dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
43714
joelb2
join:2015-03-27

joelb2 to ra88442

Member

to ra88442

Re: ATT Gigapower speeds

Most likely the place you are testing at has saturated their download some due to customers being internet users. This is probably because their end users are sucking down youtube/netflix/dropbox towards the isp hosting the test, so when you try to sent to them for the upload test you have to deal with their congestion.
ra88442
join:2003-03-29
Austin, TX

ra88442

Member

Hi F100, thanks for sending Justin's new speedtest link! Im getting the following error when I tried to access it Unfortunatley I dont have any friends at UT to try your other suggestion
ra88442

ra88442

Member

Hi OpTiC if you look earlier in the thread at the last 3 speedtests I posted (2 from speedtest.net and one from comcast) I noted that these are all with my iMac directly connected to the RG
ra88442

ra88442

Member

Hi rolande, it makes sense what you said:

"However, there are known issues with the stack as you near 1Gig speeds that can cause wait behaviors which pause the stack. It is possible that is what you are running into that is causing the perceived throughput restriction. It is also possible the hardware can't keep up that I/O rate."

If you are correct and it is the hardware that cant keep up with the I/O rate, then why do I see close to 1 gpbs downloads and a reduced speed in the upload. Shouldnt both be reduced due to the preceived throughput restriction?
ra88442

ra88442

Member

Hi iRob, I used a different cable with the R7000 speedtest and the NVG599 speedtest above. Yes it is possible it is the cable still and thanks for the suggestion but my gut tells me it is something else.
ra88442

ra88442

Member

Hi Joelb, do you know of a good speedtest site that I can try to avoid the congestion that you described?

iRob
join:2014-08-19
Mckinney, TX
MikroTik RB3011 UiAS-RM
Ubiquiti EdgeRouter Lite
ARRIS SB6183

iRob to ra88442

Member

to ra88442
I'm feeling this is something else as well. Either it's NVG599 or something on the ONT or anything (just installation specific) or it could be your neighborhood. You have seen uploads go up to 800 and you used 2 different devices, different cable, tested directly to the NVG599. Its either your neighborhood specific or its something with your equipment that was installed by at&t. The other user that was from Austin successfully got above 900 on both download and upload (even though he did say that it took time for it to improve). I've also taken the guess that you have done tests through different times of the day (morning, night, evening, primetime) and if they come out similar, I wouldn't think it be the speedtest server (night one's would possibly give you faster speeds, evening/primetime would give you slower speeds).
I'm just thinking the only fix to this problem might fall back on at&t, either it's the equipment that they installed/the current setup or drop possibly(I'm doubting that it's anything with the ONT or the NVG599 or so), or its something that might be affecting your neighborhood. Whatever it is, I'm feeling like you can't get this fixed by yourself, and it might take few months for it to get it fixed like the other user from Austin had described.
I guess in meantime, test through out the day at different times with different speedtest websites to see if you see any improvement.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande to ra88442

MVM,

to ra88442
said by ra88442:

If you are correct and it is the hardware that cant keep up with the I/O rate, then why do I see close to 1 gpbs downloads and a reduced speed in the upload. Shouldn't both be reduced due to the preceived throughput restriction?

Downloading data for a speed test does not depend on your local hardware for read performance. It is only transmitted across the bus from the network interface to the socket and then sent to the bit bucket after it is timed and measured. Uploading data requires a process to generate data to write to the buffer to then be read to the socket for transmission. That typically uses more local resources to perform that function, depending on how it is written. At speeds above 400-500Mbps with hardware that old, it does not surprise me that you saw throttling occur. However, it is possible due to the delayed ACK problem you are seeing momentary throttling that affects the overall performance but only in one direction.
ra88442
join:2003-03-29
Austin, TX

ra88442

Member

Hi Rolande, I ran the following command then did a speedtest:

sudo sysctl -w net.inet.tcp.delayed_ack=0

I didnt see any improvement with my upload speed. Is this what you were referring to when you said, "However, it is possible due to the delayed ACK problem you are seeing momentary throttling that affects the overall performance but only in one direction."

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

said by ra88442:

Hi Rolande, I ran the following command then did a speedtest:

sudo sysctl -w net.inet.tcp.delayed_ack=0

I didnt see any improvement with my upload speed. Is this what you were referring to when you said, "However, it is possible due to the delayed ACK problem you are seeing momentary throttling that affects the overall performance but only in one direction."

