Global Server Load Balancing (GSLB) is an enhanced DNS-based solution for server load balancing and failover in globally distributed server environments. In recent times, companies are migrating from traditional on-premise single data infrastructure to geographically distributed multi-data center environments by using hybrid on-premise and cloud infrastructure or entirely single/multiple provider cloud infrastructure. The demand for load balancing and failover in such complex environments requires a specific solution that is not available through the use of standard data center load balancers.
Global Server Load Balancing collects status information from servers located in different data centers, processes this information and uses the standard DNS infrastructure to optimally distribute the servers IP address(es) to clients. Generally, the server information collected consists of server status (available/unavailable), current server load (CPU and/or memory load measured in [%]) and/or server’s service response time measured in [msec].
The traditional Global Server Load Balancing uses a variety of techniques for server load balancing such as round-robin, weighted round-robin, least connections and random. Server status information is collected by using built-in GSLB functionality (load agent installed on a physical server and/or custom techniques for service response time measurements) or by using GSLB Rest API functionality and externally collected and delivered server status data.
Another very important part of GSLB functionality, except server load balancing and failover, is optimal server selection for clients. Since each client has a unique network distance to each distributed data center (server), a very important aspect of GSLB functionality is to select the data center (server) with the shortest and fastest network distance for each client. The traditional GSLB uses server-side approximation, such as geographical proximity by using static proximity or geotargeting. A modern GSLB uses client-side real-time network distance measurements and client-side algorithms to accurately measure the network distance to each offered A /AAAA IP address and to use this data together with server-side information (server load and/or server network service response time) to determine the optimal server for each particular client based on its topological (network) distance from the server and current server status. Such enhanced capabilities enable companies to transform their traditional active-passive data center functionality into active-active to meet customer needs at full capacity, lower service response time and reduce costs at the same time. Client-side GSLB also enables companies to reduce the use of traditional server load balancing by making servers directly available to their clients, where client-side GSLB functionality, along with the server-side GSLB information provided, selects the optimal server for each client. Always. Anywhere.