Our Juniper EX switches are configured for sFlow sampling, and send this data to an NPM instance running NTA 4.1.0. This works fine in principle, but I wonder if the data is correct:
Juniper switches use an adaptive sampling rate, per interface, depending on interface load. In NTA, under NetFlow Settings -> Manage NetFlow Sources, it seems that Orion assumes a global sampling rate for each system.
Is that just a display problem, and NTA correctly calculates with current rate for each sFlow sample, or does it use the auto-detected sample rate for all data for that switch?
Example from show sflow interfaces on a Juniper switch:
Interface Status | Sample rate | Adapted sample | rate | Polling interval | ||
Egress Ingress | Egress | Ingress | Egress | Ingress | ||
xe-0/0/0.0 | Enabled Enabled | 5000 | 5000 | 80000 | 80000 | 30 |
xe-0/0/1.0 | Enabled Enabled | 5000 | 5000 | 40000 | 40000 | 30 |
xe-0/0/10.0 | Enabled Enabled | 5000 | 5000 | 160000 | 160000 | 30 |
xe-0/0/11.0 | Enabled Enabled | 5000 | 5000 | 80000 | 80000 | 30 |
xe-0/0/12.0 | Enabled Enabled | 5000 | 5000 | 160000 | 160000 | 30 |
xe-0/0/35.0 | Enabled Enabled | 5000 | 5000 | 320000 | 320000 | 30 |
At the same time, NTA shows this (when I expand one of the systems, I see the above sFlow interfaces, but no sample rate attached to the interfaces themselves: