[rrd-users] Neuro data in RRD - ns time steps?

[Peter Passaro]

Hi All,

I'm an academic neuroscientist who is looking for an efficient way to  
store and retrieve recorded neuronal data for use in an intranet data  
browser app. After an initial read through the documentation and  
tutorials, RRD seems to be very close to what I had been thinking of  
with one exception, our data is recorded at 5-25KHz, so we need time  
steps of 50-200 nanoseconds. So I am wondering... has anyone modified  
RRD to work with smaller time steps? If not, I am curious about how  
much coding would be required to do this (i.e. how deeply embedded is  
the time structure currently used in the database)

My alternatives would be to use a BLOB system in MySQL, or just use  
file references in a database and build a custom system to scan the  
binary data files and return the requested chunks.  So I would also  
be curious if anyone has any opinions (performance, scalability,  
etc.) of different ways of building time series data browsers.

By the way, our files are large - each experiment is recorded on 60  
channels for 2 hours, which we break into consecutive 2 minute files 
(~ 70Mb for a 5Khz recording) for ease of use. I would ideally like  
to have an RRD for each 2 hour experiment, with each channel as one  
GAUGE DS containing 36-180M data points (in the native format theses  
are 2 byte unsigned integers). Are these sorts of numbers worrisome  
in terms of performance?

Thanks,
Peter

Peter Passaro
Research Officer
Dept of Informatics/Dept of Biology
Pevensey III
University of Sussex
Falmer, Brighton BN1 9QH, UK
Office + 44 (0)1273 877590
P.A.Passaro at sussex.ac.uk


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20070828/56f2513c/attachment.html