Quite often there is a confusion when comparing legacy network load balancing with broadband bonding technologies. The truth is that although these technologies look similar there are many differences between them. The ability of broadband bonding to simplify WAN management and boost the network performance has made it already the new trend in corporate networking.
Both broadband bonding and network load balancing spread the network load on individual links while ensuring that the data is routed to the functional connections isolating the non-functional ones. The main difference lies in the level of the implementation of the function of sharing the burden of network data traffic.
Why WAN load balancing cannot cope with modern business connectivity needs?
Conventional load balancers spread the load at the session level, i.e. by choosing different paths / broadband connections for routing TCP or UDP sessions. This means that each session is assigned a specific broadband connection with a maximum speed of that single connection chosen. In case a broadband connection fails, the session will be terminated and a new session has to be initiated which will be routed through a functional connection from the pool of the available connections. In practice this means that if, for example, a local network user had opened a VoIP session ( e.g. Skype call), and the connection which was used for this session was interrupted, the session would be discontinued and the telephone conversation would be terminated. The user can call back the caller and a new session will be initiated through another functional link, but it is easily understood that this may not be acceptable in a corporate environment.
How broadband bonding is different than load balancing?
With comBOX broadband bonding technology, the network load is distributed on the packet level (TCP or UDP packets), which means that a session can be routed through all the available broadband connections, essentially by distributing its data packets to all connections. This results in utilizing the available bandwidth of all individual broadband connections simultaneously, achieving high speeds, both uplink and downlink, even for a single session. In case that some of the available connections fail, the session is not interrupted but only limited by their speed contribution. In a practical scenario, if a user on the local network has initiated a VoIP session, data packets of this session will be distributed to all available broadband connections. Thus, the failure of some connections will cause the packets which were transmitted via the malfunctioning connections to be rerouted via the functional connections and the connections’ failure will not be perceived by the user as the telephone conversation will continue.
Another aspect that should be taken seriously into account, is the capability of prioritizing applications for business critical operations. Defining QoS prioritization classes in a WAN load balancer, is a quite complex process as it has to be laid down for each individual broadband connection. Thus, a single centralized policy cannot be clearly defined and applied, as the total available bandwidth capacity is not known. Additionally, when using DSL type connections whose performance is fluctuating, QoS priority rules cannot be applied correctly due to unstable and problematic performance of each individual connection. Using the broadband bonding technology that is incorporated in comBOX services, priority rules are defined for a single data channel while the capacity is calculated in real time allowing the reliable use of QoS classes.
So, when someone finally compares the two technologies, the conclusion is that the only common element is the use of multiple broadband connections to increase overall network capacity and reliability. However, broadband bonding technology has more features in common with leased lines as it creates a single connection with aggregated speed and reliability.