Wireless Data: Does Speed Really Matter

from the kerton-wireless-review dept

The latest Kerton Wireless Review column (you can get these columns emailed to you by subscribing) asks the question of whether or not speed really matters when it comes to wireless data. It goes through some important misconceptions on the issues related to wireless data use, and talks about why there are different kinds of wireless data speeds that are important in different situations. Click on "read more" or "comments" below to read the full article.Wireless Data: Does Speed Really Matter?
by Derek Kerton

Speed enables richer content, more functional applications, and shorter wait times, but what speed factors are important, and how much speed is enough? It turns out that there are multiple speed factors to measure in wireless data:

1. Average Throughput: This is the speed factor we all know, a "56 Kbps" dial up modem actually connects at between 33 and 48 Kbps while a DSL modem gets 300-500 Kbps. In the wireless camp, circuit-switched (2G) cellular dial-up connections have provided between 9.6 Kbps and 14.4 Kbps for over 10 years. Last year, GPRS deployment bumped this up to 20 - 30 Kbps while 1xRTT network support 40-60 Kbps. Verizon Wireless is launching EV-DO service, which support DSL-type speeds.
2. Connection Delay: This factor, highly responsible for the pain of mobile data users, is the delay between the request for a mobile data session to the time when the data connection is established. In the circuit-switched (2G) world, these delays were between 20 and 40 seconds! In 2.5G GPRS and 1xRTT networks, carriers would have us believe that this delay has been eliminated - that our phones are 'always-on'. Close, but not true. Establishing a data session in 2.5G phones still requires some set-up time, usually between 3 and 10 seconds.
3. Network Latency: This third speed factor, occurring when customers are already connected with a wireless data session, is the network delay between when a data request is received and when a response is delivered. This delay is usually between 200 ms and 1 second.

Understanding the three types of delay, we can better analyze which has the largest impact wireless data services. In so doing, an interesting result emerges: the delay factors differ in relative importance depending on the application. For example with Push-To-Talk voice messaging services, connection delay is the most important, while for streaming media, maximum throughput is the big daddy.

Continue the analysis, assign blame: connection delay, not low throughput, caused the majority of the wireless Internet's early problems, particularly in the original launch of WAP. The WAP mobile usage pattern often calls for quick access to small bits of information - where bandwidth is less important than connection delays. Take the example of looking up a stock price or sports score during a discussion with friends: up until 2002, connection delay caused 30 sec. of delay while throughput was a non-issue - even this entire article could be delivered to a handset in 0.5 sec on a 9.6 Kbps connection, once the connection delay was past. WAP suffered because few would wait 30 seconds to connect to a tidbit of information.

RIM's Blackberry e-mail solution won hearts and minds because it hides the delay. Blackberry retrieves mail at regular intervals, so the device handles the delay and stores the mail in local memory. When the user grabs their Blackberry, they experience no delay. Great design, resulting in many Blackberry users who assume it has a fast network behind it. Actually it has been running on 9.6-19.4 Kbps data networks since 1999.

Now that 2.5G networks have reduced the connection delay, throughput rises in relative importance. But how much bandwidth is enough? While it's certain that anyone who says, "xx is enough bandwidth" is destined to be proven wrong, I will make the 'qualified' claim that in May 2003, in the USA, 50 Kbps is enough mobile throughput for mobile phones. This is because most current handsets do not have the memory, screen, interface, or processor to exploit any more bandwidth.

We are at a critical juncture in the wireless data evolution. Connection delays have been radically reduced by 2.5G, throughput is adequate to service the devices in hand, and the devices are finally color and virtual machine-capable. If the industry can also provide compelling content and applications, the wireless Internet should take off by 2004. If it does not, our industry is running out of excuses.

---

See Derek speak at...
Wireless Landscape Summit, June 3, San Jose Wyndham:
Summit on opportunities and trends in 802.11 and Wireless LAN.
Register at www.mobiletechforum.com. Early bird tickets $295.

---

*Mr. Kerton is a regular contributor to Techdirt and is a Principal Consultant at The Kerton Group based in San Jose, CA. For more info, visit http://www.kerton.com