Jump to content

Not getting max speed


Recommended Posts

I have a 50/50 connection, and I seem to be getting low speeds with all speed tests. All of these tools and FAQ info seems to be worded quite well, so it seems that the makers of testmy.net know what they're talking about.

 

Anyway, here's my tracert to the dal server

 

Tracing route to testmy.net [75.126.77.87]
over a maximum of 30 hops:
 
  1    <1 ms    <1 ms    <1 ms  10.0.0.1
  2     <1 ms     1 ms     1 ms  66.x.x.1
  3     2 ms     1 ms     1 ms  x.x.x [x.x.x.x]
  4    11 ms    12 ms    12 ms  te-9-4.car4.Chicago1.Level3.net [4.71.102.197]
  5    13 ms    13 ms    11 ms  ae-31-51.ebr1.Chicago1.Level3.net [4.69.138.30]
  6    13 ms    12 ms    11 ms  ae-6-6.ebr1.Chicago2.Level3.net [4.69.140.190]
  7    12 ms    12 ms    11 ms  ae-1-51.edge4.Chicago3.Level3.net [4.69.138.134]
  8    14 ms    12 ms    12 ms  te2-5.bbr01.eq01.chi01.networklayer.com [4.53.98.26]
  9    13 ms    13 ms    13 ms  ae7.bbr01.eq01.chi01.networklayer.com [173.192.18.170]
 10    34 ms    35 ms    36 ms  ae20.bbr01.eq01.dal03.networklayer.com [173.192.18.136]
 11    36 ms    35 ms    35 ms  ae6.dar02.sr01.dal05.networklayer.com [50.97.18.193]
 12    76 ms    40 ms    36 ms  po2.fcr04.sr03.dal05.networklayer.com [173.192.118.151]
 13    35 ms    36 ms    37 ms  75.126.77.87-static.reverse.softlayer.com [75.126.77.87]
 
Route seems quite nice. I did a speed test to the Dal server, and kept ping running during it.
 
Ping statistics for 75.126.77.87:
    Packets: Sent = 254, Received = 254, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 34ms, Maximum = 38ms, Average = 35ms
 
jwyuL4E.png
Looking at the numbers I was getting on my network graph, it pretty much lines up with the result, so that's good.

 

 

Here's my upload

HA4t2K8.png

 

Something is wrong with this result. My below graph shows my upload as green. As you can see, it was VERY stable. It was a hair over 11mb/s, but my result was 5mb/s. I noticed after my upload bandwidth dropped to 0, my test didn't finish. It sat there for as long as the upload took.

https://lh3.googleusercontent.com/-OgQqcrlSxjE/Uo69emC7_MI/AAAAAAAABZA/I2TOMFneU60/w157-h223-no/TestNetUpload.png

It seem it counted the time the test it sat idle as part of the calculation for throughput. So about 24 seconds of 11mb/s and another 30 seconds of 0mb/s, averaged out to about 5mb/s, my final result.

 

 

Here's another photo of my network graph of me downloading a file from NewYork City around 9pm, many months ago.

 

 

https://lh6.googleusercontent.com/-8jeh-Z2kQpM/UZV1EcR6KzI/AAAAAAAAA98/88GxCu3M6NU/w159-h230-no/DownloadGraph.png

 

Here's a photo of my network graph when I tried to load test. I was downloading from Steam around 40mb/s, then loaded up uTorrent and tried seeding about 100 Linux ISOs. This was also around 9pm.

 

https://lh6.googleusercontent.com/33nNmK6Wz2kvpwVFGDpTNKOa3rNxEVV9eVJwhAXRS00=w440-h207-p-no

 

 

Here's the results of running ping against YouTube for the past 4 days. Btw, my ISP does not peer with anyone or has any caching servers. All data is over their trunk.

 

Ping statistics for 173.194.46.41:
    Packets: Sent = 337658, Received = 337415, Lost = 243 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 10ms, Maximum = 26ms, Average = 11ms
 
I'm curious as to why my up and down speed tests are so low.
 
Thanks   :smiley:
Link to post
Share on other sites

Not sure if these links will help or confuse you more.

 

ICSI Netalyzr http://netalyzr.icsi.berkeley.edu/ will show a lot of info on ports that could be blocked and additional network info about my head. Once you run the test

 

For info on your ISP check out Hurricane Electric BGP Toolkit http://bgp.he.net/ this will lookup you ISP and show you how many peers they have.

