Tuesday, 14 June 2016

The Cost of IOT - What is involved


The Cost of IOT – What is involved

 Aside from capability, one of the big ticket items that will hold IOT back is the cost. All the hype in the world won’t propel IOT to stardom if the cost does not match the delivery benefit. And right now, no-one seems to be talking cost so we will lay out some of the components for you to consider. But as you will see below, don’t start asking about cost until you know your provider understands the business end to end.


1.            The Sensor
Most unfortunately, this is the murkiest of corners in the IOT space. You can get cheap one-wire based temperature sensors for $5 but it takes an electrician (or experienced data cable handler) to wire it in. $65 gets you a light bulb monitor but there are a lot of light bulbs to hook up. A decent soil moisture probe can cost anything from $400 to $2000. Measuring temperature, humidity, pH or water quality is similarly variable except now the price range is $5 to thousands. Setting up a door switch has no sensor cost but probably an hour’s labour to fit.
The further problem is that if you use LoRa fitted transmitters, you may be constrained to purchasing from a limited range of compatible devices and there is often a price penalty for this. Non-conforming devices are still possible but you’re up for an installation charge. When it is LoRa there is a per sensor charge which might be $150/year. If you use 3G/4G connectivity, there is usually a single $150/year charge which can cover up from one to 20 or more sensors. Those sensors will mostly incur an installation charge but not an additional monthly charge.
We can leave the discussion around sensor technology to another day but suffice to say there are four main types: digital, one wire, low res analogue and hi res analogue. The technology behind the brain dictates which of these technologies are possible but also the technology of the remote sensor plays a huge part. Cheap sensors are only possible where the main controlling unit allows them to be connected and the right I/O technology is in place to utilise them. So when you are discussing your solution with your IoT provider, you need to know what sensors they are proposing for you, what they will cost and if they come with monthly charges.
2.            The Brain
Almost all IOT implementations have some central controller. Whether it is LAN, LoRa or 3G based, there is always a cost in the order of $2000 per year. Sometimes this is a monthly based charge where you never own the unit and others are outright purchase. The LoRa brains communicate wirelessly to their sensors whereas the 3G ones are mostly hard wired up. In something like a hothouse where 20 sensors are common, it is unlikely to be serviced by LoRa technology. Grain silos separated by distance might be more economically connected via LoRa.
One of the key attributes of the brain is its capacity to run code and here is where there is a massive difference in capability. With Yun or Arduino combinations, the code base possible is extremely limited because the memory space is limited – 256KB. The code, often C based, is very limited. More complex boards such as the Raspberry PI and DragonBoard allow the deployment of a “full” operating system such as Linux or Windows 10 IOT Core. They have up to 1GB of on-board RAM and now permit the execution of complex code. This allows us to contemplate pseudo sensors such as airlocks or pressure differential sets and proactively manage outgoing data streams. Assuming there are applications written to exploit the bigger operating system, there is a clear advantage with Dragon Boards or Raspberry PI but there will also be a higher cost.
3.            Installation Costs
We can’t escape this and essentially is a standard electrician’s rate. The only difference between a LoRa solution and a 3G one is that there is more wiring for the typical 3G because each sensor needs to be wired in to the brain. Having said that, the 3G sensors are often a fraction of the price of LoRa sensor probes and do not incur monthly charges on a per sensor basis. A complex hothouse might take 20-30 man hours to wire up so there might be a once off cost of $2000 to contemplate but there are no monthly sensor fees that would easily exceed this on an annual basis.
Any solution should expect at least $400 for installation of the brain.
4.            Bandwidth charges (SIM cards)
We have already covered the monthly charges found on both 3G and LoRa installations but what must be considered is excess charges in the 3G environment if the data flow is not contained. If your solution uses for JSON, which has a data density of only around 5%, you could be up for GBs per month per device. Now your bandwidth charge per device is looking more like $30/month instead of less than $15. LoRa will never suffer this expense because it cannot transmit that much data. This also implies that LoRa would never be deployed in environments where detailed trending data needs to be collected.
The takeaway is that data cost is proportional to the data need. More data means more cost. When choosing your IoT provider you will need to know what data volume your hardware will pass or what it will cost you if it is variable.
5.            Platform Charges
Comparing platforms is quite difficult. Here are some examples for sort of equivalent IoT suites:
·         Amazon - $8 per million messages
·         Cloudera – Well they won’t tell you directly because their data hub is kind of a one stop shop but it runs on Azure. Essentially the pricing model is per minute but as it’s based on MS Azure, you can expect it to be about the same
·         Azure -  $50/month for up to 400,000 messages per day ($4.10 per million messages)
And there is a curious thing going on here. Find IoT platforms and find their pricing! If you search for IoT frameworks in your favourite browser you will get the likes of Dell, IBM, Bosch, Rockwell, Cisco, Intel, GE and even Facebook, all touting their expertise and capacity. Their websites are full of spin and promise and whilst they may be investing in devices, networks and sensors, they are not building platforms the way Amazon and Microsoft have. Well at least not yet and nothing that provides a pricing model. For now, this places us back to the above three.
Amazon charges per message whereas Microsoft charge per service. Each are scalable but in different ways and on a different basis. For example, if you have a slow message day, the Amazon cost is lower but you can’t automatically ramp up from one pricing tier to another. Microsoft sells you a pipe of a certain width and you scale within automatically. But it isn’t as simple as this and there are other services and capabilities you also need to take into account.  On top of the platform there are storage charges ($5/month per client) and web service charges (that sort of work out to around $15/server). In a genuine multi-tenanted approach, it is conceivable that we are looking at a further $15 - $25 per month per device/customer, depending on how the environment is configured.
The trick here is to find a provider that can roll all this together for you and offer a multi-tenanted solution so that a bunch of IoT users get to share in the lower pricing possible with scale. If you are to do this yourself, you will have to come to terms with this complexity.
6.            BI and Visualisation Charges
Not surprisingly, this can be the most expensive component of all. If you want to use Microsoft’s BI and Stream Analytics services, as powerful as they are, they are also very expensive. You would be looking at $25/user per month per service. Amazon, Cloudera, DigitalOcean and others are all just as pricey. However, if there is purpose written software servicing the specific market vertical being monitored, there must be a software charge of some kind. Most of the companies racing to build IoT style sensors do not also build the software applications to manage and trend the collected data so this makes finding a solution partner difficult. Equally, software developers, who usually do not have industry IP to start with, do not readily jump into bed with hardware partners because it locks your software tight against a specific hardware strategy. I am predicting a lot tears and wrist slashing will take place over the various partnerships that seem to have formed to date. Most of the agricultural based software I’ve seen is seriously expensive – out of reach to the mass market. Similarly, wiring up your home, be it for medical monitoring or turning your heating on, costs a bomb and those who dipped their toes in this market have all gone broke.
7.            Research and Development
If nothing else, I hope I have conveyed to you the complexity involved. How you go about finding a technology partner to build an IoT solution will be the subject of another blog but your first priority is to establish your role in the chain. Are you the consumer or the provider? Are you doing hardware or software? Which platform are you going to use? Who will be your project integrator? Are you going to do all this yourself? You can see that the sheer research and development involved is not trivial. Most companies like ourselves have been toying with this for years and it isn’t being made easier by changing technology – at the sensor and brain end as well as the platforms in the middle. This research is expensive and that cost must find its way into the solutions somehow but there is no doubt a world full of companies racing forward to get your attention. Just be prepared for the cost.
Geoff Schaller
@IOTRemote

No comments:

Post a Comment