Anybody Running 1 GBPS At Home?

[alarmguy]

Ok, tried the first test and turned off everything on the firewall, ran the test over and over to see if I could get the same results both with the firewall on and off. At first I thought there was a 10% difference with the wall off but then I could not repeat it. So safe to say no effect. Peak buffer on the worst test was 775 but most all others both with the firewall on and off was between 650 and 700 when on the download phase of the test.
On the standard test with fusion, no change in the ping, jitter or download speeds. Of course the high buffer was only on downloads, being upload speeds are only 12Mbps the buffer reading was almost zero most all times and if I had to guess, less then 10ms even when it wasnt close to zero.

Interesting thought about the Archer c7 and funny as I said to my wife early for fun it would be cool to hook it up and run the tests. Whether I ever do it I dont know, I get off work early most days but wife works from a home office sooo .. will see ... I did watch a video on the buffer thing and some really quick reading. Im not in anyway an IT guy. I was good at getting customers wifi cameras to work properly and known for trouble shooting in my company,(aka going back on other peoples installs) at the same time, we didnt play IT nor have the knowledge for it and in unfixable cases told customers to call their internet provider OR if a business, their IT company/dept. or get a new router and call us when you do. I am sure you can imagine the garbage in some households systems.

Anyway, if I understand it correctly and think as you stated, the buffer is really how quickly the in this case, consumer router can sort and get the information to the computer. SO it would be interesting to see how the older Archer C7 handles this. I do have one question, from what I THINK I read, isnt it possible it could also be a high buffer if the computer equipment itself can not accept the information as quick as the router can deliver? I know maybe more unlikely but ?

AS far as consumer grade $99 Motorola MR 2600 Im thrilled with its performance in our home. The reality is we aren't down loading huge files or even a combination of information from our 20+ devices all at one time. I do find it interesting, love stuff like this.
(I did turn the firewall back on after the dozen or so tests that I ran) Except DMZ I left off
Ps. I have also noticed the router settings do not correctly adjust or show the correct settings when using Safari on the Mac, I confirmed it on my wifes Windows Computer with Edge and Chrome AND also with Opera on my Mac which all display and behave correctly.

 

[OVERKILL]

Anyway, if I understand it correctly and think as you stated, the buffer is really how quickly the in this case, consumer router can sort and get the information to the computer. SO it would be interesting to see how the older Archer C7 handles this. I do have one question, from what I THINK I read, isnt it possible it could also be a high buffer if the computer equipment itself can not accept the information as quick as the router can deliver? I know maybe more unlikely but ?

Given your Mac is MUCH newer than my 2010 vintage beast, I think it is safe to assume it isn't the source of the issue. You could definitely try it on another computer though within your LAN just to rule that out.

But yes, that's what that parameter measures, the rate at which your device doing NAT/PAT can translate packets passing through it with your connection under high utilization, that's why the upload side yields significantly better results as the device is under a fraction of the stress due to the lower available bandwidth.

Here's the results through an MX84 on a 200/200 fibre link w/5Mbit reserved for SIP:

I'll be able to test an MX84 on a 500/500 (its limit) in the next few weeks if you are interested.

 

[Brybo86]

I've had comcast 25/2 for the last 10 years.
Typical speed test is 30/2.5 ping 8-10ms.
Using Asus RT68U 2.4/5ghz
Same modem the entire time too. Sb6120
I pay 19.99/mo, no taxes.

We have unlimited 200mpbs+ LTE via phones at home.

Only thing that uses home internet is the
TV for streaming
PS4 game once a year (I'm 34 with 3 kids)
computer email/banking
or an ipad zoom call.

Only time I may appreciate the faster speed would be when downloading a very large app/game via wifi

I do plan on paying off my mortgage in 7 years though.

 

[ZeeOSix]

But yes, that's what that parameter measures, the rate at which your device doing NAT/PAT can translate packets passing through it with your connection under high utilization, that's why the upload side yields significantly better results as the device is under a fraction of the stress due to the lower available bandwidth.

Using http://www.dslreports.com/speedtest my bufferbloat is basically zero on download, and around 250-290ms on upload. Gives a "C" score for bufferbloat.

