Two Types of Scaling
Two methods of scaling are used to address fluctuating enterprise computing demands.
Vertical scaling is often also referred to as scaling up. When a system is scaled vertically, more powerful hardware is employed to provide enhanced capabilities. The limits of the hardware define the limitations of a vertical scaling initiative. Vertical scaling is easier to implement and can often be done on-the-fly but may eventually bump into the physical constraints of the hardware.
There are many specific ways a system can be vertically scaled. Moving to a new virtual machine with more memory or a faster CPU may solve problems with compute capacity. Additional storage can be provisioned to handle the greater size of transaction logs during database usage peaks. Improved network connectivity can be implemented to handle a surge in customers without degrading system performance.
Horizontal scaling is also called scaling out. Rather than building more capacity into an existing system, this method deploys additional systems to help process the workload. Horizontal scaling requires an architected solution and does not lend itself to being done quickly. The benefits of horizontal scaling is that when architected correctly there is virtually no limit on the increased capacity available to handle the workload.
You may also see reference to side-by-side scaling which involves cloning or spinning up replicas of a system for testing and development purposes. This type of scaling does not address the needs of variable seasonal demand.
Three Ways to Take Advantage of Cloud Scaling
Organizations can take advantage of scaling in three ways. In the following descriptions the presumption is that the solution has been adequately architected and is ready to be operational when needed.
Manual scaling requires human intervention to manage the scaling operation. This requirement open the door to potential human error, such as failing to scale down which can result in unnecessary financial expenditures.
Scheduled scaling occurs based on s set schedule defined by the customer. This method removes the chances of human error but may be too inflexible in some instances. Changes in demand for computing resources may not conform to a regular schedule.
Automatic scaling or autoscaling is done based on predefined rules and thresholds. In many cases, autoscaling is the preferred method of scaling computing resources. It ensures that the necessary capacity is there when needed and scales down to minimize costs.
Businesses that have seasonal workload demands can use cloud scaling techniques to maintain high performance and meet customers’ expectations. With the judicious use of scaling, organizations can keep costs under control while obtaining the competing resources they need to thrive. It’s just another reason that the cloud is seen as such an appealing destination.