This blog post was authored by David Thorman, who leads Ops at Kloudless.

In our last post we covered the different options for deploying Kloudless Enterprise. When deploying Kloudless Enterprise, you would start off with a single server. However, when running any application, there is a limit to how much work a single server can handle. A server can be scaled vertically by allocating more CPU cores and memory, but eventually those limits will be reached and another mechanism to increase capacity must be found. This is where clustering is useful.

Clustering allows your application to scale out horizontally rather than up which leads to two benefits: higher throughput capacity and high availability. Clustering is made possible by running more than one Kloudless Enterprise instance behind a load balancer. A load balancer is a server that accepts requests from clients and then forwards them to the backend Kloudless Enterprise servers. This allows the client to take advantage of multiple backend servers without having to manually keep track of their hostnames. The distribution of work allows the cluster to handle a higher number of requests than a single server would, resulting in greater throughput.

The cluster can continue to serve requests in the event of a failed secondary instance, ensuring high availability. If the primary instance fails, a secondary instance will be promoted to primary, ensuring smooth disaster recovery. Each node is capable of handling similar work, and is configured to be the same size. This means the failure of a single node will not result in the cluster not being able to serve requests.

Service interruption can be minimized by ensuring that the load balancer only sends requests to nodes that it realizes are healthy. This is typically achieved via health checks that the load balancer performs. For example, in AWS, the Elastic Load Balancer’s health checks take the form of an HTTP request to each individual node in the cluster. Nodes that either don’t return or return a non-200 status code are marked as unhealthy. This allows the cluster to handle requests successfully even though there are failures.

Thanks to the HA/DR features above, the cluster can be dynamically scaled up or down without interrupting service to clients. Adding new nodes increases capacity when your service experiences periods of higher load. Removing nodes allows costs to be reduced during periods of lower load. On certain platforms ,these changes in capacity can be handled automatically either on a timed schedule or based on metrics gathered from the cluster itself. This allows you to take full advantage of the elasticity of modern IaaS platforms such as AWS without disrupting service to your customers.

This post has covered the high-level benefits of clustering and the flexibility of a Kloudless Enterprise deployment. Our configuration guide covers the technical details of deploying Kloudless Enterprise as well as a walkthrough detailing how to deploy Kloudless Enterprise in an auto-scaling cluster. The guide can be accessed by emailing us at hello@kloudless.com. Feel free to reach out to us or comment below with any questions.

Thanks for following our Kloudless Enterprise Series!