Enforcing Subscriber Plan Speeds and Optimizing QoE
Preseem is a Quality of Experience (QoE) monitoring and optimization platform developed for WISPs. This blog post highlights how traditional plan enforcement tools like Mikrotik simple queues or Access Points (APs) can negatively impact subscriber QoE.
There are many different ways to enforce subscriber plan speeds. Some common options for WISPs include Mikrotik Simple Queues, Access Point or CPE shapers. All of these do a good job at making sure that subscribers are getting their contracted plan speeds.
The problem, however, is how these systems achieve subscriber plan limits and what happens to the QoE for subscribers once the plan speed is hit.
How Traditional Shapers Enforce Plans
Traditional shapers have a single buffer for each subscriber’s traffic. Incoming packets go into the buffer and it releases 5 Mb/s of data (for example) on the other end. If more comes in than goes out, the buffer starts filling up. As the buffer fills up, it adds latency to all the traffic in the home.
In this example, from a WISP perspective, the subscriber plan speeds were enforced. However, from the subscriber’s point of view, QoE was bad and this generally leads to support calls or churn.
With traditional shapers, as soon as one internet application maxes out the connection for a home, every other application in the household becomes slow.
How Preseem Handles Plan Enforcement
Preseem solves subscriber plan enforcement problems associated with traditional shaping tools in two ways:
1. Instead of using a simple buffer that fills up as packets come in, Preseem uses Active Queue Management (AQM). AQM knows exactly how many packets to buffer or drop, and applies TCP signaling to get heavy traffic to “back off.” The net effect is that instead of adding 150-200 ms of latency, Preseem adds less than 10 ms when it kicks in. Even if a subscriber is doing a big download like an operating system update, the latency with Preseem stays low. This ensures that the overall QoE is high, meaning far fewer support calls!
2. Preseem also uses fair queuing to automatically distribute the bandwidth fairly between different applications. The platform automatically figures out if an application qualifies as a bulk flow or an interactive flow. This is completely based on the size of the flow, i.e. how many bytes that flow is transferring. If it’s large, that’s a bulk flow. If small, it’s considered an interactive flow. For every bulk flow, Preseem automatically creates one queue to manage that flow independently. Then, it automatically balances or fairly divides traffic between those bulk applications. So, Netflix and a download for example will both get half the traffic. But any interactive flows like gaming or VoIP calls don’t get queued at all. These smaller interactive flows get prioritized straight through
So, no matter how busy a link is due to bulk applications like Netflix, interactive applications will always “feel fast,” thanks to Preseem’s advanced AQM techniques. This leads to a much better experience for subscribers.
Causes of Poor Subscriber QoE
Some of the situations WISPs shared with us before Preseem was deployed, include:
- One household member watching Netflix leading to a poor experience for another member playing games online.
- When downloading a new OS update, all other applications in the house felt slow, leading to high support costs.
- Background downloads by heavy applications affected the overall internet experience, leading to frequent complaints.
Users in all these situations reported better QoE after Preseem replaced traditional shaping and plan enforcement methods across WISP networks.
A Preseem Case Study
The above shows the latency and bitrate at a WISP with more than 2,500 subscribers based in Amarillo, TX. In this network, the plan rate was enforced at the AP. Very similar results are found with Mikrotik-based shaping.
Once the plan speed of 3 Mb/s was hit, latency for the subscriber shot over 200 ms. A quarter of that was the AP itself (base latency), with at least 150 ms being the buffer for the traffic management. As soon as Preseem took over the shaping for this network, total latency came down to around 55 ms, with Preseem adding fewer than 10 ms of latency. This lower latency is a result of Preseem’s use of AQM techniques, based on FQ-CoDel, which ensures that the overall QoE for a subscriber remains high despite a maxed-out connection/link.
To find out how your traffic management and subscriber plan enforcement tools impact subscriber QoE and how you can fix this problem, contact us for a free 30-day trial of Preseem.