WirelessMoves

Recently, Techcrunch and Gigaom reported a couple of very interesting technical details on the amount of data and number crunching over at Facebook. Here are some:

• Current amount of data stored: 100 petabytes. That is 100.000 terabytes! Imagine the physical amount of space required to store such an amount of data…

• 500 terabytes (i.e. 0.5 petabytes) of data added each day, including 300 million new pictures.

• 105 terabytes of data analyzed every 30 minutes.

Anyone's got a link to original material on this, I'd like to hear/see it from them first hand!?

If 0.5 petabytes are added per day that's 182.5 petabytes per year if user behavior stays the same even without adding new users. Makes one wonder how long they can keep up with the amount of data they have to store as cost for it will rise as well. I wonder if the storage and power costs per TB of data is decreasing as fast their data store is increasing. Kind of a life and death question if you don't want to throw away data at some point unless you can increase your revenue at the same time with storing more and more data.

 

Many years ago when I first ventured into mobile I got myself a SIM card reader with a serial interface because I found a neat program called "SimSpy" that could read lots of fields on the SIM card and interpret their contents. Time has moved on, serial interfaces on computers became a thing of the past and my serial SIM card reader would not work any longer without a USB adapter. In other words, I pretty much forgot about it. But recently, I had the need for a SIM card reader again so I bought a new one straight with a USB interface for 16 Euros as the description mentioned that it supported a number of, let's say, 'traditional' API standards. For details see the (German) description on Amazon. And indeed, the included driver installed flawlessly even in a virtual machine running Windows XP and the latest version of SimSpy had no problems finding the card reader and reading the SIM cards I inserted (see picture on the left, with a couple of sensitive fields blanked out…). Very neat indeed!

They
are installed on pretty much on every bridge over major roads these days: Traffic
jam sensors. Recently, I got to see one close up and found it quite
interesting that they are totally autonomous, no network cable (of
course) but also no power cable. Have a look at the picture on the left.
According to the website of the company that has deployed those
sensors, data is sent over the GSM network when necessary, i.e. only
when the traffic flow changes. No mention whether SMS or GPRS is used
for the data exchange.

Then I wondered how much power the sensor
draws and if the solar panel could supply enough power for a round the
clock operation? The module has a size of about a quarter of a square
meter so how much power could it deliver per day? From experience with a
solar panel that had a size of about 4 square meters that could deliver about 500 watts under best conditions this module could perhaps deliver around 30 watts under the best conditions. Now the sun is not always shining so perhaps the module delivers 100 Wh per day. Is that enough?

That depends on how much power the device draws. The radio module itself is probably pretty efficient. Idle mode power consumption of embedded modules is typically less than 10 mW. So letting it run around the clock would result in a power consumption of the radio module of around 0.24 Wh per day plus the increased power consumption for communication. I assume the device only communicates rarely so this value should not increase much. And then there's the application processor board and the sensors themselves. As their main job is to count and analyze impulses it gets when cars move under the sensor I can imagine that its power consumption is equally low. But even if it takes a full watt, the solar panel and a battery to store energy for nighttime should be way sufficient to drive the device. And a full watt is already quite a lot, even a Raspberry Pi Model A under full load is specified to take only 1.5 Watts.

With all that is going on at the moment in the UK with the Orange/T-Mobile merger, LTE in 1800 MHz, upcoming auctions for 800 and 2600 MHz spectrum in 2012, potential sale of 1800 MHz spectrum of EE to H3G, etc. it's a bit difficult to keep track of who actually has spectrum where and how much. Is your head spinning? Mine does.

But here's an interesting document released by Ofcom that shows who has what kind of spectrum in 2012 that sheds some light on the current situation. The details are on page 8. Quite insightful.

It looks like even though SMS messages can't be sent free of charge, SMS spam and fraud seem to be on the increase and I suspect a significant part of the population may be vulnerable to be tricked into calling premium numbers that way.

When I recently talked to a somewhat elderly person I was told a story about an SMS the person had received which looked like it was sent by a popular parcel service company. The message said that they had tried to deliver a parcel but found nobody at home to receive it. The message then requested the person to dial the number in the SMS message (or click on it) to call the parcel service to arrange for another delivery. The number, of course was a premium number but not all people recognize such numbers, especially in countries that don't use 0900 or other well known codes for such numbers.

In this particular case, no damage was done because when the person tried to call the number it was already deactivated. It looks like the fraud was discovered and shut down.

What is really bad about the whole thing is that even after I explained that the SMS message was a scam to trick people into calling premium numbers, the person still insisted that the message has come from the parcel delivery service. I am not sure my explanation has sunk in 🙁 And the person I was talking to was not dumb, just not born in the mobile age…

When watching the Sol 4 press conference for the Mars Science Laboratory (MSL) mission, better known for the Curiosity rover, the person responsible for the on-board computers was saying that the on-board computer was less powerful than an iPhone. One reporter in the audience was totally baffled by this and asked how it was possible to fly to Mars with something that had less power than the smartphone in his pocket.

