Some workload disruptions can devastate a business, so IT organizations usually protect critical workloads via various strategies collectively known as high availability.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
One of the most common strategies for high availability is server clustering. Two or more physical servers process the same IT workload. As far as the clients -- end users -- are concerned, the cluster behaves as a single server with one IP address
Although it is possible to configure an active-passive environment where one server -- node -- does the work and others stand by, it is increasingly common to establish active-active clustering where each node shares the workload. An active-active approach offers more computing capacity because multiple servers are available to do the work. It also provides redundancy -- if one server fails, the others continue to operate and service the workload without any disruption to users.
The key to any server cluster is load balancing, which channels traffic to and from the workload, distributing that traffic across the numerous cluster nodes. In an active-passive configuration, the server load balancer recognizes a failed node and redirects traffic to the next available node. In an active-active configuration, the load balancer spreads out the workload's traffic among multiple nodes. Distribution may be equal, called symmetrical distribution, or uneven –asymmetrical -- depending on the computing power of each node or how an administrator prefers for the active-active cluster to behave. For example, an older server in the cluster may receive a lesser percentage of traffic while a newer server receives more.
Some load balancers distribute traffic dynamically in response to server loads, so overworked servers with falling response times temporarily receive less traffic, streamlining traffic and optimizing overall workload performance. This makes the most out of the server cluster's total compute capacity
Selecting a load balancer for data center server clusters requires a keen knowledge of traffic requirements and a clear understanding of the necessary feature set. Clusters are not difficult to set up, but optimum load balancing and active-acting clustering performance demand testing and IT architectural design expertise.
Load balancers are an integral part of any enterprise application cluster, whether deployed locally or in the cloud. The issue of cloud services simply adds another layer for system and software architects to consider in relation to workload availability. Enterprises that rely on the public cloud for enterprise application deployment typically include the cloud provider's load balancing services to distribute application traffic among multiple redundant compute instances. For example, Amazon Web Services' elastic load balancing distributes incoming workload traffic across AWS EC2 instances for greater application availability. Google Compute Engine also provides load balancing services. IT organizations can implement load balancers in hybrid cloud environments to distribute traffic to applications spread between cloud setups.
How can I test server high availability?
Three management concerns for hybrid cloud
Load balancing in the private cloud
Dig Deeper on Real-Time Performance Monitoring and Management
Related Q&A from Stephen J. Bigelow
OpenStack scheduled numerous hypervisors for deprecation in 2014's OpenStack Icehouse, but no others are scheduled for future releases, up to and ...continue reading
There are many differences between OpenStack-supported hypervisors, but only some features are mandatory. Adopters need to review feature sets as ...continue reading
VIC supports container creation and image deployment through virtual container hosts, which suit well-proven workloads, or Docker container hosts, ...continue reading
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.