Influx on micro ssd

Grafana InfluxDB
1

Hello

Anyone has experience with influx on micro SSD? or what is the smallest device you have installed it on?
We are looking at Influxdb for quiet a lot of IoT devices that might be in regions where internet bandwidth could be low at times and were thinking of saving the data on IoT device itself and do a dump when connection is good type of arch design approach

2

Hello

Anyone has experience with influx on micro SSD?

Do you mean “micro SD”, or is “micro SSD” something I haven’t yet upgraded to?

or what is the smallest device you have installed it on?

I use it on Banana Pi BPI-R1 - Banana Pi Wiki with 4Gbyte SSD.

We are looking at Influxdb for quiet a lot of IoT devices that might be in
regions where internet bandwidth could be low at times and were thinking of
saving the data on IoT device itself and do a dump when connection is good
type of arch design approach

The biggest questions which come to my mind about that are:

  1. You say “bandwidth could be low at times”. Do you mean “low” or do you
    mean “non-existent”? What do you mean by “times” (duration)?

  2. You talk about “saving the data on IoT device itself” - how much data? At
    what rate are you acquiring this data?

  3. I assume that when bandwidth is not “low” then it’s sufficiently good to
    catch up on the backlog of data and send it to the central collector.

I think the calculation is pretty simple, provided you have the necessary raw
numbers:

a) Tmax = maximum amount of time a device can be offline (I’m going to ignore
the concept of “low bandwidth” for now - either it has a “sufficiently good”
Internet connection or it doesn’t)

b) Rdata = quantity of data collected by the IoT device per time period

c) S = storage capacity of the IoT device

If Rdata x Tmax < S then you’re okay. Otherwise you will lose data.

I know the above may seem somewhat trite, but without hard numbers I can’t
think of a better way to express it.

Antony.

1 Like
3

micro SSD is what I mean :wink: no need to upgrade champ. SSD one can install operating systems etc. But maybe influx as well, I was aiming for something tiny whether sd or ssd :wink:

No triteness detected I appreciate your questions, put my toes to the fire

  1. You say “bandwidth could be low at times”. Do you mean “low” or do you
    mean “non-existent”? What do you mean by “times” (duration)?
    Sometimes it could low and sometimes non-existent. At times the device could be in a location where cell data is low or any other means of getting internet/data is low or geo location issues, weather etc
  2. You talk about “saving the data on IoT device itself” - how much data? At what rate are you acquiring this data?
    I need to figure the exact size for sure. But data is coming at per mili/sec rate so it will be a lot. Just queried for 1 device and I see 32k “rows”. If that helps.
  3. I assume that when bandwidth is not “low” then it’s sufficiently good to
    catch up on the backlog of data and send it to the central collector.
    Yep that is the approach I am recommending.

Tmax = this could vary as the device might be idle for some undetermined amount of time If low bandwidth we will consider it no internet. Don’t want to wrangle with it

rdata = 32k rows per 24 hour period. I pulled this data from one of our ES that pulls data via kafka. Will need to convert this to GB

S = we will not be stingy on this so depending on the IoT itself we will see what the max is we can put on the device so as not to also add weight etc. Need to cal all these for best fit

Thanks for walking me through this! I have it in my brain but needed to sort it out with a peer.

This has been a help also