So far I’ve looked at devices with which we’ll interact with information in 2020. Now I’m changing gears to look at the backend.
Fry’s is selling Terabyte hard drives for $69.99 this week. A Terabyte is, give or take, a billion pages of text, or a million books. An entire library. Or a month’s worth of video. At the current rate, give or take, we’ll get a factor of at least 100 ($0.69 per Terabyte) by 2020; chances are it will be all solid state memory instead of spinning disks, but I don’t care too much about this distinction in this blog. I do care that we will also get substantially higher communication bandwidth and better broadband coverage by 2020, which can be used to instantly access more storage in the cloud. Note that much of the need for storage on client-side devices is driven by a lack of good high-speed connectivity: if I can easily get at my iTunes collection over a network, there is not need to carry a copy of it in my pocket, a major driver for client-side storage needs today.
Together, this means that by 2020 we’ll essentially have unlimited storage capacity for free on our personal devices for any foreseeable application. I’ve been attempting to come up with applications that require multi-terabyte storage on a client-side device, and have come up with basically nothing. Visual input is the highest-bandwidth input channel to humans, and there is no way even a single terabyte is insufficient as a cache to a high-speed communications link for video. Sensors on devices are unlikely to produce higher-bandwidth data either.
If we go into the realm of the very speculative, we could think of some wearable health monitoring systems (like “portable MRIs”) that could possibly produce higher data rates, but for a variety of reasons I think those are unlikely by 2020.
So the price points for client-side storage will gradually approach zero because there’s no need for higher-priced storage on the client. One lesson to draw: you won’t get rich as a vendor of client-side storage in 2020.
P.S. as usual, please tell me that I’m wrong if I am!
Comments
2 responses to “Unlimited Free Client-Side Storage”
I think Johannes is right. I’ll go even a step further: Since client-side processing power also will increase, the combination will bring algorithms that today only work in the cloud down to our pockets. More processing power means smarter caching and background processing. We’ll collect continuous streams of high-resolution data from our bodies and surroundings, crunch it to do things we haven’t thought of yet, and archive the data in the cloud when bandwidth is available.
DanLyke is right that there will still be limits to portable bandwidth: Bandwidth will always be better in some fixed locations than in arbitrary or remote locations or when moving at speed. But more processing power means we’ll need less (and less reliable) bandwidth, and more storage means we’ll be able to cache more data more intelligently, syncing (merging?) with the cloud later.
A short-term example: Google Nav on Android is great, but even in metropolitan areas, its need for a continuous network connection is a severe limitation. My Garmin GPS, with its client-side database (only 2GB), still works way better. It seems obvious there are people at Google right now working to build predictive caching of map data and points of interest, using client-side storage (my Droid came with 16GB) and processing to hide the fragility of the connection to the cloud.
I still wonder at the notion of essentially free bandwidth for our portable devices. The “n to n” network problem still looms, and shows its head every time I try to watch a YouTube video, and with AT&T whining about network congestion due to iPhone usage, this is a problem I don’t see as solved in ten years.
I know I already have various strategies for caching stuff on and off my portable devices, and I wish my iPhone were much much better at that (I’d love to be able to cache a couple of gigabytes of Google Earth data for areas I’m in or am likely to be in). I spend a moderate amount of time in places where data connections aren’t just slow (ie: San Francisco), but non-existent (large swaths of western Marin and Sonoma counties), and that’s even more true for places I pull out my camera or other data acquisition devices (admittedly, all of which have far far lower data use than the camera).
So I guess I’m not disputing your notion that nobody’s going to get rich selling client-side storage in ten years, but I think we’ll be demanding a couple of hundred gigabytes of client-side on those machines.