I want to know more about how auto-scaling load balancing works

Load-balancing has traditionally been difficult to do effectively

In a web hosting environment, when you need to scale to more than one server, a naïve approach is to simply split your websites so that you have say 100 located on one server, the next 100 located on another server and so on. This is how many web hosts have operated in the past.

However, a problem comes when two or three of the websites on a particular server become very busy and cause it to slow down. The problem is, you have spare capacity on other servers, but you are not able to utilise that capacity because the three busy websites in question all happen to be on the same server. In the past you might have manually moved these websites onto separate servers to more evenly spread the load and make more effective use of available hardware, but the limitation of this approach is that it’s a manual process that requires constant management as the workload grows and changes.

What you really need is a way of automatically balancing the load on your servers so that no single server is ever getting so much load that it becomes unresponsive or worse, crashes. In the past a number of different approaches have been taken to solve this problem, but all of them suffer from limitations and flaws.

Hybrid Web Cluster’s load balancing is done entirely in software – no expensive hardware devices are required, just normal commodity servers or virtual machine instances. This eliminates a single point of failure in the hardware load balancer itself, and makes Hybrid Web Cluster a more robust system.

Hybrid Web Cluster offers a revolutionary new type of load balancing

The load balancing system in Hybrid Web Cluster works by splitting websites over a number of servers, say 100 on each server as in the example above, but the crucial difference is that Hybrid Web Cluster keeps a close watch over the load generated by each individual website or database, and its sophisticated algorithms make judgements about when to transfer websites from one server to another to spread load more evenly. These transfers happen transparently to the end user, competely automatically, we call this process site-juggling.

This means a website can go from having no traffic at all and sharing a server with 99 or more other websites to getting a massive spike in traffic, say if the website address is featured on TV – Hybrid Web Cluster simply reacts by moving all of the quieter sites hosted on the same machine to another node, leaving the original machine as a dedicated server for the busy site in question. The same process happens for databases, meaning the system can automatically scale from a site requiring virtually no resources, to that same single website requiring a dedicated database server and separate dedicated web server, in a matter of seconds.

Hybrid Web Cluster runs equally well on real hardware and cloud server instances, this means websites can be juggled from one to the other by the load balancing algorithms according to conditions that you specify. This allows you to spread load evenly between real servers and cloud, or use cloud servers as an emergency overflow to cope with peaks in demand.

Hybrid Web Cluster’s powerful visualisation tools allow you to view all of this happening live on your cluster, and of course monitor disk and bandwidth usage of each website or user. This gives you all the business intelligence necessary to make decisions on pricing or account quotas.