For what I have to handle, yes, there are a lot of attributes (daily) in addition to the daily prices (OHLC). At securities level, SharesOutstanding, TradedVolume, ShortInterest. At the company level, even more - MarketCap, DilutedSharesOutstanding, P/E, P/B, DividendYield, etc, etc.. Seems like each attribute would be a CF.
On Thu, May 13, 2010 at 5:15 PM, Miguel Verde <miguelitov...@gmail.com>wrote: > I agree that it's more normal in a columnar store than in an RDBMS, but in > my experience modelling similar data, the vast majority of the time I want > all of {high, low, close, volume} and optimizing for that would be my goal. > It does seem like Steve has more expansive attributes to track (e.g. > sharesOutstanding) but just these could be in additional CFs. > > > On Thu, May 13, 2010 at 3:58 PM, Benjamin Black <b...@b3k.us> wrote: > >> On Thu, May 13, 2010 at 12:45 PM, Miguel Verde <miguelitov...@gmail.com> >> wrote: >> > I also think that's not a good design, but only because the typical >> query >> > would have to hit several column families instead of just one. >> > >> >> This is completely normal in a columnar store. You query at least one >> index CF, then use the response(s) to query at least one data CF. >> IMO, the problem with the original schema proposal is that it tries to >> do entirely too much in a single CF. Break the problem into pieces >> and it becomes easier. >> >> >> b >> > >