Hardware Sizing
Sizing is driven by two values: the number of agents classifying concurrently at peak load, and the total number of idle or connected agents. Select the row that covers your peak concurrent
classification count. If you are unsure, assume 20-30% of your total agent population will be active at peak.
| Concurrent classifying agents (peak) | Idle/connected agents | vCPU | RAM |
|---|---|---|---|
| Up to 80 | Up to 22,000 | 24 | 96 GB |
| Up to 110 | Up to 30,000 | 32 | 124 GB |
| Up to 160 | Up to 45,000 | 48 | 192 GB |
| Up to 400 | Up to 110,000 | 120 | 480 GB |
Sizing assumptions:
-
1 classification per 50 seconds per agent at peak (~0.02 RPS / agent)
-
Latency SLO: 95% of classifications complete in under 5 seconds
-
Maximum of 5 DLP policies enabled
-
'Concurrent classifying' refers to the volume of active classification occurring at peak times, distinct from total licensed seats. Typical office workloads see 20-30% peak concurrency among licensed agents.
-
Heartbeats configured at 60 seconds
-
Larger deployments scale by adding clusters; each cluster handles the per-row agent count.
Note: The above values are baseline minimums. Review capacity utilization every six months and adjust as data volumes, policy counts, and endpoint numbers scale. Contact Forcepoint support before
making changes if you need guidance.