I'm a new user for about one and a half day so I haven't found much time to play with the system other than clicking around in the Dash. This revealed some annoyances:

- no history: I can see minutes resolution for the last hour but not the one before. Some for days etc. It would be very useful to compare with my usage of previous days (one a fine scale!). Come to think of it:

-feature request: Overlay daily usage graphs self with self-1day. In a similar fashion I can compare Fluksonians.

-bug: If my hourly usage shows a spike of say 2KW this spike will show dampened on the daily graph. Now I understand why this is and it should be changed, but: The Max and Min values below the graph are calculated over this low grained data instead of the *real* data. In my opinion their meaning is completely lost.

-inconsistency: the monthly graph shows 8 weeks.

yschaeff's picture

Err, please bare with the pre-coffee grammar. plus I meant in the 4th paragraph that it should _not_ be changed: "Now I understand why this is and it shouldn't be changed"

gonium's picture

Hi yschaeff,

welcome to the flukso community! I can understand your desire to have more fine-grained data available, but technically, this is not that easy to implements. The RRDTool Bart uses to store the data allocates a fixed size of memory. As new data comes in, old data is aggregated and stored in the compressed way. This is nice from a provider point of view because each additional fluksonian needs a constant amount of disk space to be served. But from a user point of view this is not what one would expect.

My current workaround for this is to export the power readings as they occur directly from the flukso to a small server at home. I have another thread open for this since there appears to be a bug somewhere in my code, but I have not found the time to investigate this further. Maybe this helps you - I will release my code as soon as I have fixed these issues.

On the long run I hope that a data export API on the flukso website will solve this in a clean way.

Just my 2ct,

yschaeff's picture

This is nice from a provider point of view because each additional fluksonian needs a constant amount of disk space to be served.

True, however just having the last week in details available would still be manageable and scalable. Worst case one would need 7 times more data. One datapoint every two minutes for about a week would take 1 megabyte perhaps? This seems very reasonable from an operators POV while offering users the ability to compare days.
I consider this important since I would expect this to be the core of Flukso, getting a grasp on your energy usage. *does it make a difference to unplug my toaster at night?*

As for exporting the measurements. I plan to do this as well, although a relational database seems overkill for chronological one dimensional data. What I would like to do is to read the data several times a second and write it to some other host. This host would optionally aggregate the data and send it to Flukso.

icarus75's picture

1/ Flukso at its core is about community metering so we have been putting our emphasis and development on comparing with other households' energy patterns. Your self-1[hour|day|month|year|night] suggestion is an attractive one. My main concern is to not clutter the user interface when implementing self-compare. So I'll do some further thinking on how to implement this in a clean way.

2/ I don't concur with your MIN/MAX bug report. The MIN/MAX values are calculated for the current graph being rendered. So the larger the time interval being viewed, the larger the aggregation and 'flattening' of peaks. The values give us a quantification of the data on display. You do have a point that the meaning of these statistics is not documented on the site.

3/ The 'inconsistency' in the month graph is because displaying a single month (4 weeks) would only show 28 data points. We did this initially, but gave us an ugly graph. So this is what we render now:
hour = 60 data points
day = 4 * 24 = 96 data points
month = 28 * 2 = 56 data points
year = 52 data points

4/ You can get a very sound grasp of your energy usage by the charts that are available now. Just take a look at your nightly graph and compare it to mine or someone else's. The 'night' graph is calculated by averaging the electricity consumption between 2AM and 5AM local time and thus gives you a good proxy for the amount of energy your house is 'leaking' at rest. If you observe a high value, then it is most likely caused by an inefficient fridge or deep freezer, not a toaster. I really need to write about this in a future blog post!

5/ Sampling electricity usage data multiple times per second is overkill.


yschaeff's picture

Thanks for the reply,

1/ Hmm, I think just having a '-1 overlay' checkbox would mean a major improvement at minimal cost.

2/ I do not agree. The values relate to the shown graph, but they do NOT (read: do less) relate to my peak usage that period. Rendering them meaningless pretty pictures. I'm not interested in pictures, I'm interested in power consumption.

3/ ok
4/ Please don't think I'm blaming my toaster. ;) I was merely stressing the importance to make comparisons.

5/ Yes, for the graphs it is overkill. I have Bigger Plans(TM).