Skip to content
Calcrivo

Load Balancer Capacity Calculator

Size a load balancer pair from connection arrival rate and hold time, and check the state table against the per-node limit.

Inputs

cps
s
nodes
connections
Mbps

Concurrent Connections

360,000

Connections per Active Node

360,000

Connection Table Utilisation

72.0%

Spare Table Entries

140,000

Bandwidth per Connection

11.11kbps

Assessment

Sized with limited headroom

Step by step

  1. Values used

    New connections per second = 8,000 cps; Average connection hold time = 45 s; Load balancer nodes in the cluster = 2 nodes; Active/standby (only one node carries traffic) = Yes; Connection table limit per node = 500,000 connections; Aggregate throughput = 4,000 Mbps

  2. Load Balancer Capacity

    concurrent connections = new connections per second × average hold time (Erlang-style concurrency); utilisation = concurrent ÷ per-node table limit.

  3. Concurrent Connections

    = 360,000

  4. Connections per Active Node

    = 360,000

  5. Connection Table Utilisation

    = 72.0

  6. Spare Table Entries

    = 140,000

  7. Bandwidth per Connection

    = 11.11 kbps

  8. Assessment

    = Sized with limited headroom

How it works

A connection table holds one entry per live flow, and the number of live flows is the arrival rate multiplied by how long each one lasts. In active/standby only one node carries the load, so the surviving node must fit the entire table on its own. Load balancers are almost never limited by bandwidth — they run out of connection-table entries first, and an active/active cluster sized at 80% cannot survive losing a node.

Formula

Load Balancer Capacity

concurrent connections = new connections per second × average hold time (Erlang-style concurrency); utilisation = concurrent ÷ per-node table limit.

holdTime
How long a connection occupies a table entry, including idle timeout
activeNodes
1 in active/standby, the full cluster in active/active

Frequently Asked Questions

How is Load Balancer Capacity calculated?

concurrent connections = new connections per second × average hold time (Erlang-style concurrency); utilisation = concurrent ÷ per-node table limit. A connection table holds one entry per live flow, and the number of live flows is the arrival rate multiplied by how long each one lasts. In active/standby only one node carries the load, so the surviving node must fit the entire table on its own.

Why does Load Balancer Capacity matter?

Load balancers are almost never limited by bandwidth — they run out of connection-table entries first, and an active/active cluster sized at 80% cannot survive losing a node.

What values do I need to enter?

This calculator takes 6 inputs: New connections per second, Average connection hold time, Load balancer nodes in the cluster, Active/standby (only one node carries traffic), Connection table limit per node, Aggregate throughput. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.

You might also need