Forum

When designing a WiFi deployment that supports applications that are sensitve to latency & jitter (e.g. voice and video), although some bandwidth requirements can be understood by perhaps testing the apps over an AP as part of the survey, how do you get a handle on the effects of increasing latency and jitter as more wireless devices use an AP?

As far as I understand it, WiFi QOS techniques provide a good level of statistical probability that a more frequent level of access to the wireless medium is provided to higher priority apps (that are latency-sensitive). But, there must come a tipping point when the latency effects of servicing other devices on the same AP will create issues for these apps - I suspect that this could be reached well before the bandwidth capacity of a well-designed 802.11n WLAN?

Testing this effect sounds like it could be quite a challenge, as it could require significant amounts of equipment to verify.

Am I looking at this issue in the correct way? Is it really an issue? Is there a way of testing, verifying or predicting this in a better way than by trying to simulate with piles of laptops passing traffic until a service affecting threshold is hit?

I am particularly interested as I am involved with a project that requires users to use (amongst other apps) desktop video conferencing. Although capturing and verifying the data rates required by the apps seems to be relatively straightforward, I'm not sure how I can predict the capacity of the network to handle the latency-sensitive traffic before its effects will start to degrade voice and video streams as more devices join an AP.

I've looked as CAC solutions, but they seem to rely on restricting bandwidth at some pre-determined threshold - you would still need to know the threshold you need to stay under before hitting latency issues that would affect traffic.

Any gurus out there interested in providing me with any enlightenment? Am I maybe over-thinking this or missing an obvious technique or solution...? :)

Thanks

Nigel.

Hi Nigel,

In my experience of deploying wireless infra that's capable of handling the triple play I really depend heavily on the site surveys. Understanding the rf and physical environment made it easy for me to do a benchmarking.

I used AirMagnet Survey Pro and Planner for the predictive and actual surveys. I used IxChariot for traffic simulations. I know these softwares are expensive. I was just very lucky that my work place have those things.

Forget about the QoS and HTs first. It is more important addressing the obvious challenges before going into the deep.

Hi Nigel,

Cool post.

We use satellite simulators quite a bit to test the effect of a satellite backhaul. NetEm/ WanEm and others that you can get for free will run on an (also free) LINUX platform. You do not have to use pre-set latency or jitter thresholds and can simulate any you want and in a few different ways, it's very useful.

The problem of scaling is much harder though imo. I agree with Art that Chariot works for that, although like he said it is expensive, and we have run into a few problems with it when attempting to test scaling, since real-life would involve asynchronous activities, etc...there are a few challenges.

Another option that may (?) be helpful is setting up a specialized VPN apps for users using those sensitive apps. That seems counter-intutive since it would add overhead, but some solutions do some tricky stuff with the packaging that actually ends up accellerating and improving performance. (Not making this up, we have run many side-by-side tests, repeatedly, in many architectures, with many types of end-devices.) The good ones do cost, but may be worth it depending on what you're trying to do. Also many of the VTC apps will allow you to fiddle with the settings to find maximum performance, and some do it for you.

Are you also speaking to the effect of wifi roaming on your sensitive apps? If so, have you looked into 802.11r (FT) /802.11k (RRM) capable devices? I have no practical experience with this but am currently studying it in my CWSP book. ;) (Check the whitepaper too.) Since I'm just getting immersed in that now I can't offer any insight this area, but if you find or experience any nuggets with these new (and the emerging) VoWiFi standards please share them!

Cheers,
Tracy