Howdy!
I'm looking to use the framework for high-performance storing and querying of financial time series, i.e. ticks arriving from the exchange. I need to be able to store 500-1000 (or more) ticks per second and would like to be able to quickly find similar time series given a date range.
I'm looking to create a cheaper alternative to www.kx.com with their KDB and K language. For each tick I need to store the datetime, symbol, trade type (bid/ask/trade), price and volume.
How would I do a similarity search? I would like to select a portion of my stock chart (for example) and find subsequences in other stocks that match the pattern.
What is the best way to take differences in scale be taken into account?
Would I need to calculate the Euclidian distance myself?
If I am to write wavelet or Fast Fourier Transform code, should I write it in MIL or use the Python/Perl/etc interface?
Thanks for your help and please use the link in my signature to read up on my project.
Thanks, Joel