Hi Alexandru, I feel Cassandra can certainly be used to solve the problem you have but if your requires are not very strict, you need very high throughput and its okay for you to lose some data occasionally due to machine crash, then I recommend you look at Redis (http://redis.io/). It is a high performant key/value store with very high throughput and used for analytics.
thanks, Ritesh On Wed, Feb 23, 2011 at 12:45 PM, Alexandru Dan Sicoe < sicoe.alexan...@googlemail.com> wrote: > Hello, > > I'm currently doing my masters project. I need to store lots of time series > data of any type (String, int, booleans, arrays of the previous) with a high > writing rate(20MBytes/sec -> 170TBytes/year - note not running continuously) > but less strict read requirements. This is monitoring data from a vast > distributed network. The queries will be something like: give me this data > between Time1 and Time2. > > The hardware that I have available is between 2 and 5 hosts. > > Questions: > > Should I use Cassandra? > > Suggestions of how to structure the data? (I read > Cloudkick's blog > https://www.cloudkick.com/blog/2010/mar/02/4_months_with_cassandra/ but I > found that it doesn't give too much detail) > > > Any help is much appreciated, > > Alex >
