My speed tests do not reflect real world speeds, especially for uploads

Let me make sure I understand the situation.  When you do a speedtest or file transfer to any other machine except the one in TX, you get full speed?  And you're convinced the speed issue between you and that box in TX is an AT&T problem and not your equipment or the equipment in TX?  What happens if the person in TX initiates the upload/download tests?

No this problem happens with a majority of my targets.  Commercial devices like Netflix, Apple, Speedtest.net work as they should.  When people try to connect to my server, or I try to connect to another user’s server, upload really takes a dive.  It’s like MSS clamping is going on here.

I can send diagnostic reports to Apple for example which uploads gigabytes of data to them and it will saturate my upload speed.  When my private server, which I’ve tested is able to fully saturate a Gigabit line, the speed the data arrives at is no faster than 30 Mbps.

Ok, got it, thanks for the clarification.  I'm far from a network expert (and I didn't stay at a Holiday Inn Express last night) but my understanding is that MSS clamping happens when there is a mismatch in the MTU between the ends of a connection like you're testing with.  I'm not sure what sort of setup could be in that path to change from the standard 1500 unless it's a misconfiguration.

What sort of networking equipment do you have at your end and at the one in TX?  Which gateways, any other higher end networking stuff like Ubiquiti or Edgerouter, etc?

I'm doubtful you'll get any useful technical info in this forum from an AT&T person.  This is network level stuff and those people tend to hide from the general public because they don't want to be bothered with questions like this.  I can't ever recall seeing someone at that level here.

I think you might have a better shot at an answer, at least from people who are experts in this stuff and perhaps from an AT&T network tech, if you post in the DSLReports Uverse forum (https://www.dslreports.com/forum/uverse).

My thought is that one of the many hops happening in the route might be congested or overloaded and thus causing slowdowns.  A node that is not used when connecting to Netflix, Apple, etc. I suspect.  

My hope reaching out here is that I might be pointed in the direction to reach out to someone who can actually look into this issue rather than a scripted person.

I personally use an ASUS RT-AC68U and a FreeBSD server.  The TX is also a FreeBSD server running on a gigapower connection.  I should be able to at the very least upload to it as fast as I download from it.  But my upload is significantly slower than download.  That just shows a big problem in transit somewhere.

Which gateways are in play on both ends?

Do you see anything odd with a traceroute to the other server?

Yeah, I think your hope has a better chance over there.  There are AT&T NOC people who lurk there and will jump in if they see something interesting to them (or if they think you're so far wrong and are unfairly accusing AT&T of an issue ).

Wow, NVG589?  How long have you had fiber?  I think it's been a long time since they used that box.  Is it FTTP?

Weird about the traceroute.  Does that happen no matter where you're tracing to?  Is there any commonality about where they fail?

@Cyberpower678 

Huge mistake to use such as old device.  Get the BGW210 back. Put it into passthrough mode.  

As for the bottleneck issue, I'd try the same test with a quality VPN and see if performance changes.