Optimal Server Selection – Why it’s so important?

In today’s world, there is a rapidly increasing demand for all types of static and dynamic content. The type and quality of the online service delivery is highly dependent on the provider, server placement, the quality and quantity of servers and the server selection mechanism. Poor service quality always leads to frustrated and dissatisfied customers and that can significantly slow the growth of the service. Most online services are provided by multiple identical servers spread across different geographical locations in order to improve the performance and stability of the service. Server placement always plays a major role in the quality of the service and it is always advisable to place most of the servers in a location that has the shortest geographical and network distance to the majority of its users.

Cloud and web hosting, SaaS, business and finance, streaming and entertainment companies compete against each other in a highly competitive market and the quality and reliability of their services and applications is often key to retaining existing and attracting new customers.  Their users are not interested in spending more than a few seconds waiting for the content to load, feature to work or even suffer from constant server outages or maintenance.

There are several factors that need to be taken into account during the server selection process. Network distance between the client and server, the service response time and the current server load are all very important in the final selection decision. If the network distance is too far and the service response time high, the client will have a very poor latency and if the server is running at maximum load and is constantly saturated, the service will have significant delays.

To handle these issues, the Domain Name System (DNS) for Global Server Load Balancing (GSLB) is used and a local server load balancer if needed, depending on the capabilities and features of the GSLB. There is a variety of methods that are used, some more precise and efficient than others. The main difference in various server selection methods is the number of parameters that are used and their type (static or dynamic). Common static server selection methods may include topology(geolocation) to assign clients to the closest server depending on the DNS resolution and topology records or a Round Robin method to alternately distribute client requests to each server and then after each server has handled a request, repeat the process again. These methods however are flawed because geolocation can often be misinterpreted and imprecise and the round robin method doesn’t take into account for how long each client will stay connected to the server which can then lead to individual server saturation.

The best way to determine the optimal server for each client is to use a Dynamic Server Selection process that is backed with real-time data about the load status of the server (CPU & RAM utilization), a measured network distance between the client and server and the service response time. Combining those three methods and parameters in the optimal server selection process will always ensure the maximum customer satisfaction with a responsive and smooth service. The distribution of online services is of great importance because it can make or break a new app and it’s something that every large company needs to optimize if they want to remain competitive in today’s market.