Thank you both for your replies.
I do not expect the RAP needs that much bandwidth to communicate with controller while idling (I’ll call it keep-alive, but I am not sure if it is the correct terminology)
Let’s exam the data path of three RAP-155s, which have the same data consumption rate: RAP-1 is having active client while RAP-2 and RAP-3 are idling. It looks like in RAP-1 tunnel 17 is keep-alive while in RAP-2 and RAP-3 are tunnel 50 and tunnel 28 respectively. We can see the large number of packets in all three RAPs for keep-alive
Note that all RAPs are working normal
(WC03) # show datapath session table | include 10.129.254.84
10.129.254.84 172.18.254.96 47 0 0 0/0 0 46 0 tunnel 17 469f 16903 3873876 FC
10.129.254.84 172.18.254.96 17 8211 8419 0/0 0 0 1 local 8 0 0 FYI
10.129.254.84 172.18.254.96 17 8209 8419 0/0 0 0 1 tunnel 82 6 0 0 FYCI
10.129.254.84 172.18.254.96 17 8209 8421 0/0 0 0 0 tunnel 82 2 0 0 FYCI
10.129.254.84 172.18.254.96 17 8211 8224 0/0 0 0 0 local 2 0 0 FYI
10.129.254.84 172.18.254.96 17 8209 8209 0/0 0 0 0 tunnel 82 1c 6 5302 FCI
172.18.254.96 10.129.254.84 47 0 0 0/0 7 24 0 tunnel 17 46a0 8924852 948888740 F
172.18.254.96 10.129.254.84 17 8421 8209 0/0 0 0 0 tunnel 82 3 0 0 FYI
172.18.254.96 10.129.254.84 17 8419 8211 0/0 0 0 0 local 9 1 161 FCI
172.18.254.96 10.129.254.84 17 8419 8209 0/0 0 0 0 tunnel 82 7 0 0 FYI
172.18.254.96 10.129.254.84 17 8224 8211 0/0 0 0 0 local 3 2 1032 FCI
172.18.254.96 10.129.254.84 17 8209 8209 0/0 0 0 1 tunnel 82 1d 0 0 FYI
(WC03) # show datapath session table | include 10.129.254.76
10.129.254.76 172.18.254.96 47 0 0 0/0 0 48 0 tunnel 50 39ae 197343 75126093 FC
172.18.254.96 10.129.254.76 47 0 0 0/0 7 24 0 tunnel 50 39af 12740453 1405272065 F
(WC03) # show datapath session table | include 10.129.254.83
10.129.254.83 172.18.254.96 47 0 0 0/0 0 46 0 tunnel 28 77f3 24116 4901394 FC
172.18.254.96 10.129.254.83 47 0 0 0/0 7 24 0 tunnel 28 77f4 16180546 1713550184 F
(WC03) #