Hey! Thanks for writing! On Friday, October 10, 2014, hustjackie <[email protected]> wrote: > > 1. in cluster mode, if one node has crash, what's the failover? is there > one mechanism to guarantees fault-tolerance? >
We currently support coarse-grained fault tolerance via program restarts. This will be soon extended to more fine grained fault tolerance, where only necessary parts of a program are re-executed. > 2. is there a performance report? > There are some new performance numbers coming up soon. I think there are also some older numbers from Robert. Let's wait what he has to say. – Ufuk >
