I have been having a lot of problems coming in on an AT&T circuit at a remote data center. The issue at hand is that we can transmit up to 35GB of data at the speed we expect before the bandwidth seems to completely die off to the point we are getting almost a 1/10th what we were getting for the first 35GB. After troubleshooting with AT&T I was sent the following email:
It has been identified that you are dropping a lot of packets due to their being marked as COS1 (Real Time) and COS2v (Interactive) while your PPCOS (Per Packet Class of Service) profile does not allow for either. Your PPCOS profile allows for 50% of your port speed (CIR) to be allocated to COS3 (p-bit marked as: 2) and 50%-100% as COS4 (p-bit marked as: 0).
If you are marking your 802.1p fields of your packets with either a 5 (Real Time) or a 4 (Interactive), this is likely causing the problem. If you can change those designations to 2’s and 0’s that should care for the performance issue. Note: 802.1p bit fields that contain no COS designation marking are treated as COS4 in our network.
Let me know if this helps.
I did what I think makes this change. I ran the command
DR-1(eth-12)# bandwidth-min output 0 0 50 50 0 0 0 0
and this produces the following output show bandwidth output command:
DR-1(eth-12)# show bandwidth output 12
Outbound Guaranteed Minimum Bandwidth Percentage
Port Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8
------ ------ ------ ------ ------ ------ ------ ------ ------
12 0 0 50 50 0 0 0 0
is this the right course of action?
the switch in question here is a Aruba 3810m and upstream port is a Gigabit ethernet port.
Any help at all would be greatly appreciated.
No comments:
Post a Comment