 

[OVERKILL]

Using http://www.dslreports.com/speedtest my bufferbloat is basically zero on download, and around 250-290ms on upload. Gives a "C" score for bufferbloat.

Maybe try another browser? Unless your router is doing something funky, it should be the other way around.

 

[ZeeOSix]

Maybe try another browser? Unless your router is doing something funky, it should be the other way around.

Same results using Firefox, Edge and Google Chrome. Snap-shot is after running in Google Chrome.

 

[alarmguy]

Thanks for your responses Overkill. I ran the same tests on my old (but not that old) gave to my wife HP desk top (not sure what Gen i5), same results as my Mac with one exception. With that exception I got to see the buffer bloat go lower on her HP. Normally her download speeds are within 10% of mine if not the same but a little more variable.

Anyway, on one of her test runs download speed dropped to 88 Mbps and the buffer bloat dropped drastically with it so I got to see the correlation. By drastically I mean (just grabbing at a number here) around 250 from over 500ms. So actually I guess safe to say by bloat looks ok but not great up to around 90 mbps (again for a consumer router)
Also of course with Spectrum my 100 mbps service is only 12 Mbps up and the bloat meter is only 0 to 5 at any given time.
All the tests I have done involved different browsers Chrome on Windows possibly edge, cant remember, Opera and Safari on Mac

I think its just a matter of time that I hook up that old Archer C7 for fun, that I just replaced and see what numbers it pulls.

(unrelated) I love this Mac mini (2018 ver) bought fall of 2019. Moving Apple into my world has been terrific after decades of Windows. I have to admit, I am almost looking forward 1 year and have a feeling I am going to give my wife this Mac mini and purchase a new Mini with the new M1 chip for fun. I was even thinking of a MacBook Air for Christmas (we both use laptops as well when not on our desktops) plus wife has Dell Optiplex for her office. Anyway, cant bring myself to do it, on a laptop having a touch screen is a must have for us and most likely go for a Windows laptop to replace my wifes or mine. Im actually thinking of a Windows branded Surface Laptop 3 or Lenovo Yogi around the same specs. Had a bad experience with Lenovo not to many years ago regarding a desktop for my wife, returned it and got a refund. *L*
HP envy and Dell xps not out of the question. Except for Apple I would like to keep laptop under or around $700 right now BB has Surface on sale 599. The laptops are just used to go "shopping" and reading so really just need to boot up fast, change programs fast and our current ones getting a little long int he tooth but still useable for another year if we dont this Christmas.

 

[Skippy722]

I played around with a Netgear AX4 dual band and X6S tri band. No matter what I tried with the X6S it wouldn’t go below 1,500ms! I forget the Asus exists, since installing it back in August I haven’t had a single issue with it. Current uptime is 25 days due to a power outage.

 

[OVERKILL]

Same results using Firefox, Edge and Google Chrome. Snap-shot is after running in Google Chrome.

View attachment 34204

Now THAT is interesting! What are you running for a router? Are you using QoS on it?

 

[wwillson]

Same results using Firefox, Edge and Google Chrome. Snap-shot is after running in Google Chrome.

View attachment 34204

Idle is not zero queuing delay, but how could it be, as they have to send packets to measure if there is queuing delay. We can easily calculate the delay queue depth if we know the upload shaper speed, packet size, and packet rate. Remember that any time a packet is being inserted onto the physical layer, there is technically congestion because no other packets can be inserted and the queue will grow.

Download queuing delay is small because the download direction isn't very congested. There can be several reasons it isn't congested, but we can clearly see that it isn't.

Upload queuing delay is large because the the upload queue is congested. Because of our relatively small upload shapers, we have to use tools to try to reduce the upload queue depth as much as possible. I upload backups to an online service regularly, so I set the max upload speed to about 50% of my Comcast shaped upload speed. This is what I call poor-man's class of service, but it works relatively well as it leaves 'room' for other applications. Another great tool is differentiated queue on routers, but it's hit or miss whether your router has that capability.

