Load Balancing vs Clustering

Two Things That Look Alike

I do find sometimes there's some confusion between what is a load balancer and what is clustering. At first glance, load balancing and clustering seem to serve a very similar purpose, and they do cover common ground, but the two are different.

What Clustering Does

Clustering happens on a service. For instance, it's very common that we cluster database services. These are machines that are working together — the services are working together to deliver these database services. Same thing: we might have a cluster of servers that are working together to perform the same function. That's where clustering services shine.

Clustering is coordinating services or applications on the server. They're set up so that they're sharing these different resources, configurations and data amongst these services.

What Load Balancing Does

With load balancing, we actually have some sort of load balancer that balances the load between these servers. There are different types of load balancers, but specifically the one that we're talking about is balancing the load amongst these different servers.

Load balancing is more about balancing the load. It creates connections between the client and the servers, so more often than not we're dealing with something like network traffic. Let me be clear that there are other types of load balancers, although the one we're talking about here is more on the network side, balancing the load of the network, versus clustering, which is happening more on the databases.

Where Each One Shines

Clustering and load balancing each provide a performance increase, availability and scalability. What I would say, though, is that clustering is a little more designed towards availability and load balancing is a little more designed towards performance. They each have their area where they really shine, or were built for, although there are a lot of parallels between them. I like to think of load balancing as more focused on the application and traffic, and clustering as more based off the database and the application.

Using Them Together, and Apart

Often we use clustering and load balancing together. That is, we need to coordinate things on a database, and so we cluster those; or we've got an application that we need coordinated — maybe it's keeping track of things like sessions — and so we're using clustering technology to do that, but then perhaps we're also using load balancing to balance our different clients across these servers.

There are definitely times that we use clustering without load balancing. There are a lot of cases where I've done this, where we might have databases on the back end. We have a primary database that has this IP address to it, and then these machines make a connection to that IP address. But when there's a failover, what happens is either this secondary database takes on that IP address, or you make a DNS change, or something happens where now things get switched over to that secondary database server and then make a connection to that secondary database server. So we definitely don't need load balancing to be a part of all of this.

There are also definitely times I've not used clustering. For instance, maybe I have apps running that are not clustered apps, and we're using a load balancer to balance our clients amongst these different servers. What has to happen is that the session stays with one of these servers, and if things need to fail over to another machine, then that client is going to have to establish a new session with that machine — or we store the session back in our database or something to that effect. So somehow we need to compensate for that, but we definitely don't need clustering services to do that type of coordination.