What "most concurrent agents" tells you about server sizing

The most-concurrent-agents figure in Maximum System Stats is the highest number of agents logged into the system at the same moment on a given day — and it is the metric that determines whether your server and agent licensing can handle the real-world peak, not just the headcount on paper.

Why peak simultaneous agents is the right measure

Most call centers staff multiple shifts and not every Agent logs in at the same time. Your total hired headcount might be 60 agents across three shifts while your peak simultaneous count is 28. The server only needs to handle the 28 at once. If you size infrastructure for 60 you are over-spending; if you forget the overlap period between shift handoffs where 35 agents are simultaneously logged in, you may hit a constraint you did not anticipate.

The most-concurrent-agents figure captures that real overlap peak. It is not a sum, not an average — it is the single worst-case simultaneous agent session count the server had to support on that calendar day.

What the agent peak actually sizes

Each simultaneous agent session places load on the server in several ways:

• One active Agent session means at least one open SIP registration consuming a SIP peer slot and a keepalive timer in Asterisk.
• Each logged-in agent has a running process in VICIdial's keepalive system monitoring that session.
• Each active agent on a call adds at least one Channel to the Asterisk channel pool, plus the far-end trunk leg.
• Agent licensing in some VICIdial configurations is enforced at the simultaneous-session level, so exceeding the licensed peak blocks new logins rather than issuing a warning.

stateDiagram-v2
  [*] --> Safe : peak below 70pct of capacity
  Safe --> Warning : peak reaches 70 to 85pct
  Warning --> Critical : peak reaches 85 to 100pct
  Critical --> Breach : peak exceeds capacity
  Breach --> Critical : resize or reduce agents
  Critical --> Warning : headroom restored
  Warning --> Safe : growth slows

Reading the peak against your seat count

Open the 30-day table and compare the most-concurrent-agents column against the agent capacity you believe you have. A healthy operation keeps its peak at around 70 to 80 percent of that ceiling on normal days, with room to absorb the occasional all-hands-on-deck situation.

If the peak is consistently at or above 90 percent, you are one shift-overlap away from a login block or a degraded Agent utilization situation where agents wait for the system to respond. If the peak is consistently below 40 percent, you may be over-provisioned and paying for server capacity you do not use.

When the most-concurrent-agents figure starts climbing month over month, it tells you staffing is growing even if no single shift feels dramatically busier. The peak is how growth shows up in the data before it becomes a problem.

Combining agent peak with call peak

The most-concurrent-agents figure sits in the same report as the most-concurrent-calls figures, and reading them together tells you about Lines per agent efficiency. If your peak agent count is 25 and your peak combined call count is 90, your dialer was maintaining roughly 3.6 simultaneous calls per agent at the busiest moment. That is a useful sanity check on whether your Predictive dialing ratio is set appropriately.

If that ratio is significantly higher than expected — say, six simultaneous calls per logged-in agent — it can mean the dialer is launching calls faster than agents are becoming available, which drives up Drop rate and pushes you toward compliance thresholds. If the ratio is lower than expected, the campaign's Auto dial level may be set too conservatively and agents are sitting idle between calls more than necessary.

The most-concurrent-agents figure also gives you a grounding point when you are evaluating whether a reported problem was a software issue or a load issue. If an agent complained that the system was sluggish at 2pm and the most-concurrent-agents column shows a new daily high at that hour, load is the first suspect. If the agent count was normal and the call peak was normal, look at software — a runaway process, a slow query, or a network event.

For the full monitoring framework see our guide to monitoring VICIdial server health and capacity. To understand how load behaves under these agent counts in real time, read VICIdial system load explained.

Want a fully monitored VICIdial box sized to your actual concurrent-agent peak? Start a VICIfast trial — your server is live in under 40 seconds with Maximum System Stats collecting from the first boot.