Wireless ISPs

Before Bequant we were constantly getting complaints like “my Netflix is buffering” or “I am running slow” even though the customers were receiving their full allotted bandwidth. In the last two weeks since implementing the Bequant solution we have not received a single complaint! What a relief. It has been worth every penny. It is a great solution for the shortcomings inherent in the WISP industry.

We tested the Bequant TCP Acceleration (BTA) product for a couple of weeks and the results were impressive. So good in fact, we decided to purchase four devices and are considering buying two more. BTA has definitely provided an improved streaming service experience for our customers across the board, resulting in fewer truck rolls and higher customer satisfaction.

Overview

A WISP network can be very challenging for the TCP protocol, due to the latency variability and packet losses that are typical of wireless communications. Because of those two factors, servers using TCP (which are the majority) will tend to send data to WISP subscribers more slowly than what the WISP network can actually sustain. However, when traffic flows through our BQN node, the Bequant TCP Acceleration (BTA) functionality will not get confused by the latency variability and packet losses of wireless networks, and will increase the speed experienced by WISP subscribers, utilizing the capacity that a regular TCP would have left unused. This will improve the quality of experience (QoE), the customer satisfaction and loyalty, and reduce customer support issues.

The DPI Analytics in the BQN node will also show which services are being used the most, and at what times. With that DPI information, the Bequant Bandwidth Management (BBM) functionality can identify and control the speed of some of those services, to avoid congestion situations and save in Internet connectivity costs.

Acceleration Results

This section shows the acceleration obtained in one of our WISP customers in commercial service. Our product usually leaves around 1% of TCP connections, at random, without optimization, to measure the speed of those downloads (and uploads) and to be able to compare them with the speed of the rest of TCP conenctions, which are being optimized, and to obtain the acceleration percentage (the increase in average speed). Additionaly, you can also look at our Netflix results in different kinds of ISPs.

This chart shows a comparison of the donwload mean speeds for different services.

This chart shows a comparison of the upload mean speeds for different services.

This chart shows average speeds over a week, with consistent acceleration effect in all network conditions. The image was generated in real time by the BQN online WEB interface, and provides 9 graphs for different categories depending on the download size and internet RTT.

Bandwidth Savings Results

The following graph shows the effects on the network throughput of limiting streaming video speed to 4 Mbps per user in a WISP network, identifying the video streaming traffic with our DPI Analysis functionality. The 4 Mbps limitation per user was enforced by our Bandwidth Management functionality during even hours of the day, while no limitation was set on odd hours of the day, and the total volume of traffic sent every hour, for the whole network, was measured for four days. It can be seen that the overall volume of traffic was significantly lower (around 15% lower), and that the peaks in traffic were also lower when limiting to 4 Mbps. All these results were obtained with no customer complaints due to quality.

In the previous experiment, we also measured the average latency from our equipment up to the network subscribers, and found it to be significantly lower, by around 5 milliseconds, during the even hours, when video streaming traffic had a speed limitation, as can be seen in the following graph. This is due to the fact that video stremaing tends to take all the available bandwidth, and that can fill up buffers in the network, which increases latency. By limiting video streaming total speed per user to a value below the subscribers' plan limit, we ensure those limits are not reached and some bandwidth is left for applications like online gaming or voice-over-IP, which will benefit from the latency reduction.

Deployment

Bequant software runs on Bequant Nodes (BQN), which are transparently deployed as a bump-in-the-wire between the access gateway router and the Internet gateway router.

The BQN only needs to be on the path of the traffic to optimize and it is compatible with other network elements such as traffic/QoS policers or shapers. The BQN supports PPPoE, VLAN, QinQ and MPLS traffic seamlessly.