I have a three-node Windows NLB Server 2012 R2 DirectAccess farm. These three single purpose servers have good specs (8 cores, 32GB RAM, etc etc). The problem that I am seeing is that the clients all have very low throughput on each session (between 6 and 8 MBit "aka 1 MByte" per second). This performance spec is linear since for each concurrent client that you add their throughput is also in that range. Clients are all high-spec notebooks with Windows 8.1 Enterprise x64. Performance on these clients is excellent except when transiting the DirectAccess server. If the same client connects through AnyConnect VPN their throughput is excellent. Additionally, when clients connect to the DA server plugged into the external traffic switch (aka same network as the DA external interface) the performance is identical so this isn't a WAN performance issue. The utilization on all devices (DA servers, DA clients, network hardware) is very low so it does not appear to be a resource problem. I confirmed that NULL CIPHER is used on clients so the traffic isn't being double encrypted. This NLB started life 2 years ago as a Windows Server 2012 environment on different hardware and I've had the exact same issue. It works "ok" but not the throughput that the capacity planning documentation indicates.
Any ideas?
Thanks,
Mark Ringo