Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

My ASA 5505 has three VLANs. One connected to the internet called outside, one for our office called office (which connects to the corporate VPN) and one for a publicly accessible resource-centre. Each VLAN is on a separate subnet.

I want to ensure that traffic from office to the vpn or internet has priority over traffic from the resource-centre. Put another way, I don't want the resource-centre traffic to overwhelm office traffic.

I can create ACLs for my two inside vlans:

access-list resource-centre-traffic extended permit ip 192.168.0.0 255.255.255.0 any
access-list resource-centre-traffic extended permit ip any 192.168.0.0 255.255.255.0
access-list office-traffic extended permit ip 172.16.0.0 255.255.255.0 any
access-list office-traffic extended permit ip any 172.16.0.0 255.255.255.0 

I think what I need to do is set up prioritization to traffic that matches office-traffic - that would mean that traffic to/from office would never be disrupted by traffic to/from the resource centre right?

I'm confused since I could also be setting up traffic policing on the resource-centre traffic, but I don't think that is what I want to do, as I don't actually care how much bandwidth the resource centre is using as long as it doesn't disrupt office traffic.

share|improve this question

1 Answer 1

up vote 4 down vote accepted

While I am a big fan of the ASA platform, I will be the first to admit that QoS paradigms and capabilities are rather limited on ASA's. Standard IOS ISR's run circles around what the ASA is capable of in regard to QoS.

If you have not read both the ASA QoS Configuration Guide and the IOS QoS Solutions Guide please read them. They are required reading for understanding what Cisco (and many other vendors) mean by the truly loaded term of "quality of service." Note that the IOS guide contains many features that the ASA does not support -- and gives examples that will not work on an ASA. However, both contain a wealth of useful concepts, terms, and details regarding various QoS paradigms and concepts.

With IOS your case would be quite simple -- configure the appropriate bandwidth on the outside interface, use the Modular QoS CLI to create a class and shape traffic that matches the resource-centre-traffic ACL, fair-queue the rest.

However, with an ASA that is not possible because traffic shaping is performed on all traffic on an interface on an ASA. Traffic cannot be shaped on a specific class on the ASA platform.

With shaping not being terribly useful in your case you are left with policing and priority queuing, sometimes called Low Latency Queueing (LLQ).

You have the following options

  • Police traffic matching the resource-centre-traffic ACL
  • Priority queue traffic matching the office-traffic ACL
  • Do both at the same, that is police and priority queue traffic matching the respective ACL's.

When it comes to QoS, the KISS principle still applies. The simpler the better. It is for this reason I would advise starting out minimally and fine tune. Start with policing first.

Policing

The following policing example will rate limit (police) traffic matching the resource-centre-traffic ACL to 1 Mb/s. Know that policing will drop packets that exceed the limit ultimately causing host network stacks to retransmit and back-off settling around the policed rate. Shaping avoids this by introducing latency and not dropping, but the ASA's shaper cannot shape based on classes.

! create class-map

class-map resource-centre-traffic-class
 match access-list resource-centre-traffic

! create policy-map, advise not using a global policy

policy-map outside-policy
 class resource-centre-traffic-class
  police input 1000000   ! rate in bits per second, 1 Mb/s listed
  police output 1000000  ! rate in bits per second, 1 Mb/s listed

! assign policy to interface, in this case outside

service-policy outside-policy interface outside

Priority Queuing/LLQ

Priority queuing is only applicable in an interface's transmit/out direction. It is important to configure Queue Limits and TX Ring Limit on the interface out which the traffic in the priority queue will be egressing. I am going to assume a 3 Mb/s transmit and create roughly 2 Mb/s priority queue based on 256 byte packet size. Size to use for the priority queue is a lot of the magic when it comes to LLQ. Note that any specified traffic that cannot fit into the priority queue is tail-dropped -- that is to say it is dropped -- likely not what you want for your office-traffic.

It is in this regard that priority queuing/LLQ is not generally used in high throughput traffic classes, but used for low-latency instead. However, I am including the priority queuing example here for the sake of covering what the ASA can do -- I would not recommend using priority queuing for any high throughput flows/traffic classes.

  • It is generally advisable to keep a LLQ as small as possible -- without tail dropping, as a large LLQ can starve the normal interface queue if too large.
  • Packets that qualify for the LLQ but won't fit (if it is full) are tail-dropped from the LLQ. This is shared in common with policing -- however with a LLQ traffic always "cuts to the front of the line" as the LLQ (a software queue just like the normal interface queue) is always serviced by the driver and placed in the physical interface's hardware queue (hardware ring buffer) first.

Numbers used for queue-limit and tx-ring-limit determined using the worksheet in the configuration guide and then massaged.

priority-queue outside
 queue-limit 500   ! based on factors listed earlier, very important number
 tx-ring-limit 20  ! based on factors listed earlier

! create class-map

class-map office-traffic-class
 match access-list office-traffic

! create policy-map, advise not using a global policy

policy-map outside-policy
 class office-traffic-class
  priority

! assign policy to interface, in this case outside

service-policy outside-policy interface outside
  • Both class-maps can be listed in the outside-policy for both policing and priority queuing at the same time.
  • You can still implement a global-policy and have it set as global -- as long as there are not QoS actions on any of the class's listed in the global-policy -- then the global-policy other actions (inspects, etc.) will remain in effect on the outside interface when the outside-policy (with only QoS actions) is placed on the outside interface.

TL;DR

ASA is really limited on the QoS front. Try policing. If that doesn't work add in or try LLQ very carefully. If that doesn't work look for an IOS ISR[G2] with shaping, CBWFQ, etc.

share|improve this answer
    
This man speaks the truth –  tacos_tacos_tacos Dec 18 '11 at 17:56

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.