What is AutoScaling and the implications on the online sector
Each time one connects to a website, he/she is connected to a specific server which hosts all of the necessary data. It doesn’t necessarily mean that the server you are connected to is the only server with such data. It is common for companies to have a number of servers spread out over the world in order to prevent dropped connections when a server is being overloaded and to limit the risks of Distributed Denial of Service (Ddos) attacks. A load balancer monitors the traffic coming to each of the servers and redirects you to a server, which is currently being used the least.
At the present point in time, there is a very distinct lack of flexibility with regards to the number of servers hosting a website and its location. Companies generally run their webserver infrastructure on a fixed number of servers. The AutoScaler is a relatively new introduction to this rigid system and involves the scaling up or down of the number of servers depending on the RAM and CPU consumption of each server.
Online traffic fluctuates significantly throughout the course of a day. The figure below illustrates this trend in a graphical manner.
As shown, one server is sufficient to handle the load from midnight to around 6am on a regular working day. The AutoScaler allows for scaling up operations at around 6am and scaling down operations at 1pm and then again at 9pm. Having a rigid infrastructure in place instead of the autoscaler would mean that all 4 or even 5 servers are running throughout the day to accommodate for this expected traffic. Not only does the autoscaler allow for savings on behalf of the company in terms of server costs, it is also a more energy efficient way to organize your web infrastructure making it a much more sustainable solution in the long run.
The AutoScaling solutions currently provisioned on the Amazon Web Services Platform and the Google Cloud platform only scale servers based on their CPU and Ram consumption. These parameters, however, are not ideal indicators of actual web user experience. A Dutch start-up company by the name of Treestle has tackled this problem with the introduction of the LiquiD AutoScaler which monitors the page load time of the end-users and scales the servers purely based on this. The CEO of trestle states: “LiquiD AutoScaler offers hands off scaling so the system administrators don’t need to think about the relationship between server statistics and actual website performance”. Many believe that this truly automatic scaling solution is the future of the online sector due to its user-centric functionality.
The increasing popularity of this feature amongst Infrastructure as a service (IaaS) providers such as Amazon Web Services, Interoute and Rackspace proves the need for flexible web infrastructure. However, big corporations such as UEFA and Nike still prefer to have their infrastructure organized in a rigid manner due to fears of a failure in the autoscaler leading to dropped connections resulting in forgone revenues. Is this new technology the future of web infrastructure or simply a fad which is bound to fail? Will the scaling operations be done based on actual web-user experience or based on server statistics, which are much easier to obtain? Only time will provide the answers.
M. Floss, 356166mf