Very fast cache / access time-indexed data in c#

I'm coding a data intensive app in c#. Currently, the app loads loads and loads of timeseries from a distant sql server, does a lot of calculation to create other timeseries, and I'd like to access to these timeseries fast.

Each timeserie has a unique identifier, and should map from DateTime to anything (mostly floats,开发者_开发百科 but sometime strings, stringarray, etc).

Do you know any library I could use for that, giving me :

• fast and parallel access to these timeseries ?
• access to the "tree" version of these timeseries, to lookup the latest date, last previous date, etc ?

I've had a look a massive parallel cache, such as memcached, tokyo-tyrant or redis, but I'd have to store a somehow serialized version of each timeseries to solve my problem.

Cheers !

Sorry if I misunderstood something crucial, but wouldn't the obvious solution be a Dictionary<DateTime, object>()?

E: If in-memory storage is out of the question, ignore this of course. A local database to store the information seems the most viable alternative if you are memory restricted.

If you used a Document Database (e.g. RavenDB) or Key-Value store you'd be able to do this.