Clustering with memory engine

[Gadi Naveh]

Hi

I have a need to be able to access large amount of info at high speed, where the info is double indexed (k1(k2, k2...) and k2(k1, k1, k1...)). for this is i am considering the memory engine. however, i am looking for a way to cluster and data partition the data so that i have a tier of web servers that talk to the data tier, where keys are distributed across multiple data servers (NOT replicated). I am looking for a solution that will constantly and deterministically determine the data location (which data server the data goes to/resides in) without a central provider for the partitioning logic. the reason for the memory engine direction is for high speed, the need for data partitioning is to be able to scale out without being bound by the memory capacity of a single server.

any idea will be welcome.

thanks

Gadi