 

I am sure other others here might be able to provide better direction but thought I could at least provide you what I found in learning about network issues.

Link to post
Share on other sites

Not sure if these links will help or confuse you more.

 

ICSI Netalyzr http://netalyzr.icsi.berkeley.edu/ will show a lot of info on ports that could be blocked and additional network info about my head. Once you run the test

 

For info on your ISP check out Hurricane Electric BGP Toolkit http://bgp.he.net/ this will lookup you ISP and show you how many peers they have.

 

I am sure other others here might be able to provide better direction but thought I could at least provide you what I found in learning about network issues.

 

bgp.he.net/ does not show a peers tab for my ISP. It does show it for many other ISPs.

 

For the Berkeley test, I got all green except my computer is 5 seconds slow and my router may have some vulnerability. I really want that Netgear Nighthawk.

 

I get almost my full 50mb/s from YouTube during peak hours, and that is not from a local CDN, but from Google in another state.

 

As for blocked ports and other things, my ISP proudly advertises that they do not block any ports, except 25 but removable upon request, do no traffic shaping, and no QoS. All traffic is best effort, even their own IPTV and VOIP services. They do not reserve any bandwidth for their own services. Except network management protocols of course.

 

Back when you needed to peer with or host a Netflix cache server, I asked my ISP if they had any plans to sign up to OpenConnect so I could watch in SuperHD. After a few days of escalation, I got a response from a higher-up who told me they had no plans to peer with or install a Netflix cache server. He said they have plenty of bandwidth and do not want to favor one company over another.

 

It seems their general attack plan is to just brute force everything with more trunk bandwidth from Level 3. I'm not complaining.

 

I talked for 45 minutes with a Senior Network tech who had been with them for over a decade. He told me they played with over-subscription and traffic shaping at one point, but it was a huge head-ache to manage properly and made unhappy customers when things went bad. So they switched over to a non-oversubscribed dedicated network. He said network management is just so much easier when you're not worrying about bandwidth. Cuts down on support calls to.

 

I asked him about their trunk, he told me they have not had any issues with congestion for a long while. They typically upgrade well before they get near capacity of their main link. They also have a secondary and tertiary link. He said they are setup such that they can bond their primary and secondary links, and could re-route some over their tertiary. Since their internal network is all dedicated and their trunk is not congested, my only limitation of speed is with Level 3.

 

Who am I to argue?

 

What I would like to know is how they keep buffering latency down. My ping is about 12ms to YouTube in Chicago. If I overload my connection with uTorrent, my ping might spike to 20ms, but that's about it. After that, it's the beginning of packet loss, no buffer bloat here. I disabled QoS on my router because it's no longer an issue. The difference between 12ms and 20ms is only 8ms. I don't understand how there can only be 8ms of buffer

Edited by CA3LE
page layout broken -
Link to post
Share on other sites

Upload issue has been fixed: https://testmy.net/ipb/topic/31239-upload-inaccuracy-fix/

 

My upload and download speed tests are now showing about the same. Still lower than my line rate, but stable and match what my network graph shows. I love this aspect, "other" speed tests I have seen my network graph showing 20mb-30mb, and the speed result listing 60mb.

Link to post
Share on other sites

ROFL.. I'm a scrub... Long story short, I was playing with my TCP settings a long while back and forgot to set them back. The reason I was playing with them is I could download 50mb/s from the Internet, but transferring a file across my 1gb network was only about 1mb/s.. not bytes.. bits. Something with Window's default TCP settings was not playing nice. A quick change and I was up over 110MB/s or near 980mb/s.

 

 

BTW, 33MB upload sizes does not seem to be enough for me to get a good consistent speed. It takes a second for my speed to ramp up the TCP send window, but at 4.8MB/s average, the 33MB is only a few seconds.

 

fwyEULO.png

 

My download is better, but still not the best

 

JTbY5qx.png

 

What tipped me off to check my TCP settings was I found another speed test site that wasn't speedtest.net, and also claimed to use HTML5. I found it strange that my upload had the same flat packet-shaped look and almost the same speed. That was way too close to be a coincidence.

 

SpeedTest.Net did not show me this issue... How I hate that place. It's "fun", and kind of useful, but very superficial.

 

Looks like the CDN gives me a better download, but the upload isn't as strong

 

L6XaSCr.py6ed1V.png

Link to post
Share on other sites
×
×
  • Create New...