Archive for November, 2014

The (best effort) Internet of Things

Saturday, November 8th, 2014

Back in the early 1990s, I was manager of new products at WilTel.

WilTel was a scrappy telecom competitor (focused on the business and wholesale markets) whose greatest claim to fame at the time was being the first carrier to launch frame relay services. Frame relay was a packet-based service that dramatically changed the data networking landscape. It delivered performance significantly better than X.25, the previous option for packet services. In fact the performance was so good that businesses could replace expensive private line mesh networks with much more affordable frame relay networks made up of virtual circuits connecting the same end-points. It was a runaway success. In fact, WilTel’s introduction of frame relay is a great case study in successful self-cannibalization. Private lines were WilTel’s bread and butter, so frame relay meant that the company aggressively introduced a product that replaced it’s core product with a less expensive alternative. Sure, for many of WilTel’s existing business customers, there was a reduction in what they paid the company for the equivalent benefit. However, WilTel was able to not only replace a single customer’s WilTel private lines with frame relay, but also that same customer’s AT&T private lines and MCI private lines and Sprint private lines, resulting in a significant increase in “share of wallet” and a meaningful increase in overall revenue from each customer. More importantly, WilTel became a credible provider to many business customers who had never before considered the company. (Sorry for the rabbit trail, but it seemed like a good teachable moment…)

I can’t take credit for the brilliance of this move. I played a very small role in the launch of WilTel’s frame relay service. (That credit goes to Joe Zell and Christine Heckart.) But I tell the story to make the point of how important frame relay and private line services were to the company in the first half of the 1990s when I moved into the role of manager of new products for WilTel and started to see the emerging potential for Internet-based services. I started preaching that, in the future, businesses wouldn’t need private line or even frame relay services. Why would they pay for all of that, when “the Internet” could just as well deliver their data from point a to point b. (Of course, I was right, I was just about 15-20 years too early. For my last 9 years at Sprint, the company was slowly moving towards shutting down frame relay and private line services.)

Anyway, in my role at WilTel, we started increasing our focus on Internet-based services. All of that came to a screeching halt at the end of 1994. WilTel was in the process of being acquired by LDDS (the combined company would be renamed WorldCom) and I put together a summary of our product development efforts for LDDS’ CEO, Bernie Ebbers. His response: “That Internet stuff – shut it down. The Internet’s a toy. Businesses will never pay for it.”

Now, I know that Bernie gets (and deserves) a lot of grief for stupid decisions he made that destroyed WorldCom (and with it the retirement dreams of thousands of employees). But I tell this story to make the point that there was a time when the Internet was broadly considered “not good enough” for “real” use.

The Internet was designed to be a “best effort” network. No performance is guaranteed. As traffic increases, the network does its best to respond and in general everyone’s performance degrades. (Net neutrality is all about maintaining this model.) Today, Internet service providers have invested so much in the core infrastructure and access networks that this “best effort” is actually good enough for the vast majority of uses that consumers and businesses have, and providers have offered dedicated services, apart from the “public Internet”, using IP-related protocols (e.g. MPLS) to ensure the performance that businesses require for their more mission-critical applications.

But it wasn’t always this way. In reality, you couldn’t count on any specific throughput or even successful delivery from your Internet service. The only thing you could count on was unpredictability and unreliability. For many uses, that was just fine, so people bought it and the Internet grew dramatically. For at least a period of time, Internet traffic grew exponentially.

Wow, that’s a long lead in to my main point…

I think it’s important to realize where we are in the “Internet of Things”. As we are rapidly deploying these technologies, it’s important to have right expectations. In time, the vast majority of IoT applications will be rock solid with high levels of performance, and even today, those applications that need it are (hopefully) being built with the right level of investment in the right kind of core infrastructure so that they will work when they need to. (I’m thinking of things like the government-mandated Positive Train Control.)

Today, the vast majority of applications are “best effort.” If it fails – oh well – we’ll try again tomorrow.

Case in point – my smartwatch. I have loved connected watches for a very long time. In 1994, I bought a Timex DataLink watch. Then in the late 1990s I started buying watches with a satellite connection to the atomic clock. By 2009, I’d written off the future of watches, thinking that the smartphone had replaced them, but around the same time I got my first FitBit fitness tracker. I was ecstatic when fitness tracking functionality became integrated into the watch form factor.

My current favorite smartwatch is the Basis band, which I’ve been wearing for a couple of years. Basis has since been acquired by Intel, and is launching an exciting new product. What I love about my Basis is that it encourages me to develop healthy habits. Everyday, I’m working towards eight different goals, from not sitting too long, to walking at least 2000 steps each morning, to walking at least 10,000 steps each day. There’s also a sense of gamification, earning points by hitting goals allows you to add more habits. It’s great – when it works. It’s absolutely frustrating when it doesn’t.

For example, yesterday, the Basis cloud (which is what rewards me for hitting my goals) told me that I walked 6,640 steps in the morning (yea! – crushed that goal), 5,570 steps in the afternoon (yea!), and only 823 steps in the evening (oh well, can’t win them all). But the really surprising and frustrating data is that Basis thinks I only walked 6,482 steps total for the day. You’re probably thinking what I always think when this happens – that logically is impossible and mathematically is about half of the actual total. But, the big deal for me is that I didn’t get credit for my goal of 10,000 steps, even though I clearly achieved it.

Unfortunately, this is not unusual. In fact, it’s happened twice this week and probably a handful of times this month. I’ve been reporting the problem to Basis for a long time. Their typical response is “do a hard reset of the device” (which really never does any good – it doesn’t recover the data for the current day and doesn’t seem to improve performance over time). The most recent time I reported the problem, after a few back and forths with customer service, I got a more complete response which included this explanation: “The Basis clears the watch memory after a sync to prevent the memory from becoming full and failing to record further data. This clearing of data is the last step in the sync process, and normally, if the sync fails, this action does not trigger. Sometimes, however, data corruption does not register immediately and the action to clear the watch memory is triggered. This results in the loss of data from the watch memory. The totals on the watch face do not reset with each sync in order to keep your daily data accurate, but if a segment of that data is corrupted, the data on the web will only show the information that was successfully uploaded to it. This is why you can see discrepancies in the data on the watch face versus what’s in the app.” Never has Basis indicated that are going to do anything to fix what would be a relatively easy thing to fix (when morning+afternoon+evening > total, update the total…). For these reasons, I won’t be buying any more products from Basis.

But my point isn’t meant to be that the Basis product is a failure. It’s that virtually all IoT products are going to have similar failures. Bluetooth, WiFi, 4G, software, hardware… they all fail at times. The fact that my smartwatch fails to give me credit for what I’ve done is frustrating, but it’s an application that is just fine for “best effort.” And for now, “best effort” isn’t all that good.

Thankfully, if history is any indication, “best effort” is going to keep improving, and IoT will fundamentally change the way that we interact with the world around us, and the way that businesses operate.