pCell is only as good as the Linux it runs on
I’m still working-away on my IBM book and it is still a week from being finished (the well-known second 90 percent syndrome). The book, if I am allowed to sell it on Amazon, will cost a whopping $3.99 and will be worth the money. But I’m still a columnist of sorts so here are my thoughts on pCell, an impressive new technology for increasing performance of LTE mobile data networks. It was invented by WebTV founder Steve Perlman, introduced two weeks ago in New York (very impressive video here, but fast-forward to 5:30) and was the talk of the Mobile World Congress in Barcelona the following week. pCell is amazing. It is also probably a security nightmare waiting to happen.
This is not me being a bad-ass or somehow wanting pCell to fail. I think it is great and I want it to wildly succeed, but there are a couple things about pCell that have been going over the heads of most reporters, security being one of them. I’ve read all the stories about pCell and the word security doesn’t appear in any of them, none.
Typically with new technologies like this the inventors haven’t thought much about security or they rely on a small installed base to keep the product or service under the radar of the bad guys. But pCell, for all it’s high tech loveliness, is a Software Defined Network proudly running in a data center on plain old Linux servers. So if there is a security vulnerability found in Linux (Red Hat found one just this week that I’ll be writing about later tonight) then pCell is vulnerable. Not might be vulnerable, IS vulnerable.
This vulnerability will probably show itself first as a disruption of service. Since the personal (that’s the p in pCell) cells are defined and controlled from the servers through transceivers scattered around town, if you mess with the server you mess with the network and can potentially take it down. So pCell, as Linux software, is vulnerable to Linux problems.
It just worries me that nobody is mentioning this because I expect we’ll see a ton of pCell announcements shortly with major mobile carriers covering densely-populated areas where this service will shine… until it is hacked.
The other point that is glossed over in the pCell discussion is total bandwidth: if you are going to have every pCell user watching his or her own HD (not to mention 4K) video signal, the network is going to have to be provisioned with enough bandwidth to cover those total users. You see pCell is about the efficient utilization of radio spectrum, not network bandwidth. If it takes two megabits-per-second (Netflix recommends five mbps by the way so my numbers here are probably too low) to practically serve what the video network vendor will claim is a 1080p HD video stream using H.264 encoding, then you’ll need two megabits for every mobile user watching video on the pCell network. So when they talk in the video about a single pCell radio helping to create a clear channel for up to 1000 users, then each of those users had better be provisioned with at least two megabits of backbone or cached server bandwidth.
Yes, an LTE connection is faster than two megabits, but unless there are two million actual bits on the other end of that connection (called provisioning) having an empty pipe is meaningless.
If there are two million people in Manhattan, 10 percent of those are on the phone and 10 percent of those are watching video on their phones, then the aggregate bandwidth for all those users had better be 20,000*2,000,000=40 gigabits-per-second (100 gigs according to Netflix). This isn’t much if video is cached and delivered from the data center, but if it is coming across the Internet backbone it’s a lot.
There are probably limitations, too, on how much bandwidth each of those pWave radios (the ones on buildings, not the ones in your phone) can handle. If as suggested by Perlman’s claim of single-virtual-channel operation, then each of those radios probably maxes out around 75 mbps based on the 10 MHz LTE standard.
So how many pWaves would it really take to serve Manhattan? At least 40gigs/75megs=533 radios. Looking at it from a different angle (Perlman’s claim of 1000 users per radio) then the number under the scenario I described is lower (200,000/1,000=200). That isn’t a lot but it is one radio for every 27 acres (one radio for every 700 acres in the latter example — did you know Manhattan has 14,526 acres?). This explains why they had several radios in the demo room on the pCell video in order to show dual 4K video streams running in 10 MHz.
Where I am going with this is that nobody really knows how many cell towers there are in Manhattan but only nine are registered with the FCC http://www.city-data.com/towers/cell-Manhattan-New-York.html and the largest number I can count on a cell tower monitoring site is 38. Whichever of my pCell numbers is right is going to require some significant changes in the way mobile carriers do business in order to best utilize this new tech in the dense areas where it is needed most.
Based on the fact that Comcast, for example, is building opt-out public WiFi access points into its cable modems, I suspect the mobile carriers are going to try piggybacking on customer bandwidth, too. Because the more the merrier when it comes to pCell. If Verizon, say, offered some of its FiOS customers free Internet in exchange for hosting a pCell, their mobile network could be locally transformed overnight. Who’s going to turn down a deal like that?