Table partition in scale out architecture

Hello,

We have a huge table with 857M records which it's key is GUID (In our reports we never select records from this table with the key\GUID) . Our system is scale out of 4 indexservers - 512G each.

We partitioned the table with 2 levels:

1. 1.HASH for the key/GUID with number of partitions equal to the number of hosts = 4 in our case.
2. 13 partitions of range according to the periods.

The table before the partition:

The table after the partition:

Questions:

1. Because the key has a key we were forced to partition according to it although we never use it, Will it improve the performance?
2. The memory consumed by the table was grown significantly from 73G to 107G and also he size on disk. Why?
3. After the partitione finished all the partition are located in one indexserver. I know I can move them but does it make sense? On the one hand the calculation will be parlalize between 4 indexservers on the other hand the traffic of data between the servers will be huge. Because the key of the table doesn’t have any meaning I guess that every select above this table will use all the 4 indexservers and have a lot of traffic.

What do you think?

Thanks,

Amir