PCA Forums

I ran the comparisons running a Steam download with my bandwidth fully saturated at 3.2 Megabytes per second and again at idle.
I used Speedtest.net, Pingtest.net and an echo request poll using Uotrace of various Game Servers host locations: ( http://www.gameservers.com/why/.)<table class=" cke_show_border" border="0" cellpadding="default" cellspacing="0" height="default" width="default"><tbody><tr><td>Test</td><td>Location</td><td>Avg. Ping</td><td>Best Ping</td><td>Worst Ping</td><td>Download Mb/s</td><td>Upload Mb/s</td><td>Jitter</td></tr><tr><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td></tr><tr><td>ping full load</td><td>Vancouver</td><td>104</td><td>52</td><td>156</td><td>
</td><td>
</td><td>
</td></tr><tr><td>ping idle</td><td>Vancouver</td><td>10</td><td>8</td><td>33</td><td>
</td><td>
</td><td>
</td></tr><tr><td>speed test full load</td><td>Vancouver</td><td>76</td><td>
</td><td>
</td><td>10.44</td><td>2.37</td><td>
</td></tr><tr><td>speed test idle</td><td>Vancouver</td><td>5</td><td>
</td><td>
</td><td>26.75</td><td>2.69</td><td>
</td></tr><tr><td>ping test load</td><td>Lynnwood WA</td><td>112</td><td>
</td><td>
</td><td>
</td><td>
</td><td>19</td></tr><tr><td>ping test idle</td><td>Lynnwood WA</td><td>14</td><td>
</td><td>
</td><td>
</td><td>
</td><td>1</td></tr><tr><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td><td>
</td></tr><tr><td>ping full load</td><td>Miami</td><td>207</td><td>153</td><td>257</td><td>
</td><td>
</td><td>
</td></tr><tr><td>ping idle</td><td>Miami</td><td>88</td><td>82</td><td>106</td><td>
</td><td>
</td><td>
</td></tr><tr><td>speed test full load</td><td>Miami</td><td>166</td><td>
</td><td>
</td><td>4.21</td><td>2.18</td><td>
</td></tr><tr><td>speed test idle</td><td>Miami</td><td>89</td><td>
</td><td>
</td><td>23</td><td>2.76</td><td>
</td></tr><tr><td>ping test load</td><td>Orlando</td><td>211</td><td>
</td><td>
</td><td>
</td><td>
</td><td>19</td></tr><tr><td>ping test idle</td><td>Orlando</td><td>107</td><td>
</td><td>
</td><td>
</td><td>
</td><td>4</td></tr></tbody></table>

I just ran the tests with Windows 8.1 stock settings. There is a huge impact in this worst case download scenario. I would hate to be playing on a game server if a game like Call of Duty Advanced Warfare, picked this computer to host an online match while downloading at full bandwidth.

That's what QOS (Quality of Service) is for, but it's a bitch to setup properly. Took me a couple of weeks to set it up properly on my new router. Not something for the faint of heart.

Yeah I have QOS running. I have one computer that does downloading and its traffic gets low priority and my PC gets max so even when I am downloading I can still play games with a 20-40 ping.

I tried about a 15 second poll of the ping to the Vancouver game server host, with a Steam Download capped at 1 Megabyte per second.
The average ping was 14ms but the best was 8 ms and the worst was 117 ms.
This should be leaving over 2/3 of my bandwidth available but it still has a significant impact on performance.
Has anybody tried polling the pings on the servers after using QOS?
I suspect there still would be a noticeable performance impact when hosting a game server and downloading.

Ironically, the reason that often happens is because the outgoing buffer in your router gets overrun (you may be downloading, but each downloaded packet needs to be acknowledged with a sent packet. QoS also solves that by allowing you to set a max upload cap on the router that's 5 - 10% lower than your max upload speed, to keep the buffer from overflowing.

That's one thing that makes tweaking QoS difficult. You need to do a lot of testing to get a good idea of your actual down/up load speeds.

But yes, QoS would solve your problem, I just will warn you it's hard to setup right. You need to know all the ports your apps use, and there is no universal QoS priority scheme - you need to tailor it to your needs, and then tweak it a bit from time to time.

You also need a router that actually does QoS (and allows you to actually tweak it - not just a QoS checkbox that turns on a hardcoded algorithm.

If you poll an echo request for 15 seconds on one of the game servers, what is the difference between your maximum ping with an idle connection and while streaming a netfix show using QOS?

I'll have to try it when I get a chance and let you know.

FlyingPenguin wrote:I'll have to try it when I get a chance and let you know.

Thanks.
I'm not investigating this because I'm having any problems. When I game online I make sure my network connection is idle. I have Steam set to not download when I am gaming, don't have any download or uploads running and there is no impact for the cable companies to record or stream shows on their personal video recorders.

I was surprised to learn that the telephone companies share the bandwidth with the internet and TV shows. It's 6 Mb/s per high quality stream, which means you would only have 7 Mb/s left over if you were recording 3 shows on a 25 Mb/s connection.

A friend asked the question why does online gaming seem to have more lag when people have 10 times the bandwidth they used to have?
In Call of Duty they used to let you run a dedicated server where you could cap the maximum ping of people allowed on your server.
Now they use public matchmaking where you have no control over who is host, what maximum ping is allowed and have a default ping search of 800. I suspect they have it capped at 18 players because they feel 1 Megabit up and down is a common network connection.

