How to reduce latency, jitter and loss across your network

Posted:

Last week I looked at SDN and what benefits it could provide for enterprise network through the use of policies. In this blog, I’ll turn my attention to SD-WAN.

SD-WAN uses a lot of the principles from SDN, including the ability to provision easily (more of a benefit to the service provider in my opinion) and the ability to define policies. SD-WAN policies boil down to the ability to define path choice based on application. Voice would be a prime candidate. If you have a branch site with two WAN links, SD-WAN would let you define a policy for latency, jitter and loss. If a circuit exceeded that policy, then voice (but possibly nothing else depending on your policy) would be diverted down the other, better performing link. SD-WAN does other things as well such as load sharing WAN links etc, but other protocols can achieve the same thing and have been implemented for longer and therefore could be considered more proven. It is also commonly associated with consumption of public cloud services, but again, a cloud first strategy can be fully supported by a variety of network technologies which we have been deploying for the last 9 years.

Some vendors cite cost savings as a major benefit through the potential to use low cost Internet circuits as an alternative to private MPLS. That is mainly true for international sites where MPLS links are expensive. In the UK, the cost of a 100Mb MPLS circuit is not significantly more expensive than a 100Mb Internet circuit, so the cost savings don’t always stack up. It depends on the network distribution and requirements. So, we have the fundamental benefit of SD-WAN being the ability to have path choice for an application based on latency, jitter or loss. These are criteria which standard routing protocols don’t necessarily react to. There is a real benefit to this ability to use the best performing path in a WAN but again, the best value comes for international sites.

International connectivity can be prone to carrier routing changes causing latency to vary considerably. In the UK though that is less of an issue. For UK based correctly sized networks, the risk is a network brownout. A period of packet loss, of only a few percent, can significantly impact application performance, but traditional routing metrics won’t recognise or respond to it. ‘So what?’ I hear you ask. Well, SD-WAN gives us path choice based on a policy. It’s particularly relevant for international sites or if you are concerned about having a brownout (no one wants a brownout) it has some similarities to SDN, but the two aren’t the same and one doesn’t necessarily relate to another.

SDN gives us policy across the entire network and for my money, it’s the way enterprise customers need to go. It may still be in an early adoption phase right now, but it’s coming fast. Any investment in the network for an enterprise, whether it’s SD-WAN or not should be considered in the light of an SDN strategy over the medium term. Making the wrong choice now could be an expensive mistake.

To discover more about ANS’ connectivity offerings here.