With ‘admissions season’ underway, parents, carers and guardians are able to register for school places. For those authorities who have moved to online admissions applications, this is a time where the webserver plays a particularly critical role.
But what would happen if, in the event of a disaster, the webserver were to fail? Our skilled members on the Service Desk and One Technical Services would immediately begin the triage process and start to investigate the issue with your IT department.
This is where load balancing, ie how you control traffic between two or more servers, becomes important. The question is to NLB or to not NLB.
Windows load balancing, known as NLB (Network Load Balancing) creates a cluster on a virtual IP address, to which webservers or ‘hosts’ are added. If you connect to the cluster you’ll get content from either host, the end user isn’t affected as long as both hosts are serving the same content.
NLB assigns priority to each host so, if there are two webservers, the priorities would be 1 and 2 with host 1 being the default host. The default host will serve content, but if it can’t do this, the next highest priority host would serve content through the process known as ‘convergence’.
NLB hosts contact each other periodically to ensure they’re still ‘alive’, known as a ‘heartbeat’. So if host 1 were to fail, and no longer responded to heartbeats, NLB would automatically set host 2 as the default when converging to rebuild the cluster. Traffic requests would go to webserver 2 and disaster would be averted with web services remaining up and running whilst host 1 is being fixed.
NLB can also remove a host temporarily for maintenance while the other host(s) remain active, perhaps to reboot the webserver or install additional software while out of the cluster. The end user would never know anything was happening. The host can then be added back to the cluster, again with no impact on the other webserver(s).
However, in the event of a disaster there may be brief outage of services as NLB reconfigures and converges the cluster. It can also take time for the cluster to detect a host is no longer alive as the host would need to fail to respond to a number of heartbeats. This is because a single heartbeat could fail due to a non-critical issue such as a brief network error which is why the default heartbeat failure can be set to a custom value.
One Technical Services strives to enable load balancing on all environments where there are multiple webservers but sometimes network limitations can prevent NLB working. If so, we would recommend a physical load balancing solution.
Even more control can be leveraged when using physical load balancing solutions. Our hosted eSuite datacentre makes use of this whereby, using the load balancer, we can disable access to a particular site for scheduled work, such as data imports. All other customer sites within the environment remain up and running with no performance impact at all. We can also disable access to all customer sites when we upgrade the environment to the latest release version and then enable access for everyone when the upgrade is complete.
Our Technical Services team can deliver the most effective support to our customers via the use of physical load balancing technologies.
NLB is key to preserving uptime or for performing seamless maintenance while applications are still available. In today’s world it’s crucial to ensure on demand services remain up 24/7 and although it’s not always possible to prevent a disaster, the potential for downtime with NLB or a physical load balancing solution is reduced significantly.
The answer to the question may not always be NLB, but you should always consider a form of load balancing to preserve uptime with multiple webservers.
If you are interested in scalability of web services or eStart hosting please contact your account manager for more information.