Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations strongm on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Cisco QoS for Speed Testing Sites

Status
Not open for further replies.

jdtp

ISP
Feb 27, 2012
2
US
I have a 3662 router with IOS 12.3(23), and multiple users on the LAN side FE0/0 with a 60M connection to the FE0/1 (WAN) side. I want to implement some Qos or policy maps to make sure when someone runs a speed test, they get the appropriately configured speed there account is set for. I'm struggling with identifying a way to flag the speedtest traffic w/o knowing the specific endpoint (speedtest server IP) that a user would be connecting to.

Does anyone have a suggestion for this process?

Thank you,
Jeff
 
Hi,

just thinking out loud here as its an interesting question. If the users on the LAN side have a specific SLA. i.e 2MB bandwidth etc can you not Shape the traffic from the specific users not to ever go above 2MB on the WAN side? So any traffic inc speed tests should never exceed that? but something is niggling me that the shaping will not kick in unless the output queue is full on your WAN interface.


Lee.

LEEroy
CCNP,CCIP
 
For downloading you would be policing on the input of the WAN interface (since you cannot shape on an input policy), the policy-map would trigger on whatever amount you specified.

But by that point it has already traversed your connection...and you're technically already using the bandwidth from your provider's perspective.
 
Thank you for the responses.

Lee, yes, and in actuality, they are limited to a specific upload/download bandwidth at the end users location via a cable modem and (lets say in this case) they all have a 2MB download and upload SLA. So I have 100 users with 2MB SLA, and of course they aren't generally using 2MB all the time, but what I'm trying to prevent is during a high usage time, where the 60MB WAN connection is being maxed out, that the users running a speed test to any of the random speed test sites on the internet, will give them the correct results of 2MB download etc.

dgrizzard, you are correct. I can do that, but more so at the moment, is how I can identify the traffic as "Speed Test" traffic of which then I can apply some sort of policy to it and prioritize it so that the router will handle it first. Maybe that isn't the solution in this case. Maybe I will have to reserver some specific amount of bandwidth thats just available to specific "speed test" traffic.

I appreciate the input and hope for more. I've implemented a policy map and applied it to the LAN interface, for traffic coming in from the LAN side and applied the same policy map to the WAN interface also applying it for traffic coming in to the WAN interface from outside. I then have it setting the precedent to 7 making it the highest priority traffic in the queue.

class-map match-all spd-test
description Match all speed test site traffic
match protocol http host "*speedtest*"
!
policy-map spd-test
class spd-test
set precedence 7
!
Interface Fastethernet 0/0 (LAN)
service-policy input spd-test
!
Interface Fastethernet 0/1 (WAN)
service-policy input spd-test

Specifically, I did a capture on a speedtest being run and identified that the hostname generally had "speedtest" in it and thought that its possibly to identify the traffic that way, of which then applies the higher precedence to that traffic and I hoped would then provide more accurate results for a user on the LAN doing a speed test at a time when the WAN link was maxed out.

Hopefully that explains it in more detail and again, I appreciate the input. Let me know if you think the configs above would work or not. I haven't been able to test them accurately in my lab yet, so I'll let you know what comes of it if I can confirm a working solution.

Thanks!
 
I think you are still going to run into potential of some of those speed tests being "starved" out especially under congestion. There might be better ways of handling the over subscription, but would you really want to put a speed test traffic as a higher priority over something else?

I think that approach is opening the doors for some complaints. A little mis-leading IMO that a speed test would say I'm getting a full 2mbps even though the ISP uplink to the internet is under congestion and over subscribed.

I can see policing each cable modem down to 2mbps for everything coming in, but if you are constantly under congestion, I'd say it's time to look at upgrading, or doing some digging to see why the connection is over-loaded.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top