cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Bookmark
|
Subscribe
|
2109
Views
0
Helpful
6
Replies

srr-queue bandwitdh and priority queue out configurations on user port but not on Trunk link

CarloSalvador
Level 1
Level 1

Hi Guys,

 

We have this configured on our User ports on Access Switch.

!
interface GigabitEthernet1/0/9
description VoiceAndData Port
switchport access vlan 80
switchport mode access
switchport voice vlan 81
srr-queue bandwidth share 1 30 35 5
priority-queue out
mls qos trust cos
auto qos trust
spanning-tree portfast edge
spanning-tree bpduguard enable
end


But on the Trunk links of the same Access Switch we don't.

!
interface TenGigabitEthernet1/0/1
description -Link to Core-
switchport trunk allowed vlan 80-82,84,88,89
switchport trunk native vlan 99
switchport mode trunk
channel-group 2 mode active
end

!
interface TenGigabitEthernet5/0/1
description -Link to Core-
switchport trunk allowed vlan 80-82,84,88,89
switchport trunk native vlan 99
switchport mode trunk
channel-group 2 mode active
end
!
interface Port-channel2
description -Link to Core-
switchport trunk allowed vlan 80-82,84,88,89
switchport trunk native vlan 99
switchport mode trunk
end


My question is.... Do we need to remove the srr-queue command and the priority queue out from the user port since it is not configured on the Trunk links?
1 Accepted Solution

Accepted Solutions

"With regards to your last response, are you suggesting to remove the these commands from all the switchports (Trunks and User ports)?"

No, I'm not.

I've only saying the 3750 series is often considered not to have a lot of buffer resources, and QoS, on those switches, if not optimal for your needs, can make the situation worse.  Conversely, though, optimal QoS, for your needs, can both better manage buffer resources and manage bandwidth and/or drops to meet different traffic application needs.

View solution in original post

6 Replies 6

Joseph W. Doherty
Hall of Fame
Hall of Fame

"Do we need to remove the srr-queue command and the priority queue out from the user port since it is not configured on the Trunk links?"

Actually, more likely the converse is true.  I.e. if using QoS on edge ports, likely should use in on uplink ports too.

My "guess" is, whoever set up QoS on the edge ports thought that they needed it more than your uplink ports because the former have much less bandwidth to work with, gig vs. 20gig.  This might be true as I don't know how many edge ports your access switch has nor how busy they become.  I.e. don't know if your uplink port channel ever congests, and even if it does, congestion might be so minor it's never adverse to your traffic.

Hi Joseph,

 

Thank you for your response. In that case then I will put back the configurations on the Trunks.

Furthermore, would you be able to advise about this outstanding issue I've been working on for several months now? The issue is the Video calling (Ring Central or MS Teams) are dropping when video is being in used. So far, I am seeing Output drops on that Trunk link on the Core Switch (Layer 2 Core - 3750X ) going to the Gateway (Juniper SRX Firewall).

 

We first thought that it is due to lack of bandwidth (100Mbps), however, we already did try to upgrade the Internet Bandwidth to 500Mbps but still getting this output drops.

 

 

Often two issues arise, from your last posting.

The Catalyst 3750 series is a bit "infamous" concerning port egress drops when dealing with bursty traffic.  Often worst case is with QoS enabled with default QoS settings (or other "poor" settings of same). Next best is, at least for drops, disabling QoS, altogether.  Best case, if often with "good" QoS settings.  (Drops can also be impacted by which port is being used as QoS buffers are allocated to bands of 24 copper ports and to the uplink ports, if any.)

Second, you mention "Internet Bandwidth".  That has many issues of its own, as the Internet, generally, doesn't support QoS.  Further, your mention of 500 Mbps implies, usually, that amount of bandwidth on a gig (or better) link.  If so, for QoS purposes, you generally need to "shape" your egress to the contracted amount, which a 3750 doesn't do very well.  That said, these drops on one an internal link, processing ingress traffic from the Internet and forwarding it to your FW?  If so, is more likely the possible culprit is what's mentioned in the prior paragraph.

Insufficient information, about your topology and its traffic flows, to make suggestions.

CarloSalvador
Level 1
Level 1

Hi @Joseph W. Doherty ,

 

Thank you so much for your responses. I really appreciate it much.

With regards to your last response, are you suggesting to remove the these commands from all the switchports (Trunks and User ports)?

srr-queue bandwidth share 1 30 35 5
priority-queue out
mls qos trust cos
auto qos trust

Thank you so much for your response again in advanced.

"With regards to your last response, are you suggesting to remove the these commands from all the switchports (Trunks and User ports)?"

No, I'm not.

I've only saying the 3750 series is often considered not to have a lot of buffer resources, and QoS, on those switches, if not optimal for your needs, can make the situation worse.  Conversely, though, optimal QoS, for your needs, can both better manage buffer resources and manage bandwidth and/or drops to meet different traffic application needs.

Thank you so much @Joseph W. Doherty for your effort in answering all my question! From here I can start to plan the next action from my end.