Tobias Oetiker wrote: > If you now store everything as integer, this resampling process > will loose precision which is realy not something you would want > todo to your data.
... > So if you insist on storeing your data as integers, you would have > 3 options: > > a) devise a magical way that enables you to collect all data at > exactly time % step == 0 > > b) lie about the data collection time, and fill in the data that > you collected at time % step == x at time % step == 0 > > c) use a generic (SQL) database where you can store the data you > collected as well as the time you did collect the data. It looks like for this special purpose I'll have to define my own data model ... > The interesting thing is to select a samping interval that is > correct for the rate of change in the data you are observing. That's exactly the point: I don't need to sample data at a 1min interval during hours where nothing happens, however, during busy times a 1sec interval might be apropriate ... So here we're talking about another requirement apart from storing exact values - I'd prefer to do the interpolation of missing values at the fetch-time, not at the store-time - saving space in the DB. I like the idea of a round-robin-database very much, however, the implementation of rrdtool doesn't always do what I would like it to. I even thought about storing the data in some other form (generic db) and put it into rrdtool just for the graphing ... So - I opened this discussion - let me close it here, it doesn't lead anywhere. I don't want to offend you since I really do appreciate what you've done with this great tool - it just doesn't meet my requirement. What I need (for this purpose) is a different database model with the same graphing features rrdtool has - that's why I thought about some changes to the underlying db and keeping a big portion of the rest. Many thanks for all. - Karl _______________________________________________ rrd-users mailing list [email protected] https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
