I have a fairly simple home network, hardwired with gigabit switches, with a wireless router that generally handles our laptops. I have a situation in which my network - wired AND wireless - becomes saturated when I really don't think it should:
When copying a large, or many small files - on the order of 5GB total - from one of the laptops to the main file server on my network, it will, after just a few seconds, bring the entire network to its knees. A video being streamed from a server across an entirely hardwired path will start to stutter and gasp until the file transfer ends.
The file transfer over wireless maxes out at approximately 2Mbps. Let's assume the video - in this case, a 480p stream, might at most want 1.5Mbps, for a grand total of a whopping 3Mbps load on the network. The wireless laptop on which I write this speedtests out on the big, ugly Internet at over 80 Mbps, and I realize that's a bit of an idealized number. But the broader point is that 80 Mbps of even theoretical available throughput shouldn't be taxed by a 3Mbps load. Yet mine sure seems to be.
Is this indicative of some misconfiguration on my part? Or am I just being a dunderhead and not understanding some basics about how data flows on the network?
EDIT: Per comments below; the only hard server on the network is a box running Ubuntu; it has two gb nics, one to the outside world, one to the inside, and it hosts two VirtualBox VM's. One VM hosts media files played back by small RaspberryPi boxes running OSMC; the other VM is the firewall box. The Ubuntu box is one I built about three years ago; Intel i5-4460, 8GB RAM with a single 2TB WD (ST2000DM001) hard drive
EDIT 2: As a test, I started streaming two HD video streams and both played flawlessly. This would reflect pulling about 3x the data from that VM across the network as in my original scenario. Are we suggesting that data going to (writing) the server while streaming is the problem? VMs are going through taps in a virtual bridge to access the network; if those taps are just implemented as file descriptors, could the two-way flow of data across that single tap be the bottleneck?