Sharemind 2.0 reaches a million operations per second

Dear Sharemind enthusiasts, 

We thought that we'll give you a feeling on how fast Sharemind private computations can be. To do that, we tested the new computation protocols on our new Sharemind experiment cluster. The cluster has plenty of CPU, RAM and fast communication and while the conditions are hard to find in the real world, the real world is catching up.

We benchmarked all the private operations and here are the results:

We also plotted the relative speed of a single operation depending on the input size.

Note that both axes of the plots use the logarithmic scale. 

The most important conclusion of these experiments is the fact that in ideal conditions, Sharemind can perform hundreds of thousands of private operations in a second. For share conversion we actually hit the 1 MIPS mark! Additionally, notice that Sharemind can privately process very large input vectors, making it usable for larger data mining tasks.

A more specific analysis of the performance together with the protocols achieving this will be published in the future.