It’s a unit named after a Danish mathematician. It has no physical mass. Yet it dictates the flow of billions of conversations every day.
The erlang (symbol: E) is the standard metric for measuring telephone traffic intensity. It doesn’t just count calls. It measures the average number of simultaneous circuits occupied over a specific period.
If you’ve ever sat through a hold queue, you experienced the limits of erlang calculations.
Why the Erlang Matters in Network Planning
Before A.K. Erlang, phone network design was a guessing game. Engineers relied on intuition. They guessed how many lines a central office needed based on rough estimates of subscriber demand. It was messy. Inefficient.
Agner Krarup Erlang changed that. Working for the Copenhagen Telephone Company in the early 20th century, he applied rigorous mathematics to telecom systems. He wasn’t just solving a puzzle. He was trying to prevent call drops.
His work birthed the “theory of traffic.” This framework allowed for standardized measurement. Suddenly, network planners could compare data across different systems. They could predict capacity needs with precision.
“An erlang corresponds precisely to one connection, or telephone circuit, used continuously for one hour.”
Though not part of the International System of Units (SI), the erlang remains the backbone of telecommunications planning. It translates chaotic human behavior into objective data.
How to Calculate Traffic in Erlangs
The math behind the erlang is deceptively simple. It is a dimensionless quantity. You get it by multiplying the average number of calls by the average call duration. Then divide by the observation period.
Usually, that period is one hour.
Consider a call center scenario. If twenty three-minute calls are made in one hour, the calculation looks like this:
(20 calls × 3 minutes) / 60 minutes = 1 erlang.
A single circuit occupied for the entire hour equals 1 E. If that same circuit is only busy for half the hour, that’s 0.5 erlang. This normalization allows engineers to evaluate performance and infrastructure costs objectively.
It works regardless of the underlying technology. Whether it’s analog landlines, digital networks, or modern IP-based VoIP platforms, the erlang model holds up. It provides a common language for capacity planning.
Underestimating traffic leads to saturation. Calls get blocked. Customers get angry. Overestimating it wastes resources. Money leaks out of the system. The erlang helps find the sweet spot.
Advanced Models: Erlang B vs. Erlang C
For basic traffic, the general formula suffices. But modern networks require more nuance. Engineers use specific formulas derived from Erlang’s original work.
Two models dominate the industry: Erlang B and Erlang C.
Erlang B calculates the probability that all circuits are busy. When all lines are taken, a new call is blocked. This is critical for voice networks where waiting isn’t an option.
Erlang C takes a different approach. It estimates the wait time for calls in a queue. This is essential for contact centers. Here, calls aren’t blocked. They wait. Erlang C helps managers determine how many agents are needed to handle the load without leaving customers on hold for minutes.
These models rely on statistical assumptions. They form the bedrock of modern network management. Without them, optimizing large-scale telecom infrastructure would be nearly impossible.
The numbers tell a story. A story of efficiency, waste, and the delicate balance of connectivity.
You might wonder if AI will eventually replace these statistical models. Probably not. The underlying human behavior—the unpredictability of when someone decides to make a call—remains constant. The math just gets better at predicting it.
But as networks evolve into complex, hybrid systems, the core principle remains. Measuring the load. Respecting the limit. Keeping the lines open.
You make a call. It connects. If it didn’t, you’d notice. You wouldn’t be reading this.
But that seamless connection isn’t magic. It’s math. Specifically, it’s the legacy of Agner Krarup Erlang, a Danish mathematician who figured out how to measure traffic in a way that keeps billions of devices from collapsing under their own weight. We use the erlang unit every single day, often without knowing the term.
In telecom, capacity planning isn’t a guess. It’s a calculation. And if you get it wrong, people can’t reach each other.
Sizing the Pipe for Peak Hours
Operators don’t just throw more hardware at a problem when things get busy. They predict.
The erlang unit measures traffic intensity. One erlang equals one continuous channel occupied for one hour. It’s a measure of load, not time.
