When we say that networks are congested, we usually mean that a combination of too much data and not enough available capacity to support the applications are involved. We’ll conjure up metaphors of traffic jams, inner-city lane closures and other practical analogies to simplify the concept to users. The ubiquity of the internet these days deems these no longer necessary. Users understand bottlenecks, buffers and crowding better than ever. Users expect fast connections regardless.
Network congestion is an issue of scale. Too much data forced through a route that can’t process that data fast enough inevitably leads to delays and packet loss. To design and monitor networks intelligently, we should understand what causes congestion.
Businesses Forget About Capacity
Congestion, specifically on office networks, has some common causes. If a business is reliant on VoIP or streaming media, they will design in Quality of Service (QoS) or fatter pipes to compensate when creating their network. It’s enterprises that aren’t used to these restrictions that end up with IT complaints. Businesses with a reliance on technology, but with a culture that allows open internet traffic, run into trouble. It’s last week’s SNL skit, this week’s annual tournament (read: Masters, March Madness, Olympics), streaming music and social media that suck up more than their fair share of capacity. When dealing with capacity issues, the use of QoS can allow for prioritization of some traffic (e.g., business-critical data) over less critical or recreational data traffic.
Network Segmentation is Poorly Designed
But capacity isn’t the only cause. Poor network design can cause additional congestion. The use of subnets (or VLANS to isolate the traffic) has long added order and better performance by segmenting traffic in larger networks. When misused, subnetting may, at best, provide human-readable segmentation of a network. When used intelligently, subnetting can split up or reduce congestions at key points of your network infrastructure. By cleverly creating subnets to minimize congestion, or employing a Content Delivery Network (CDN) to serve external requests, companies can reduce the network load and route traffic more efficiently.
BGP Is Dumb ( and Outdated )
The Border Gateway Protocol (BGP) used in networking seeks the shortest path between two endpoints. It is, by design, not an intelligent system—and hasn’t been updated since 2006. In order to be fast, BGP pays no attention to the amount of data sent on network routes. This ultimately leads to congestion on some of the most common pipes in the backbone of the internet. If the shortest path from you to your customers always uses one of these common pipes, BGP will route around it only when it fails.
In practice, the combination of intelligent network design, the use of QoS segmentation and an understanding of internet routing are the keys to success against congestion. The final component is visibility. If you can’t see what’s happening on your network, then you’ll find that you’re asleep behind the wheel.*
*Apologies. I had to make at least one automotive reference.