I want to scale out this account to many hundreds of devices, but I have a serious concern about data usage.
I have active devices on this account that are calling in, on average, 4 times a day. Each payload on call in is well less than 1K. To my surprise, most chips are showing consumption usage in the many MBs, some of which are showing 21MB, 19MB, 14MB, lots north of 10MB. That seems wildly wrong. Something is going on.
The problem is now my particle account is deactivated because I shot WAY past the 100MB limit (currently at 360MB). Even though per call in is tiny (less than 1K) and I've only accrued 4,249 total data operations (4% of 100K total limit). So, I'm way, WAY under data operations limits. Yet, my cellular data consumption is skyrocketing!
To put this in perspective, I have a client I set up on Particle who is now paying for over a hundred devices on the system. Last month they used 12K data operations and consumed only 68MB.
I need to solve this problem before I scale up this account for my client. Something is wrong. Data consumption is way too high and I can't afford to put on hundreds of devices in the account at this consumption rate.
Can you help me figure out why this is happening and how I can reduce that cellular data per month. I've included a screenshot of a report I generated showing data consumption per device (with sensitive data invisible). To a particle employee I can make this data fully visible so you can help me track down whats happening. The account is under rfs@risingtiger.com
Looking at some of the greatest offenders, these are devices that are really struggling with signal and are likely disconnecting and connecting to their local towers endlessly and consuming data overhead. We'd recommend taking a close look at these devices and their available signal quality and either changing their connection logic (using SEMI_AUTOMATIC and a limited number of retries) or adjusting their antenna. Connections less than 10-15% quality are somewhat likely to consume additional data overhead, though this tends to be highly variable.
So, are you saying that data is consumed every time the particle chip must reestablish a connection to the local cell tower, even if no payload is transmitted up to the cloud? How much data is that typically?
Connecting to a tower specifically (blinking green) doesn't actually consume data, however other parts after that in the process do.
Under ideal situations, a session resume can be done, which uses less than 200 bytes of data. This is also the overhead when waking up from sleep with cellular off.
There can be overhead for sending things like device vitals, which do not consume data operations but do use a small amount of data.
Under worst case conditions, repeated failure to connect successfully could cause a session invalidation, which would require up to 5K of data to authenticate with the cloud again.
I see. That does help clarify where the data consumption is happening.
I'm still baffled as to why the consumption is high. My other client's account has devices all over the United State's and some have very, very poor cell service, yet they're all staying within a reasonable data usage.
The only data I'm sending is via Particle.publish. But I am using a library called 'PublishQueueAsyncRK' that supposedly retries publishes if one fails.
Hey @risingtiger, given the nature of distributed systems on these networks, these sorts of issues are unfortunately quite common in the IoT world (Particle or otherwise!). Hopefully we can give you some confidence that there are both technical and commercial options to alleviate your concerns.
From the technical perspective, I'd love to learn more about the environment in which you are typically deploying, and your antenna choice, configuration, and placement. While I can't provide a detailed RF review here on the community, there may be some pointers and suggestions we can provide to get you some better performance. At the end of the day, these variables carry the most influence over the symptoms you are seeing.
Also to double down on something Marek shared, signal quality (RSRQ) is the biggest contributor to poor connectivity. You mention you have other sites & clients with worse connectivity — what metrics are you using to quantify that? I'd be curious how signal quality compares across those devices and the ones with high consumption.
On the commercial side, what sort of scale are you hoping to reach in the next year? Our commercial plans don't all scale linearly like the Growth plan you are currently on, and we may have options that would give you additional headroom and confidence in your deployments going forward.
