I wanted to write about this problem I ran into in the hopes that anyone else that encounters it will have a much easier time dealing with it. Maybe the devs might have an idea as to what’s going on too…
Anyways, here’s our setup. We have two offices connected together with a 1.5meg point-to-point T1 line for voice and data communications between offices. It is not used in any way for internet traffic. All of our production servers are located in the main office, and our sales staff works from the second office logging in each day to a Windows 2003 Server running terminal server dedicated to doing only that. Spark runs on the terminal server under each users session, and does not run on any of their local machinies in the sales office.
When none of the remote users have spark running within their sessions, and there’s no voice traffic, the T1 utilization runs at about 3-5%.
When the remote users have Spark 2.5.5 running, and no voice traffic, the T1 utilization runs at about 25-30%. Still well within an acceptable range.
However, when we upgraded the terminal server to run Spark 2.5.6, the T1 utilization jumped up to close to 90% causing all kinds of bottleneck issues.
Uninstalling 2.5.6 and reinstalling 2.5.5 caused the utilization to drop back down to about 30%. So there’s something with Spark 2.5.6 that’s causing Terminal Server to greatly increase the ammount of traffic sent to each client loged in. Presumably it’s something with the way 2.5.6 handles screen redraws, but I’m just guessing there.
Hopefully this will save someone alot of headaches.