 

[OVERKILL]

Idle is not zero queuing delay, but how could it be, as they have to send packets to measure if there is queuing delay. We can easily calculate the delay queue depth if we know the upload shaper speed, packet size, and packet rate. Remember that any time a packet is being inserted onto the physical layer, there is technically congestion because no other packets can be inserted and the queue will grow.

Download queuing delay is small because the download direction isn't very congested. There can be several reasons it isn't congested, but we can clearly see that it isn't.

Upload queuing delay is large because the the upload queue is congested. Because of our relatively small upload shapers, we have to use tools to try to reduce the upload queue depth as much as possible. I upload backups to an online service regularly, so I set the max upload speed to about 50% of my Comcast shaped upload speed. This is what I call poor-man's class of service, but it works relatively well as it leaves 'room' for other applications. Another great tool is differentiated queue on routers, but it's hit or miss whether your router has that capability.

Excellent detailed explanation.

So how does that tie into what we are seeing here? Well:

As I noted previously, the purpose of the bufferbloat test is really to highlight hardware limitations at link saturation, which, as you've noted, is due to excessive queue depth as a result of insufficient resources to maintain that at a reasonable level. How that typically manifests is in what alarmguy is seeing, as consumer connections are by and large asymmetrical with significantly higher bandwidth available on the downstream channel versus upstream. This puts a much higher load on the unit and will expose its inability to sort incoming packets without imposing undue delay in the process. The queue gets quite deep as the CPU gets overwhelmed. The upstream, having far less bandwidth available, doesn't stress the unit as much and the queue depth is kept at a reasonable level so latency stays low.

Most consumer-grade routers advertise the ability to push X number of packets but there's absolutely no mention of how the unit is actually going to perform when doing so. A good example using non-consumer hardware to illustrate this is to look at a Cisco ISR. You can take an 860, which might be recommended for ~25Mbit and push mid 50's through it, and it will do it, even doing NAT/PAT and some basic firewall rules. But, if you monitor the CPU when doing so, you are pegging it. You don't lose packets, bandwidth appears OK, but the queue depth gets quite deep as the unit simply lacks the processing power to keep up. Same can be illustrated on a more powerful 1921, which has Gig-E ports. It'll NAT/PAT 200Mbit, but it won't do it very well, as you'll be hitting high 80's on CPU utilization and that reflects back in latency. But raw throughput is still excellent, and the unit will provide good, low latency when not trying to move packets at that rate. Going back to the 860 for a moment, if you were to have an 80Mbit symmetrical connection attached to it and you ran something that pegged the CPU, like say downloading a massive ISO, you'd run the unit into its CPU limit. Then, if you tried to browse the internet, you'd find it noticeably delayed, this is despite you not actually saturating the limit of the connection provided to you or the connection medium (100TX).

ZeeOSix's results, in this context, don't make sense. He's experiencing high latency on what I assume to be his lower bandwidth leg (assuming a typical asymmetrical connection), so either his router is doing something strange with the upstream path (QoS?) or there's something else in that path that is delaying upstream packets, like a software firewall on his computer for example, something between the browser and the router. If done from a different computer with no firewall/AV software in play, those results would help determine if the issue is with the device (router) or something on his local PC.

 

[ZeeOSix]

Idle is not zero queuing delay, but how could it be, as they have to send packets to measure if there is queuing delay.

Idle bufferbloat is not zero, as the graph in Post #46 shows a small sliver of bufferbloat at idle.

The speed test was done on my Win10 laptop via WiFi. I'll try the same test on my Win7 desktop via Cat5 cable connection.

 

[ZeeOSix]

Now THAT is interesting! What are you running for a router? Are you using QoS on it?

It's an ActionTec modem/router on DSL service. Not sure what the QoS is set to, I'd have to look.

 

[OVERKILL]

Idle bufferbloat is not zero, as the graph in Post #46 shows a small sliver of bufferbloat at idle.

The speed test was done on my Win10 laptop via WiFi. I'll try the same test on my Win7 desktop via Cat5 cable connection.

Excellent. Also, are you running any software firewalls or antivirus that has an integrated one?

