QoS and real internet connections

I've noticed that a difference between IP on a regular DSL modem, and IP on a T1 is what happens when you swamp the line. swamp the upload on a DSL connection, and your downloads die. Swamp the upload on a T1, and your in game ping barely increases, and you can still make UMA phone calls, and you don't even realize that your brother is playing robots via VNCoVPN.

That given, what changes when you go to QoS the T1? higher bandwidth % let through?

What's the best way to QoS on a line that doesn't suck?

I'm also wondering, does gaming prioritization work FOR game servers that you're running, or only for game clients (me serving TF2 vs me playing TF2). I'd like to make sure that someone on youtube doesn't create terrible lag for people in games on my server.

Synchronous vs Asynchronous Signaling

As I understand it, typical DSL is actually ADSL, which uses asynchronous signaling. This ties your upload to your download in that, like you said, if you swamp one, yet get poor performance with the other.

A T1 connection is synchronous, though, allowing for simultaneous unhindered upload and download, regardless of how much of either you are using.

I believe the gaming prioritization only has settings for game clients. If you want your servers to be uninterrupted, you'll have to create QoS rules that specify the ports defined by whatever the server is running. For instance, if you are running CS:Source, you'd need to make rules that put incoming and outgoing TCP 27020-27050 and UDP 1200 + 27000-27015 as higher than all other traffic. I'm not sure if the ports for TF2 are the same, but, all it would take is a quick trip to valve's site.

I am under the impression that the difference between ADSL and T1 (mine is provided via HDSL), isn't so much the bandwidth difference, but how the equipment buffers data going out. saturating your upload is saturating your upload regardless of whether or not it's symmetrical. on a 384Kb upload, limiting to 35KB/s upload makes a substantial difference to what comes down, whereas with the 1544Kb line, 100% utilization both ways is possible with significantly less impact.

I have noticed that if i'm moving 75 GB of HDTV across the network (from a separate machine to the one I'm running the game server on), that has a noticeable effect on perceived latency, even though it's a 100Mb/s full duplex connection, probably due to the way packets are scheduled.

As for QoSing TF2, I'd think i'd gain more by having the data bypass the UTM, since I'm only at about 70% utilization in 6v6, QoS would only come into play when someone starts browsing youtube, or checking email, etc. for match games, people get particular about game ping times.

HDSL is NOT a T-1. Those marketing people LIE! Maybe T-1 speeds, maybe symmetrical, but a T-1, a true T-1, is two pair of copper and requires CSU/DSU's on both ends. The cheapest I have ever heard of one is over $280/month

Quote:

---

Originally Posted by mikepb

HDSL is NOT a T-1. Those marketing people LIE! Maybe T-1 speeds, maybe symmetrical, but a T-1, a true T-1, is two pair of copper and requires CSU/DSU's on both ends. The cheapest I have ever heard of one is over $280/month

---

I'm paying $240 / month.
it's real. here's the pudding.

CSU/DSU
http://liquidflame.net/u3b3rg33k/tech/IMG00013.jpg

spare line card.
http://liquidflame.net/u3b3rg33k/tech/IMG00014.jpg

it's only DS1 between the line card and the CSU. AT&T uses HDSL2 to save pairs. if you could get the same amount of data over half the pairs, you'd do it too.

Pardon the mess, I'm redoing the bulkhead so I can run more wires without having to dig up the foundation again and again.

But its really very costly i think so

You have no antecedent. there are 2 pairs between the line card and the CSU/DSU. Anyways, back to QoS?

Should I even be bothering with using UT for QoS, or should I go to pfSense or m0n0wall for this? which side of UT should they be on if I did?