Throw in the fact that lots of people are streaming Netflix and downloading on their network while they are gaming and I think it makes it worse.

It's a good thing Call of Duty Advanced Warfare lets you host a private match on your own PC with AI players, where you have some control over who you let on your server. It can still cause a problem where a friend of a friend could have a terrible connection and ruin everyone's experience with the lag compensation. So far I've only played with people I know so the private matches are smooth as glass.

Okay, here's a nice dramatic result that shows how effective QoS is. I must say I was surprised at the results myself...

Steam download, No Bandwidth Limit set in steam. Ping poll for 15 seconds to Miami Gameservers 104.156.244.241

Download was happening on the gaming PC. Ping poll test was being run from my workstation, so that eliminates any possibility of the computer saturating itself.

WITH ROUTER QOS ENABLED:
Avg Ping=20ms, Best/Worst=16/45ms, Packet Loss 4%
Steam download was being capped at 2.9MB/s by QoS.

WITH ROUTER QOS DISABLED:
Avg Ping=415ms, Best/Worst=16/486ms, Packet Loss 8%
Steam download was going as fast as it could at 3.2GB/s

AND FOR A BASELINE THIS IS WITHOUT THE DOWNLOAD GOING ON:
Avg Ping=21ms, Best/Worst=17/39ms, Packet loss 4%

Now notice the packet loss wasn't all that different (4% vs 8%)? BTW I had 4% packet loss even with no download running, so that's just normal Internet bottlenecking. Yet the average ping without QoS and a download running was a horrible 415ms. That's nearly a half second. It would make modern web pages with the arse-load of scripting they have, load slow as molasses. It would also make online gaming unplayable on another PC. Anything over 250ms is pretty much unplayable IMO.

When QoS is working properly, as it is here, it prevents any one connection from saturating all the bandwidth. In the QoS settings there are usually 10 "Classes": Class 1 thru 10, which designates priority (where class 1 has the highest priority). Some routers use letters. Some routers like mine, actually give each class a name which I can edit and is basically for my benefit. You can see in the screenshots below where the top class "Service" has the highest priority and the bottom class "Crawl" has the lowest priority.

You create rules for types of packets or, more commonly, groups of port numbers or port ranges, and assign them to a class. The router will come with some pre-defined rules (for most chat apps, streaming media, PS4, Xbox, and even some common PC games for instance). I needed to add rules for specific ports I use: my specific games, Teamspeak, a stand alone web cam I use as a security camera, etc. Anything that is not assigned a rule automatically gets treated as Class 10 (lowest) so there's no real need to assign a port for P2P for instance, because if it doesn't fit any other rules it will get lowest priority, and that's what you want.

Then you can assign a minimum and max % of bandwidth to each Class. They come with some numbers already plugged in for you in the default classes, and you can adjust them.

By default max bandwidth is set to 100% (meaning the max it will allot that class is 100% of the bandwidth), because there's no real reason not to allow a single connection to have all the bandwidth it wants if nothing else needs it.

The minimum bandwidth is the magic Ju-Ju number. This is the minimum percent of bandwidth the router will allot to that class. You start with the default which will probably work okay, but you want to tweak it as time goes on.

Also people have different priorities. NO ONE in my house ever uses chat programs like AIM and Messenger, so I decided to give my "Messenger" class a very low minimum of 5% instead of the default 50%. Teamspeak is assigned to the VOIP/Game class which I gave a 90% minimum. Yeah I'm a gamer, so I want my games to get all the love when I play. I probably need to lower that a bit to allow the wife enough bandwidth for Netflix, but I seriously doubt my playing a game would max out the bandwidth anyway, and she hasn't complained so far.

In the screenshots below you can also see there are Max bandwidth limits set for incoming and outgoing. YOU need to calculate this using speed tests, and enter a value 15 - 20% less than that to allow some buffer. I use a value of 15% less than the average I got using several speed tests, and a large download from a very fast FTP server I own.

Notice my incoming and outgoing minimums are the same for all classes except for "Media". I don't stream any video out, I just stream video in, so that gets a higher minimum. The lowest two classes are set to 1% because everything else should have priority (a P2P download can get throttled down to the bare minimum if I need most of the bandwidth for gaming and Netflix for instance).

I'm surprised the QOS results are that good with a heavy download in progress.
I think we've proven our theory that public matchmaking can have abysmal results if the connection isn't idle or running with QOS properly setup.

I haven't seen any packet loss on any of my tests even the ones where my bandwidth was completely saturated. I just tried a Ping poll for 15 seconds to Miami Gameservers 104.156.244.241 with my connection idle.
Avg Ping=105ms, Best/Worst=98/125ms, Packet Loss 0%

Is it just the Miami server you are seeing 4% packet loss or all of them?

I assume there was a bad router between me and Miami. I was going to try a different server when I get a chance.

Im totally lost in all of this....sheesh Im so outta touch in the IT world these days.

QoS is a pretty arcane art. Honestly, I only REALLY learned the nitty gritty when I upgraded to a new router last year. My old Linksys with DD-WRT had QoS enables, but it was just using the default settings which worked well enough for me, except that I usually had to avoid big downloads or P2P while gaming.

The new router's default settings weren't doing it for me, so I dove into the subject and figured it out by trial and error.

I think we're going to start seeing a lot more high-end routers, and custom router firmwares, support adaptive QoS which will basically learn from your habits.