Link balancers are routers, gateways and firewalls that are engineered to be able to spread Internet traffic over two or more Internet lines. This is a fairly straightforward functionality on paper, so why then can link balancers have significant performance differences from vendor to vendor? We will explore some of the inner workings of a link balancer with the goal of shedding light for IT professionals who are adding reliability and performance enhancements to their networks.
As we mentioned, link balancers are defined as network appliances with two or more WAN ports that spread network traffic between the WAN connections. However, one needs to realize that the distribution of traffic on various WAN Internet connections is usually not that straightforward, simply because of the varying nature of the WAN connections. In other words, WAN connection performance will change depending on the time of day, the backbone network conditions and congestion, the cross traffic on the office network and various other factors (for example if a WAN has a time-varying transport, such as 4G LTE or 5G wireless, those transport changes will directly impact the WAN connection characteristics as well). These time-varying fluctuations of WAN links will result in changes to network performance parameters such as throughput, latency and packet loss.
Legacy Link Balancers
As an IT manager, if you wanted to install a link balancer to your network several years ago, chances are you were presented with several link balancing options in the configuration such as round robin or weighted round robin. These round robin algorithms are meant to assign the traffic sessions going through the link balancer to a specific WAN link so that in aggregate, for round robin, the sessions are equally distributed over the WAN links, while for weighted round robin, the links are distributed according to a pre-defined weighting function. Of course in an ideal world where WAN links have unchanging performance and uptime, these approaches would work fine. However, in practice, real WAN links quickly will create problems when (and the term is “when” not “if”) networks fluctuate in their performance or get into brown-out or even blackout situations. In that scenario, the round robin and weighted round robin algorithms will fail to load balance effectively, significantly impacting the end-user experience.
Another problem with legacy link balancers is the fact that traffic assignments are not done intelligently. As an example if you have an application with several sessions that require those sessions to present the same IP address, this will present a challenge in legacy link balancers as the sessions may be distributed among different WAN links and therefore different IP addresses.
A similar example is your STMP email traffic. Usually, the email servers (sometimes maintained by your ISP) would do source IP filtering and therefore will expect and accept emails only from the specific IP address that they configured for you. As you guessed, this will break if the email sessions are link balanced over different WAN links having different IP addresses.
Modern Link Balancing
Fast forward to a decade later (by the way, the underperforming legacy link balancers still exist in the market), and a new breed of link balancers are available. Usually grouped under the umbrella name of SD-WAN (Software Defined WAN) or WAN Virtualization devices , these modern link balancers try to address the shortcomings of legacy link balancers in various ways. If the SD-WAN is capable of HTTP bonding, the device can adjust the parts of the download to be spread among available WAN links and dynamically adjust the sizes of the pieces of the download so that full utilization and dynamic rate adjustments can be done. This significantly outperforms legacy link balancers since the SD-WAN router with broadband bonding is capable of achieving the sum of the available link rates even for single file transfers, which is not possible with legacy link balancers.
SD-WAN on auto-pilot
Similar to Tesla’s self-driving electric vehicles, not all SD-WAN devices are created equal. Some SD-WAN solutions are equipped with advanced, self-driving capabilities. The auto-pilot analogy applies to the dynamic and adaptive adjustments that the SD-WAN link balancer can implement on the fly depending on the changes in the WAN conditions. As an example, a VOIP optimized bonding tunnel (such as VOIP Armor, by Mushroom Networks) will measure the WAN link quality with respect to the parameters that matter for the application at hand (VOIP), namely the packet loss, latency and jitter that constitute the MOS score. What this means is that variations in the WAN performance, including brownouts and blackouts on individual WAN connections, can be disguised from the VOIP packets, as the SD-WAN link balancer routes around those network problems without requiring any human intervention.
The application centric overlay tunneling presented by modern SD-WAN routers combined with the ease of use of these devices are making zero-touch networking, not just of installation but also for network operations, a reality. Network admins now can truly set their networks on autopilot and convert their emergency support tickets into schedule maintenance calls with their service providers.
Cahit Akin, CEO, Mushroom Networks, Inc.
Mushroom Networks is the provider of Broadband Bonding appliances that put your networks on auto-pilot. Application flows are intelligently routed around network problems such as latency, jitter and packet loss. Network problems are solved even before you can notice.