I find it amusing that the tech industry starts a topic and tries to make it a market space or definition that just does not make sense. If we look back over the last 10 years, we have many examples of pundits trying too hard to push ideas that are essentially fads or are not a sufficiently differentiated. (Full disclosure: I am one of the industry analysts that contributes to this frenzy at times, so I am criticizing myself to a degree).
When it comes to machine to machine (M2M), I do not understand why it is getting so much attention in the networking space. Sure, I see the forecasts predicting that there will be 10x the number of machines than people and there will be a need for 70B additional machines for connectivity. Even if the forecasts are right, these machines will not create very much network traffic. The average device will not need constant monitoring and will not send a steady stream of data. How much data do you need to get from a refrigerator or thermostat? Temp, humidity, amps, can be monitored continually but even home energy monitoring software only needs that data once a minute at most. Plus, we forget, they are machines and they need bits and bytes, not pretty pictures and video to get the relevant information.
How much data would this mean for the network? Even, very generously, if a device were to send data once every 10 seconds, there would 8,640 transmissions per day times 1 kbyte per transmission would equal 8.64 MB per day or about 3.2GB. This is the equivalent of one Netflix movie per year. Multiply that by the expected number of connected machines per household then we might see the equivalent network load of 64GB per year per household in 5 to 10 years.
This is being generous because many of the devices will not have to transmit all of the data over the network; much of it will be kept locally with aggregated data sent on a schedule or on demand. As a sanity check, I asked a developer in the smart grid/demand side management sector what their traffic pattern looks like. They have devices and meters that talk to each other (over Zigbee, not WiFi). One device acts as a gateway and sends I.D. and usage data every 15 minutes over IP to centralize server. Using these assumptions, the data transiting the Internet would be 100 to 1000 times less.
When I mention this data to some proponents, they respond that video for home monitoring is going to be the big M2M app. My argument is that it probably does not make sense to be broadcasting all of this data across the net if no one is watching. It will work much like webcams and video on demand do today; the stream will be set up when requested.
This also brings me to my other point. M2M does not deserve its own category; it is a technology that will be deployed as a part of other application areas such as home monitoring and smart grid to serve a purpose for the application and is not an end into itself.
To put this in perspective, OTT video is happening now, and it is 40 percent of all traffic (Netflix alone is 30 percent of prime time traffic) and will grow fivefold over the next five years. By my rough estimation, the M2M traffic — if it takes off — will take three to five years to add the equivalent of one streamed movie per year per household. Other issues such as addressing, security and management will not be big issues either, because practicality dictates that most devices will use a low power, self-organizing wireless protocol such as Zigbee or ZWave and will therefore not need an IP address.
Do you agree? I look forward to a debate on this.
When it comes to machine to machine (M2M), I do not understand why it is getting so much attention in the networking space. Sure, I see the forecasts predicting that there will be 10x the number of machines than people and there will be a need for 70B additional machines for connectivity. Even if the forecasts are right, these machines will not create very much network traffic. The average device will not need constant monitoring and will not send a steady stream of data. How much data do you need to get from a refrigerator or thermostat? Temp, humidity, amps, can be monitored continually but even home energy monitoring software only needs that data once a minute at most. Plus, we forget, they are machines and they need bits and bytes, not pretty pictures and video to get the relevant information.
How much data would this mean for the network? Even, very generously, if a device were to send data once every 10 seconds, there would 8,640 transmissions per day times 1 kbyte per transmission would equal 8.64 MB per day or about 3.2GB. This is the equivalent of one Netflix movie per year. Multiply that by the expected number of connected machines per household then we might see the equivalent network load of 64GB per year per household in 5 to 10 years.
This is being generous because many of the devices will not have to transmit all of the data over the network; much of it will be kept locally with aggregated data sent on a schedule or on demand. As a sanity check, I asked a developer in the smart grid/demand side management sector what their traffic pattern looks like. They have devices and meters that talk to each other (over Zigbee, not WiFi). One device acts as a gateway and sends I.D. and usage data every 15 minutes over IP to centralize server. Using these assumptions, the data transiting the Internet would be 100 to 1000 times less.
When I mention this data to some proponents, they respond that video for home monitoring is going to be the big M2M app. My argument is that it probably does not make sense to be broadcasting all of this data across the net if no one is watching. It will work much like webcams and video on demand do today; the stream will be set up when requested.
This also brings me to my other point. M2M does not deserve its own category; it is a technology that will be deployed as a part of other application areas such as home monitoring and smart grid to serve a purpose for the application and is not an end into itself.
To put this in perspective, OTT video is happening now, and it is 40 percent of all traffic (Netflix alone is 30 percent of prime time traffic) and will grow fivefold over the next five years. By my rough estimation, the M2M traffic — if it takes off — will take three to five years to add the equivalent of one streamed movie per year per household. Other issues such as addressing, security and management will not be big issues either, because practicality dictates that most devices will use a low power, self-organizing wireless protocol such as Zigbee or ZWave and will therefore not need an IP address.
Do you agree? I look forward to a debate on this.
David Dines
ddines@acgresearch.net
www.acgresearch.net
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.