Autoscaling

 * Without Autoscaling: The fixed cost approach

 * With Autoscaling: The variable cost approach

Autoscaling is one of the remarkable advantages of using the public cloud because of its elasticity of the volume usage. This factor is also beneficial by means of the financial accounting.  However, Autoscaling is not that perfect as much as the salespersons from the public cloud providers advertise.  This short essay picks up the typical benefits and challenges of installing the public cloud especially because of its Autoscaling.

Autoscaling is a notable advantage of the public cloud which is usually emphasised in the introductory guideline of the public cloud advertisement. First of all, it allows its users to elastically adjust their quantity usage at each time period, instead of fixing the quality usage, depending on the fluctuant quantity demanded at each time period. This is useful to avoid both the dead-weight of the volume usage while the low access traffic and the volume shortage while the high access traffic.

Focusing on the accounting aspect, it can covert its usage cost to the variable operational cost instead of the capital or fixed-cost = liability. In many cases, it is often advantageous because it is likely to avoid paying the high tax levied onto the capital expenditure. By contrast, the cost covered for the public cloud is often accounted as the variable operational cost which is actually useful to reduce the total tax payment. Regardless of the total cost amount, in many administrative accounting cases, it provides individuals and their firms with the tax payment advantages.

At the same time, Autoscaling does not perfectly work as the textbooks and the introductory guideline of the public cloud explains. In reality, there is a limit of calling for extra instances at once because there is a certain maximum number of the server instances within a certain time period e.g., being able to call 5 instances per minute. There is a physical barrier of scaling the instances depending on each situation.

Furthermore, there is a case that the load balancer misjudges the health check of each connected server instance and the access control. This load balancer can accidentally call unwanted instances by falsely detecting as broken, deleted, or over-capacity or some errors of the monitoring tools and the metric/log information. In addition, there can also be a human error of typing the programmes of directing the traffic control or the security/monitoring setup. This may incur a considerable financial disadvantage of excess instances called by Autoscaling. 

All in all, the Autoscaling mechanism is still an attractive characteristic of using the public cloud to optimise the operational availability as well as the financial efficiency. Nevertheless, it must take an account of the dysfunctions of Autoscaling such as the latency of calling the instances and the extra possible costs caused by both machinery and human errors. On the top of what the salespersons and the textbooks indicate, the users had better refer to the actual experiences of its real application.