I would go with C,E,F.

A - We do not know what tier is being used, and most likely that would not affect performance too much
B - We have no data about the copy command used
C - It could be the Internet connectivity ( @Eikistein, they are not saying that the VPN is complicated - rather the Internet connectivity backing it ) that is adding additional delay
D - Again, we have no data, and this would be very easy to fix
E - This is the one I have doubts about. While a separate storage layer could be slower, we do not have enough data to know whether this is the case here
F - Throughput is limited for CloudVPN

A, B, F ? not sure

C = I don’t think we can say CloudVPN is complicated
E = I think a separate storage layer outside the VM is recommended, anyway I don’t think that would be a bottleneck
D = that could be a bottleneck
--------------
A = all current databases are SQL-based (see case study), so GCS is not a good option here, you got to transform the data…
B = could not find anything to support to that,
F = more tunnels, more throughput

C, E, F should be correct
A - we don't know storage type and it shouldn't impact throughpul
B - we do no know what copy command was used (gcutil and tranfer service or even appliance)
D - adding machines won't improve throughput virtual cores are counted for egress not ingress

But generally this one is complex

Provided Jencomart has 4 datacenters and want to add Asia as another location, BCF should be the answer?

i beleive B, E, F.