 

[OVERKILL]

It's an ActionTec modem/router on DSL service. Not sure what the QoS is set to, I'd have to look.

What's your upstream bandwidth BTW? Is this ADSL2+ or VDSL? One of my clinics is fed by a 58/58 symmetrical VDSL link, that's why I ask

 

[PandaBear]

The way I see gigabit internet: you cannot use that much without blowing pass the data cap anyways. If you are not connecting between 2 places that both have this speed (at most only 1 can be cable, because upload is capped, so you need cable-fiber or fiber-fiber) it is sort of pointless.

Even if you are torrenting stuff, you need to upload enough to get the download that can saturate that big of a pipe. No video content delivery will waste so much bandwidth to stream you stuff that will use that much data.

I think what remains as the only valid use case over a 300mbps or so speed, is if you are supporting a large facility like school or office. In that case you have higher priority than just speed, like security and reliability.

 

[wwillson]

Panda,

For residential you are correct.

For enterprise, it's a completely different ball game. There are many enterprises with 10G internet that fill those pipes with hundreds or thousands of users and even 100G internet connections are becoming more common.

 

[Pew]

The way I see gigabit internet: you cannot use that much without blowing pass the data cap anyways. If you are not connecting between 2 places that both have this speed (at most only 1 can be cable, because upload is capped, so you need cable-fiber or fiber-fiber) it is sort of pointless.

But it's nice downloading all the Call of Duty updates (seriously, like 50GB updates ever other week.)

 

[alarmguy]

This has been a really great thread, Im interested in this stuff, learned a thing or two as well and IF I ever take the time will experiment, hook up and see what the older TP Link router does as far as buffering. Im not a gamer and extremely happy with my setup and service but always looking to see how I can take it up another level if there is one at the same cost. Of course the most simple was proper placement and setup plus what I believe to be a super cheap icing on the case and installation of a ton of snap on ferrite noise filters on everything electronic in the home, even the doorbell for the doorbell camera, everything and I mean everything ... *L*

The reason I became a fan of these is because I saw it actually fix an issue with my security system. Two Radio controlled security system touch pads were not arming and disarming the system reliably due to low signal strength as shown on the signal meters built into the pads, I pulled them off the wall and installed the ferrite filters on the power leads and signal went up to all 5 bars from 1 to 2. Been years now, never another issue.
Since then, I have them on anything and EVERYTHING electronic in the home and we have a lot *L* I have since doubled and tripled some of them, even put them on the electric outlet wiring in the wire box where a power line adapter is plugged into my wifes office work computer. Again, I saw it work there, not as drastic as the security system but we would get on any given day anything from a full green light to sometimes yellow on the "signal quality" of the Netgear power line adapter, since installing the ferrite, always green.
Im skeptic by nature yet, then I started noticing these Ferrite things come on the power cords from the factory of some expensive stuff, like our SONY TV has a huge one and much other stuff, also noticed currency counters at banks, ya da, yada ... data cables, quality HDMI cables have them too. So I knew it wasnt a snake oil thing even though I already saw the results myself. My other security cameras pull up the images so fast that I have to believe the installation of the filters at time of install also help. Anyway, needless to say I have multiple ones on eery single cable that has to do with internet as well. *L*

To Skippy, interesting feedback on the Netgear. I was REALLY trying to buy American, have nothing against Netgear but I wasnt reading what I wanted to read/hear about the Netgear plus throw in the higher cost and why I ended up with the Motorola MR2600 which of course is own by China based Lenovo.

 

[ZeeOSix]

Excellent. Also, are you running any software firewalls or antivirus that has an integrated one?

Just running Windows Defender and Maywarebytes Premium. I've never changed any major settings in them. I think I have my modem/router set to like a medium security level for the firewall setting. Wouldn't a firewall effect downloads more than uploads?

What's your upstream bandwidth BTW? Is this ADSL2+ or VDSL? One of my clinics is fed by a 58/58 symmetrical VDSL link, that's why I ask

I think it's ADSL2+ (two pairs of phone wires are used), and it's only 60 Mbps download and 5 Mbsp upload speed.