Traffic and Request Rate Limiting

Pavonis supports two types of rate limiting: outbound traffic limiting (bandwidth control) and request rate limiting (QPS control). Both are configured globally under ResourceLimitConfig and take effect independently — either can trigger a rejection.

Outbound Traffic Limiting

Uses a token bucket algorithm to control outbound traffic rate. How it works: the bucket fills at traffic_avg_mibps tokens per second; sending 1 MiB of data consumes 1 token; excess tokens are discarded when the bucket is full, with a maximum capacity of traffic_burst_mib.

Field	Description
traffic_avg_mibps	Long-term average rate (token fill rate, in MiB/s)
traffic_burst_mib	Allowed burst capacity (bucket size, in MiB)
traffic_max_mibps	Peak rate cap (in MiB/s); unlimited if not set

resource_limit:
  traffic_avg_mibps: 100   # average 100 MiB/s
  traffic_burst_mib: 200   # allow bursting up to 200 MiB
  traffic_max_mibps: 300   # peak never exceeds 300 MiB/s

All three fields are optional. Omitting a field means no limit on that dimension.

Request Rate Limiting

Limits the total number of inbound requests globally. Supports per-second, per-minute and per-hour granularities simultaneously:

Field	Description
request_per_second	Maximum requests per second
request_per_minute	Maximum requests per minute
request_per_hour	Maximum requests per hour

resource_limit:
  request_per_second: 50
  request_per_minute: 1000
  request_per_hour: 10000

All conditions are enforced simultaneously. A request is rejected if any condition is exceeded. Omitting a field means no limit on that dimension.

Reference

• ResourceLimitConfig full documentation