I just did my own tests with SEP 14 MP1 (14.0.2349.0100). Created two identical Server 2012 R2 VM's with 2 vCPU and 4GB memory each. Our hosts have two 10 Gb data interfaces, and the VM's use the VMXNET 3 interface, which shows a 10 Gb connection.
My test conclusions:
1) Running all components with their policies had noticeable effect on throughput, reduced to about 1/3 original throughput.
2) Withdrawing IPS, Firewall, and ADC policy gave some improvement, but still only about 1/2 original throughput.
How I tested and results:
I used LAN Speed Test (Lite) from one server, with the target being the C$ admin share on the other server and a packet length of 1000 MB.
First test before SEP installed:
Writing: 3000 - 3700 Mbps, Reading 2500 - 3200 Mbps Readings varied fair amount, our hosts do have a good amount of load on them.
Install SEP with all server components (no email scanners) on both servers:
Writing: 900 - 1100 Mbps, Reading 900 - 1100 Mbps Readings more consistent
Withdraw all policies except for Virus and Spyware Protection, LiveUpdate, and Exceptions. (No exceptions made for these servers):
Writing: 1400 - 1600 Mbps, Reading 1400 - 1600 Mbps Readings more consistent
Uninstall SEP and reboot: Readings were same as first test, before SEP was installed. Nice to know that uninstall didn't leave any slowdown components.
Install SEP again with only Virus and Spyware Protection: Readings about same as first test (no SEP installed).
Of course your tests may vary, and this is for one large file transfer, but i thought it was interesting and decided to share it. Maybe others could share their testing results?
Robert.