Yep. I assume it was set to 3 by default. Is the upload test providing variable results? Is it always around the same threshold, regardless of time of day? It is possible you are getting throttled by a buffer on the remote end
ra88442
join:2003-03-29
Austin, TX

ra88442

Member

Yes it was set as 3. I tried setting it to 2 or 1 also, but didnt see a huge difference. Yes the upload test was providing variable results. I was running the speedtest after midnight and sometimes I would see uploads in the 650 range but then it would spike into the 700-800 range on other tests and gradually decrease as it ran. There are other folks that are reporting speeds of 900/900 on gigapower so it cant be from the remote end can it? Or do you think my iMac ethernet card needs tuning?
ra88442

1 edit

ra88442

Member

Click for full size
Are the number of receive drops and receive errors on my gigabit line normal?

iRob
join:2014-08-19
Mckinney, TX
MikroTik RB3011 UiAS-RM
Ubiquiti EdgeRouter Lite
ARRIS SB6183

iRob

Member

For safety reasons, you might want to blur out your IP address on that image.
For the response of your question, I can't tell what is and what is not normal for FTTP, sorry. I'm just posting this mostly concerning about your IP address being visible.
Uverse guy
join:2015-03-20
Carrollton, TX

Uverse guy to ra88442

Member

to ra88442
Different servers will have different speeds. I have seen 300 Mbps service get .38 mbps down to certain servers. Not all servers have a true 1G uplink. AT&T can only guarantee the service from the RG to the VHO/SHO. We can not guarantee any performance off our network.

On a side note, anything below 600 Mbps on either the down or up is low. There may be an issue with the service. But, again, AT&T guarantees a certain range for Gigapower going from a company owned laptop meant ONLY to test Gigapower speeds to the IP backbone. Nothing else can be guaranteed.
brianlan
join:2009-10-12
Garner, NC

brianlan

Member

You are correct. When you have such a higher than 'normal' link you really learn what the real world is like.

The fact of the matter is the current internet infrastructure can not and will not support the Giga-Revolution anytime soon. If you thought your 20 Mbps line was throttled along the way, wait to see what your real-world 1 Gbps line preforms at. Yeah, regionally its pretty good, 600+ Mbps down and 300-500 Up, across the country more like 150-250 Mbps down and 50-100 Mbps Up. That's my reality here in NC on 1 Gbps fiber that preforms at 95% (915Mbps Down/930Mbps Up) of rated speeds to AT&T Test Locations, off AT&T's network all bets are off.
ra88442
join:2003-03-29
Austin, TX

ra88442

Member

Hi Uverse guy,

Thanks for the message! Im going to assume you work for ATT My download appears to be consistently at 900 mbps as I expect but upload is averages in the 600 mpbs range which seems to be lower. ALso, I am not just using any speedtest, but the ATT one below:

»www.att.com/speedtest/

Can you explain how the company owned laptop is tuned to a gigabit internet connection? I would expect if a ATT customer has a gigabit ethernet card in your personal computer connected to the RG, it should show the same speed test results as one from the company owned laptop.
ra88442

ra88442

Member

Hi Brianlan thanks for the message. You wrote:, "hat's my reality here in NC on 1 Gbps fiber that preforms at 95% (915Mbps Down/930Mbps Up) of rated speeds to AT&T Test Locations, off AT&T's network all bets are off."

My reality is that on AT&T Test locations, I get 930 Mbps down/600 up. Off AT&T Test locations, I get the same speeds.

I feel if my connection was 100% working then I would see the same results as you where the AT&T Test locations showed 95% of 1 gbps speeds.
brianlan
join:2009-10-12
Garner, NC

brianlan

Member

But your assuming that their is no saturation on your data path to the speed test.
ra88442
join:2003-03-29
Austin, TX

ra88442

Member

Hi Brianlan so if there is saturation in my data path to the att speedtest server (»www.att.com/speedtest/) who would I need to contact to get it resolved? I do not believe there are a lot of people in my neighborhood that has ATT gigapower so do not think it would be due to a lot of network traffic in my area...
brianlan
join:2009-10-12
Garner, NC

brianlan

Member

Its not just gigapower customers, all Uverse customers are sharing the same uplink at some point locally to you.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

said by brianlan:

Its not just gigapower customers, all Uverse customers are sharing the same uplink at some point locally to you.

But they all share the same downstream path too. It all goes across the same physical fiber ports...WDM or not. It still enters the same forwarding queue across the OLT backplane to hit the same uplink fiber to the network distribution and core. The upstream and downstream bandwidths are symmetric. Users tend to use much less upstream bandwidth in general. So, it does beg the question, where is the buffer saturation. Is it on an endpoint or at some transit point in between?

Have you tried using the new DSLR speed test? It uses many distributed simultaneous data streams from different cloud locations to aggregate the data throughput to achieve the high aggregate throughput. If this still exhibits slow performance in either direction at a variety of different test times during the day or night, then I would err towards your client endpoint being a likely culprit. That's not to say some transit hop in between couldn't be an issue but the likelihood of that being the case is very low.

why60loss
Premium Member
join:2012-09-20

why60loss to ra88442

Premium Member

to ra88442
Well now I wonder what AT&T is going to do about Comcast's fiber to 18 million homes by the end of the year plan. Is AT&T going to ramp up gigapower deployment speed and reach or let Comcast run them over with upto 2Gbs internet. (Not saying AT&T needs 2Gbs internet, but I get the feeling AT&T don't do well in those markets if it doesn't really start building out gigapower quickly with more reach.)

