo-distributed clusters.
- Azure public MEC is an edge computing solution that brings together a portfolio of Microsoft compute, networking, and application services that are managed from the cloud.
- Azure Traffic Manager is a DNS-based traffic load balancer. You can use it to direct incoming DNS requests based on a routing method that you choose.
- Azure Load Balancer provides high availability and high performance for your apps.
- Azure Virtual Machine Scale Sets enables you to manage and scale up to thousands of VMs.
- Actively replicates VMs from the Azure public MEC to the parent region and makes them available to fail over and fail back in case of outage.
- Backs up VMs to prevent data corruption or loss.
This approach costs less than the one described previously because there are no idle resources. This alternative is suitable only for applications that allow for higher RTOs.
Azure backup and disaster recovery for Azure public MEC currently supports only virtual machines.
Potential use cases
Use this architecture when you want to deploy workloads in active/standby mode to achieve high availability and disaster recovery. This scenario is ideal for the telecommunications industry.
These considerations implement the pillars of the Azure Well-Architected Framework, which is a set of guiding tenets that can be used to improve the quality of a workload. For more information, see Microsoft Azure Well-Architected Framework.
Azure public MEC is designed to host latency-critical applications. Because failover to a secondary region increases the latency of the workloads, it might not provide the same level of performance. Depending on the application and its sensitivity to this increased latency, you need to decide which of the services, if any, should fail over to the region.
Data replication and backup are important when you rely on database failovers. Most Azure PaaS services have built-in support for geo-replication and creating read replicas across regions and geographies.
Azure public MEC doesn't currently support PaaS services like SQL Managed Instance, SQL Server on Azure Virtual Machines, Azure Database for MySQL, or Azure Database for PostgreSQL. Third-party offerings like Couchbase, MongoDB, and Apache Cassandra can provide infrastructure as a service (IaaS) services that support geo-replication.
Traffic Manager supports multiple routing methods: performance, geographic, priority, and more. To best support low latency applications, dynamically send data to the region / Azure public MEC that's closest to the user. Performance routing isn't currently supported on Azure public MEC. The next best option is to statically prioritize the best location for an application.
For a globally distributed application that has workloads distributed across multiple Azure public MECs and regions, use a nested routing method. Use geographic routing to split traffic to the correct region, and then use priority routing to further split the traffic.
After the workloads in Azure public MEC are back up, Traffic Manager probes detect that it can take requests and automatically reroute traffic back to Azure public MEC.
Cost optimization is about looking at ways to reduce unnecessary expenses and improve operational efficiencies. For more information, see Overview of the cost optimization pillar.
Azure public MEC is primarily used for low latency and real-time computation scenarios. Data is processed by the compute instances that run in Azure public MEC. This architecture uses active/standby with a hot standby. That is, workloads in the secondary region aren't used unless there's a failover.
This approach to deploying workloads as a standby incurs Azure deployment costs even though the workloads aren't used.
For more information about pricing:
- See Azure pricing.
- Use the Azure pricing calculator to estimate the cost of implementing this solution.
For information about creating a cost-effective workload, see Overview of the cost optimization pillar in the Azure Well-Architected Framework documentation.
This article is maintained by Microsoft. It was originally written by the following contributors.
- Adhip Gupta | Senior Program Manager
- Azure public MEC
- What is Traffic Manager?
- What is Azure Load Balancer?
- What are virtual machine scale sets?