Multiple Node High Availability with SQL Always On

Our customers are always requesting solutions with high availability and minimal downtime but for CCG we have limited options for this, our standard installations are usually a single server which gives us also a single point of failure.

The only options validated today is to run CCG in a Microsoft Windows Failover Cluster however this is often giving problems of its own since the clustering has strict dependencies on the hardware.

This request is to add the option to run CCG with high availability without Microsoft Windows Failover Clustering. One idea Arnaud Gallois and I had is to separate out the components:

SQL - SQL would not be installed on the same server as CCG but instead the CCG DB would be installed on a SQL Always on Cluster for high availability.

CCG Application - We would have 2 or more CCG application servers running independently in either an active / active scenario for higher throughput with high availability or in an active passive scenario for only high availability. The multiple CCG application servers would be accessed via a virtual IP on a load balancer and any TCP Port / Thread could be alive and processing messages on any CCG application node connected to the same SQL DB - like this we get high performance from running multiple applications nodes and high availability in case one application server needs to be removed for servicing (Windows updates etc).

SQL is supported like this already but the change here is around support for multiple CCG application servers. This will help with our very large customers like EMRAD who need faster processing of messages than a single CCG can handle and offer a strong high availability solution that is available today with Microsoft Windows Failover Clustering. We pay significant amounts of penalities due to performance and stability issues with CCG in EMEA and this would go some way to help reduce those - ~£250,000 in 2022