Maybe AT&T will just bump users to 2Gbs in Comcast markets or would they need to install new gear to do that if they chose to match Comcast on speed. I really don't think they would need to though, just wondering what that would take.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

The downstream port for GigaPower x-GPON is 10 Gig shared. If AT&T wanted to play the same over subscription game as the cable providers they could easily go 10Gig/1Gig today. But they won't because they engineer to a much different spec than Cable companies. So instead they offer 1Gig/1Gig for now. I will not be surprised if they wave their hand and that becomes 2Gig/1Gig with the push of a button. Eventually GPON will be 40 or 100 Gig ports. Cable will be stuck with DOCSIS and behind the 8-Ball. Comcast is not putting in FTTP. As far as I'm aware it is all FTTN. So, that is a bit of a misnomer for marketing purposes. The word fiber and the meaning behind it is practically more misused and abused than HD.
silbaco
Premium Member
join:2009-08-03
USA

silbaco to ra88442

Premium Member

to ra88442
At&t is actually using xgpon? I was under the impression it was just gpon. That is what Centurylink is reportedly using for their 1 Gbps service.

OpTiC
Premium Member
join:2014-03-08
West Covina, CA

OpTiC to rolande

Premium Member

to rolande
said by rolande:

Comcast is not putting in FTTP. As far as I'm aware it is all FTTN. So, that is a bit of a misnomer for marketing purposes. The word fiber and the meaning behind it is practically more misused and abused than HD.

Comcast deploys FTTH in northeast markets for extreme 505. They are not using pon. That is how they have 2/2 services. This is the ont they are using.
»www.ciena.com/products/3 ··· verview/

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

said by OpTiC :

Comcast deploys FTTH in northeast markets for extreme 505. They are not using pon. That is how they have 2/2 services. This is the ont they are using.

And they are deploying this to 18 million homes by the end of this year?! Bull-dookie.

That is cool they are converting. I was not aware they were actually deploying ONTs. If it isn't PON, what is it? Long haul/Metro Ethernet? I'm curious what the optical Layer 2 infrastructure looks like and how they bring it back to the head end gateway.

why60loss
Premium Member
join:2012-09-20

why60loss

Premium Member

said by rolande:

said by OpTiC :

Comcast deploys FTTH in northeast markets for extreme 505. They are not using pon. That is how they have 2/2 services. This is the ont they are using.

And they are deploying this to 18 million homes by the end of this year?! Bull-dookie.

That is cool they are converting. I was not aware they were actually deploying ONTs. If it isn't PON, what is it? Long haul/Metro Ethernet? I'm curious what the optical Layer 2 infrastructure looks like and how they bring it back to the head end gateway.

Read the press release: »corporate.comcast.com/co ··· take-you

"We'll first offer this service in Atlanta and roll it out in additional cities soon with the goal to have it available across the country and available to about 18 million homes by the end of the year.

Gigabit Pro is a professional-grade residential fiber-to-the-home solution that leverages our fiber network to deliver 2 Gbps upload and download speeds. We've spent a decade building a national fiber backbone across 145,000 route miles of fiber. This new service will be available to customers that are within close proximity to our fiber network."

That is what Comcast is claiming it is going to do. Any ISP in the wake of that better think a plan and quick. It isn't 18 million just passed, it is 18million that will be able to order service by the end of the year.

Comcast has put a unheard of number homes on a FTTH upgrade in an unheard of amount of time. We only need a few months to see if this is in fact BS, until then AT&T and other ISP's for that matter might want to think of a reply to this mass fiber to the press release.
lkmd
join:2002-02-03
Dallas, TX

lkmd to ra88442

Member

to ra88442
I have compared browsers and OS. 10.9 faster upload than 10.10.
Google chrome on 10.9 I got 920/915. Chrome and OS 10.10 920/350. ! Not sure what Yosemite is doing to upload but it dose reduce speeds. I also noticed that the airport devices reduce speeds so I switched to ASUS routers and not getting max speeds.
Something is the router and OS software is reducing speeds. Im not sure gigapower is worth the extra cost....

rtfm
join:2005-07-09
Washington, DC

rtfm to why60loss

Member

to why60loss
said by why60loss:

Well now I wonder what AT&T is going to do about Comcast's fiber to 18 million homes by the end of the year plan.

The same thing they are doing about the aliens on the flying saucers headed for Timbuktu... and for the same reason.
Expand your moderator at work