Consistent vs. Best Effort: Building a Predictable Enterprise Network

For decades, one of the primary distinctions between MPLS and internet-based connectivity was guaranteed latency. Why is this guarantee so important and why do you need carrier provided MPLS-service to get it?

Latency is the time it takes for a packet to travel between two locations. The absolute minimum is the time it would take for light to go across the distance. The practical limit is higher (roughly 1ms round trip time for every 60 miles or so). There are two elements that impact latency: packet loss and traffic routing.

Packet Loss

Packet loss often occurs in the handoff between carriers. In MPLS-networks handoffs are eliminated or minimized thus reducing packet loss. Internet routes can go through many carriers increasing the likelihood of packet loss. A packet loss requires retransmission of lost packets and has a dramatic impact on overall response time. Advanced networks apply Forward Error Correction (FEC) to predict and correct packet loss without the need for retransmission.

Traffic Routing

IP routing is much harder to control. It is determined by a service provider based on a complex set of business and technical constraints. Within an MPLS network, the provider controls the route between any two sites, end-to-end, and can provide an SLA for the associated latency. Public Internet routing crosses multiple service providers which are uncoordinated by design, so no SLA can be provided.

A case in point

Cato Research Labs recently did latency tests between three IP addresses in Dubai and an IP in Mumbai. These three IP addresses are served by three separate ISPs in Dubai. The round trip time (RTT) was 37ms, 92ms, and 216ms, respectively. In this case, the choice of local ISP could drastically impact the customer experience. What determined this outcome was the internet routes used by each ISP.

While public internet routes have no SLAs, customers can do a “spot check” and get pretty good latency between two locations over the internet. However, the latency observed during the spot check, is not guaranteed. In fact, latency can vary greatly, what we call jitter, day to day and month to month, depending on the providers that get involved and various conditions on the internet.

Cato provides a guaranteed, SLA-backed latency. Our cloud network relies on tier-1 carrier links and has self-optimization capabilities to dynamically identify optimal routes and eliminate packet loss. Regardless of how we compare with the various connectivity alternatives at any given point in time, the customer is assured that latency will not exceed our SLA. This is especially critical for long haul routes across America, Europe, and Asia but also for latency-sensitive applications like voice and video. At Cato, we strive for consistency of network latency at an affordable cost, rather than making customers pay the “MPLS premium” for the absolute lowest latency figures, or relying on a best effort, inconsistent, transport like the public internet.   

Cato Networks provides an affordable, low jitter and low latency enterprise network. To know more about the trade-offs for your next generation WAN, download our “MPLS, SD-WAN, Internet, and Cloud Networks” ebook here.