A very revealing question as it shows that the general public has no idea of how much processing power is required if no screen is attached to a device and software is highly optimized for a specific purpose. It would probably be even more incomprehensible to him that when flying to the moon, the Apollo Guidance Computer did not run at 1 GHz like his iPhone, but at a mere 2 MHz (500 times slower clock speed), and instead of Gbytes of RAM and flash it contained the equivalent of a meager 4 kbytes of RAM and 64 kbytes of ROM. For the details see here and for the details on the bit level have a look at this excellent book which I just recently finished reading. In other words, the specs are similar to those of a 1980's Commodore 64, even though it's difficult to compare those two computers due to their very different purposes.

But anyway, this made me wonder how many times you could store a C64 game that filled the better part of a 5.25" disc or tape in the days of the C64 on an iPhone with 16 GB of flash today. Let's take the original Giana Sisters game on the C64 for the calculation. The disk image has a size of 174 kbytes to play it in a C64 simulator today. And that's already quite big as the C64 only had 64k of RAM so perhaps it was compressed in some form or shape in memory. But anyway, 174 kbytes it is, more than the ROM of the Apollo Guidance computer could have held only a decade and a half earlier. For the comparison let's take an iPhone with 16 GB of flash, which is roughly 16 000 MB or roughly 16 000 000 kb. The game would fit into the flash memory 91 954 times!

Stunning!

An annoyance when working on the settings of a network is that the PC used for the task becomes part of that network and depending on the job, looses Internet connectivity. Even if a 3G stick or second network adapter is used, DHCP settings, which are part of the setup mingle with the PCs settings and again cut the Internet connection. A hassle I have now found a remedy for: Wi-Fi in a Virtual Machine.

A few weeks ago I had a post on how to do 3G testing in a virtual machine but this did not extend to working in wired or Wi-Fi network. But by simply getting a USB Wi-Fi stick and making it available to the guest operating system running in the virtual machine, working on a network is completely isolated from the host computer. The picture on the left shows how this looks in practice. The host operating system uses the built in Wi-Fi chip while the guest operating system (Windows XP in the window) uses a USB Wi-Fi stick and connects to the network to be configured. No matter what is done there and how that affects the client operating system, the host operating system (Ubuntu in my case) is entirely unaffected from the changes as it uses a different network and Internet connectivity remains in place.

And if you have enough RAM you can even launch two or three virtual machines simultaneously, each using a different Wi-Fi or 3G dongle connecting to different networks. Great for testing VPN solutions, NATing, etc. etc. The more I play with Virtual machines, the more uses I find for them.

Now here's the perfect gift for a telecoms man who doesn't really have all that many wishes for someone to fulfill easily. Every now and then when tracing signaling messages, being able to spot or decipher numbers in binary is essential. But when not doing it regularly, the process is a  bit cumbersome. But here's the solution: Train you binary skills with a watch that displays the time of day in binary. After two days I was fluent again 🙂 Priceless…

An unexpected result: Chip, a popular computer magazine in Germany, noticed when doing their yearly cellular network measurement campaign, that LTE coverage from one of the leading network operators was present on 95% of their route all across Germany. For the details see here (sorry, in German, but Google can help with the translation…). And for comparison, here are last year's results.

And two other especially noteworthy data points:

1. Call setup success rates >99%, drop rates of ongoing calls <2%. Values customers in other countries can only dream of…
2. Wireless Internet connectivity of the leading operators is still as good or even slightly better than last year. The number of smartphones and data usage is rising but the numbers show that even without LTE taken into the equation, well designed networks are capable of dealing with the load.

In other words: It can be done, don't let those who can't, make you believe otherwise!

P.S.: Always looking for links to similar tests done in other countries!

One of the things a well configured GSM, UMTS or LTE network does is to give the mobile device clear and precise instructions of when it should select another cell or even performs a handover to a better suited cell during an active communication session. There are plenty of standardized parameters and algorithms based on the signal strength of the current cell, the neighboring cells, offsets before a lower speed technology network is selected, interference, etc. etc. When a mobile ends up on the Wi-Fi layer, this kind of sophistication abruptly ends as I recently experienced.

On the Wi-Fi layer it's completely up to the device to decide when it is time to reselect from one Wi-Fi access point to something else. The device I played around with clung to the Wi-Fi access point right down to the last dbm where communication was hardly possible anymore, despite an excellent other Wi-Fi network with a different known SSID in range. I manually had to reselect to the other Wi-Fi to continue working. Also, reselection from Wi-Fi to the cellular layer is probably also only done once connectivity with the Wi-Fi network is lost, which often happens much later than moving out of the "usable range" of the network where data rates are still acceptable.

Sure, Wi-Fi was never designed to include that kind of functionality and 99% of home users would likely be unable to make the required settings. Also, the break incurred in terms of IP connectivity and a different pricing between cheap home Wi-Fi and a more expensive cellular layer makes the decision to move from Wi-Fi to cellular as late as possible understandable. Nevertheless from a usability point of view it's far from ideal. In other words, the user has to make sure the Wi-Fi signal is strong enough everywhere in the house or appartment so devices never leave the usable range.