When engineers design a local switch, they don’t ask “how many users might call?” They ask “what is the peak traffic volume in erlangs?” This number dictates the physical infrastructure. How many circuits do you need in the switch? How much bandwidth do you allocate for data transmission?
In mobile networks, it’s even more granular. Cell towers serve specific geographic zones called cells. During rush hour, or at a massive concert, the load spikes. Operators use Erlang’s formulas to model this congestion. They anticipate where the pressure will hit hardest. Then, they adjust.
This isn’t just about voice anymore. The convergence of fixed and mobile networks means data, video, and voice share the same pipes. The concept has expanded. You’re not just measuring talk time. You’re measuring synchronous communication load. Whether it’s a VoIP call or a video conference, the traffic behaves similarly. It consumes resources. It blocks others if the pipe is full.
Call centers rely on this too. Emergency services. Customer support. They use queueing theory derived from Erlang’s work to staff their lines. Too few agents? Wait times explode. Too many? Money burns. The goal is efficiency. The metric is the erlang unit.
Beyond Telephony: A Universal Metric
Erlang’s influence stretches far beyond telephone exchanges. His approach combined rigorous probability with practical engineering. That combination is rare.
Today, you see his methods in computer science curricula. In applied math. In logistics.
When a company needs to optimize shared resources, they look at queueing systems. Hospitals use similar models to manage patient flow. Server farms use them to allocate CPU cycles. The erlang unit provides a unified language for capacity. It allows engineers to compare different types of traffic on a common scale.
This flexibility is why the theory survived the digital revolution. When networks went digital, the underlying math didn’t break. It adapted. Virtualization made infrastructure fluid. Cloud computing scaled resources up and down. The need to measure load remained constant.
Erlang’s contribution shifted from voice-centric to service-agnostic. We now use these principles to manage video streaming loads. To distribute computing tasks across servers. To balance bandwidth for real-time applications. The core idea is identical: predict the peak, build the capacity, avoid the block.
The Cost of Ignoring the Math
Understanding traffic in erlangs isn’t academic. It’s economic. It’s existential for connectivity.
If an operator fails to model traffic accurately, the result is dropped calls. Slow data. Frustrated users. In a competitive market, that’s business suicide. Users don’t tolerate poor service. They switch providers.
But it’s also about resilience. Our society depends on constant connectivity. Power grids. Financial systems. Emergency response. All of it relies on robust communication networks. Planning capacity using erlang unit metrics contributes directly to this resilience. It ensures the system doesn’t just survive the peak, but thrives.
The tension between investment and performance is constant. Operators must justify capital expenditure. They can’t build for the absolute maximum theoretical load every day. That would be wasteful. But they can’t build for the average load either. That would cause daily failures. They build for the statistical probability of congestion. That’s where Erlang’s formulas shine. They provide the probability of blocking given a certain traffic volume and number of servers. It’s a precise trade-off.
A Living Legacy
The term “erlang” is more than a historical footnote. It’s a living tool.
As global demand for connectivity grows, the complexity of networks increases. 5G. IoT devices. Smart cities. The variables are multiplying. But the fundamental challenge remains the same. How do you ensure quality of service when demand is unpredictable?
The answer lies in the science Erlang established. It’s taught in engineering schools. It’s implemented in network software. It’s the backbone of internet server sizing.
The erlang unit bridges the gap between theoretical probability and physical reality. It turns chaos into manageable data. It allows engineers to design systems that scale.
We are entering an era of hyper-connectivity. The stakes are higher. The margins for error are thinner. But the tools are still rooted in that early 20th-century work.
Erlang’s legacy isn’t just in the past. It’s in every connection that holds. Every call that connects. Every second of uninterrupted service.
The math still works. The network still needs it. And we still take it for granted until it fails.
Sources like the CNRS highlight the ongoing synergy between fundamental research and industrial application in electronics and future telecommunications. The collaboration between scientific communities ensures these old models remain relevant. As technology evolves, the need for rigorous traffic modeling only intensifies.
The numbers don’t lie. The connections work. Because the math holds.
