Jul 09, 2013

This blog series looks at the success factors behind making the right choice for a wireless technology and product.  The first article of this series will tackle the biggest elephant – performance.

Performance, or more precisely ‘application performance’ is the single most important factor in choosing a wireless network. The most common pitfall that customers fall into is to evaluate a product or technology solely based on point-to-point throughput delivery. But that’s only one aspect of it. Also take into consideration what services you’re going to run on the network –WiFi, Video surveillance, VOIP calls, all of the above? Each of these adds another dimension in the evaluation process.  So let’s look at five such important factors that will earn you your next promotion.

Raw Point-to-Point wireless performance –  Quad Test Matrix

Of course you need to do this. But use this ONLY to baseline the product. Make sure that you get numbers for both TCP and UDP. To really check the resilience of the product to changing RF climate, test it under heavy and light interference and short and long distances. Ideal would be to get a quad test matrix of Distance vs. Signal Strength vs. Interference vs. Performance.

Latency –  PTZ ready

Most wireless products work on store-forward technology which introduces latency. Pick a product that suits your latency needs. If your video camera has Pan Tilt Zoom (PTZ) capability or you plan to run VOIP calls, you have strict demands for latency or else you start seeing hoppy video or dropped VOIP calls. Most wireless test tools allow you to measure round trip delay as well as delay variation. A Layer 2 based switched network will offer lower latency and delay variation than a layer 3 based routed network.

Single vs. multiple hops – potential for hidden costs

As you scale your network, be sure to determine whether the technology can allow you to scale your network. What are the fiber requirements? Technologies such as PTP and PTMP work great in single hop but they can potentially have serious performance degradation across multiple hops also could need additional fiber. This will mess up your budget and limit your scale.

Big/small packets – bigger is better?

Many products work best when you test with the largest packet size of 1518 bytes. The throughput can get reduced to as much as 20% of the claimed number once you drop the packet size to 64 bytes. Video and data use large packet sizes. But that’s not true for application control packets (keep alives) and voice – which are small 64 bytes usually. As part of the qualification process, remember to test a mix of big and small packets.

QoS – Video falling apart

Security Personnel to IT manager:  “Hey, what happened to the cameras?  They were working so well last night.”  Sound familiar? Choosing products with badly designed or non-existent QoS will hand you this Christmas present in July. Video works well by itself but when you add data/WiFi on top of it, video starts deteriorating. Test the product with video in foreground and jack up the background data traffic to see the effects on video. You’ll be surprised in most cases.

Many of these test criteria may seem obvious, but time and again we come across customers stuck with networks that fail to deliver their application needs. So next time some vendor comes to you with a performance number, make sure you ask for five.