Load balancing between lines

Speed reduction experienced by an Internet server due to information congestion can be normally fixed by updating the processor or by expanding either the hard disk or the memory. An alternative would be hiring additional lines to enhance the access to the network. However, this is usually a temporary solution because of the gradual and constant growth of traffic and its higher concentration in specific areas. In this context, load balancing between lines emerges as an attempt to tackle this issue, and this is why at Televes we would like to give you some keys.

What is load balancing?

El balanceo de carga entre líneas consiste en la distribución de las peticiones de internet a través de una línea de servidores. El proceso se lleva a cabo gracias a la capa de red (nivel 3) de OSI. La forma de llevarlo a cabo es a través de varios mecanismos. Por un lado, encontramos el más simple, conocido como “Round Robin”. Este consistirá en repartir todas las peticiones que lleguen de internet entre los servidores disponibles para dicho fin.

Load balancing between lines consists on distributing Internet requests through a line of servers. This process is performed thanks to the OSI (level 3) network layer by means of several mechanisms. The most simple one is known as “Round Robin”, which implies distributing all the requests from the Internet to the available servers aimed at that purpose.

Another alternative is using equipment to receive such requests and collect – in real time – the information about their operative capacity. This information is then used by routing those requests individually to the server offering the most optimal service and featuring the best disposition.

Load balancers can bring either hardware solutions – such as switches and routers including optimized load-balancing software – or software, like installations at the servers back end.

This is how you obtain a system that allows connections to be distributed among the different links reaching a maximum connection speed equal to ___ the line assigned to such connection.

Bonding method or LACP

This mechanism operates at layer two or link level. Accordingly, several lines can be used as a single one, bringing about an increase in the bandwidth. This method is mainly used to interconnect equipment at great speeds, e.g. switch-OLT or switch-switch, among others. Therefore, by implementing this system, traffic aggregation can nearly multiply the number of links depending on its capacity.

A good example of this would be the use of two Gigabit Ethernet lines to connect two switches to each other, allowing for a performance of nearly 2 Gigabytes per second. In the same way, those ports involved in LACP or bonding will bundle so as to perform with a single logical bonded interface.

Using several balanced lines at the Internet access

Thanks to the load balancing mechanism you can obtain faster connection speeds as compared with those offered by the lines of the available operator. Their functioning is based on a process implemented by the main router of the installation where the different contracted network-access lines will be connected. Load balancing may be, thus, enabled in that router to start distributing traffic. This router will route the different packages of each user connection through an interface, distributing traffic among the lines.

A user with one single connection will not benefit from the sum of the line speeds, but from the speed of the line assigned to such user. However, the speed of the said line may be fragmented, as it might be routing other connections.

Even though, the effect will be highly satisfactory for multi-user environments, as thanks to this method, performance will be quite close to the sum of speeds of the lines distributed among them. Depending on the balancing configuration, further characteristics such as fail-over or ponderation may be implemented.

Drawbacks resulting from the use of load balancing between lines

Despite the number of advantages of load balancing, it is worth considering some aspects. First, balancing must be configured appropriately, otherwise the router may not ensure connections always routed through the same interface.

In some cases, when logging in on a specific website for authentication and sending of new requests, connection might be performed via other lines. Origin IPs seen by the web server will be then different to the ones from the authentication phase, which might lead to the cancellation of our user session for security reasons. Another possible scenario would involve an internal server publishing to the internet; it might send the packages requested by the user through the non-associated interface with the remote request.

In conclusion, we can opt for the operator’s solutions in case of a high-capacity request. This could be a real aggregation if it offered bonding or LACP to our main router. The second option would be for us to hire several lines and perform the load balancing between them by ourselves. Each manufacturer will have their own configuration mechanisms and we should know them all to perform an optimal configuration of the service.