Necessary Roughness

UPI reports that Japan has launched the WINDS satellite, which promises internet service faster than today’s normal broadband service.

“Among other uses, this will make possible great advances in telemedicine, which will bring high-quality medical treatment to remote areas, and in distance education, connecting students and teachers separated by great distances,” the statement said.

While I was teaching another part of our corporate workflow software, I got exposure to HAL’s Real Time Reservoir Solutions capability, which involved satellite internet. Back then, the speed was pretty slow, maybe 2-3 times faster than a good dialup connection. I’m sure that’s come a long way, but there’s another fundamental problem that I haven’t seen solved yet.

When talking about data speed one has to be aware of two contributing factors: the carrying capacity of the data during each transfer (sometimes referred to as packet size), and the latency.

Think about a semi versus a motorcycle. A semi with a trailer can haul a lot of material, but a motorcycle can’t. The motorcycle, though, will usually get to a destination more quickly. The ideal method of transfer would haul a lot of material very quickly, without getting anyone killed. Packet size is like the semi, while latency is like the motorcycle.

You can imagine that if you need to move a trailer’s worth of goods, you’d employ the semi, because even with the motorcycle’s high speed it would make many more trips and thus end up finishing the task more slowly. If all you wanted to do was visit somebody and get back home, you’d take the bike (or other preferred smaller and faster conveyance).

Data has to make trips too. The satellite technology I’ve seen has focused on the capacity, getting a lot of data to its destination, but physics has always hurt the latency. Information has to go from ground to satellite and back to ground again. Even a Low Earth Orbit satellite is still 250 miles up (Wikipedia). DirecTV and Dish Network are possible because of continuous large-capacity trips going one way. The fastest possible response time we saw, latency measured by pinging, was about 1.5-2 seconds. For comparison I can ping the Drudge Report from my hotel room in 15 milliseconds, at least a hundred times faster.

Because of high capacity (good) but high latency (bad), one can move a lot of data, but control systems are a pain. In the oilfield, an engineer can monitor a job and offer suggestions from thousands of miles away, but there has to be a service supervisor on location to make that split-second decision to shut things down if something goes wrong. Sometimes the engineer will forget that (it wasn’t me, but I saw it happen).

It will be interesting to see how/if the Japanese can solve the latency issue.