jimharle

So I've been testing our office Comcast Business 150/20 recently, and it's been performing adequately. Single-thread TMN downloads I can't get to go above ~108 Mbps for the average, but they do peak at close to 150. We can consistently fill it up with multi-thread downloads, and almost get to 170 (testing our own CDN and other third-party downloads). This is good for us, but alas I don't have room to complain to Comcast about performance. Still, it does indicate that TMN measures various Comcast connections reliably.

Well, corporate people understand spreadsheets, and that is how we initially presented the problems with our last provider...listing ten "known fast" download sites, and how our results compared to other ISPs we had access to. We could always do something similar with participants here, as many TMN "members" use Comcast - come up with a matrix of how your performance differs from others using the same download sites. But then again, that's the whole point of TMN in the first place, and it is unfortunate that they don't take it seriously. Veracity_Testing_with_Traceroutes.zip

Retransmission count? The purpose of the exercise is to see if packet loss is evident; and secondarily, to compare your download rates with "known fast" sources. Even if no packet loss is present, these two tests have already demonstrated that your transfer rates are far slower than they should be. We could do this forever...how many sites does it take to convince Comcast that the real-world downloads are slow? Granted they are also blaming your computer, so another key is demonstrating the slowness using multiple computers, each wired to the cable modem to eliminate any other possibilities. If I were a Comcast tech, that's how I would test it. The key is finding someone who will do it properly. As I mentioned before, our Comcast 150/20 service in our office doesn't seem to be performing as it should (for downloads), and I will have an opportunity to test it tomorrow (straight from the modem with multiple computers, bypassing our local networking equipment). If those tests pan out that something isn't right, I will raise the issue with them, even if they can't replace any cabling or whatnot until Spring. And once a real problem has been established, I will stick to it and keep on them....I won't back down from a technical fight.

So definitely slow, but no perceived packet loss. Let's try something further away for both of us, mirror.flhsi.com [208.65.55.17] in Florida. http://mirror.flhsi.com/centos/7/isos/x86_64/CentOS-7.0-1406-x86_64-NetInstall.iso (same file as the other mirror site). host 208.65.55.17 Tracing route to mirror.flhsi.com [208.65.55.17] over a maximum of 30 hops: 1 <1 ms <1 ms 1 ms firewall.ibsutah.net [192.168.1.1] 2 10 ms 10 ms 9 ms 67.186.246.1 3 9 ms 10 ms 9 ms te-0-6-0-7-sur02.sandy.ut.utah.comcast.net [68.87.220.221] 4 10 ms 10 ms 11 ms te-0-7-0-0-sur01.sandy.ut.utah.comcast.net [69.139.247.17] 5 13 ms 14 ms 14 ms te-0-9-0-0-ar04.saltlakecity.ut.utah.comcast.net [68.87.170.1] 6 26 ms 27 ms 30 ms he-0-2-0-0-cr01.denver.co.ibone.comcast.net [68.86.90.233] 7 23 ms 38 ms 22 ms xe-0-0-2-0-pe01.910fifteenth.co.ibone.comcast.net [68.86.84.26] 8 41 ms 42 ms 41 ms te0-0-0-0.rcr11.b006467-1.den01.atlas.cogentco.com [154.54.10.33] 9 41 ms 41 ms 42 ms te0-6-0-3.ccr21.den01.atlas.cogentco.com [154.54.45.181] 10 43 ms 42 ms 46 ms be2128.ccr21.mci01.atlas.cogentco.com [154.54.25.174] 11 48 ms 47 ms 48 ms be2156.ccr41.ord01.atlas.cogentco.com [154.54.6.86] 12 61 ms 60 ms 66 ms be2098.ccr41.atl01.atlas.cogentco.com [154.54.29.98] 13 68 ms 68 ms 68 ms te0-0-0-0.rcr11.jax01.atlas.cogentco.com [154.54.29.130] 14 75 ms 73 ms 81 ms te3-2.ccr01.mco01.atlas.cogentco.com [154.54.29.138] 15 73 ms 74 ms 74 ms 38.104.88.94 16 73 ms 74 ms 74 ms eth-ge0001-car1.orld-st.flhsi.com [208.90.222.250] 17 72 ms 72 ms 72 ms s06-sf0001-car1.xmas-tw.flhsi.com [208.90.222.193] 18 80 ms 74 ms 73 ms s06-sf0003-car1.coco-tw.flhsi.com [208.90.222.210] 19 72 ms 69 ms 72 ms s18-sf0001-car1.coco-vr.flhsi.com [208.90.221.201] 20 73 ms 72 ms 74 ms eth-ge0001-far1.coco-vr.flhsi.com [208.90.221.26] 21 74 ms 73 ms 76 ms mirror.flhsi.com [208.65.55.17] Trace complete. I can still download in under 60 seconds (more that 50Mbps). I had 17 retransmitted packets.

How fast was your download?

Looks like packet loss to me, but let's do a controlled test. Please perform the following steps: 1) Traceroute to mirror.anl.gov [146.137.96.7] which is the Argonne National Laboratory in Illinois, which should be a lot closer to you than me. Here's mine: Tracing route to mirror.anl.gov [146.137.96.7] over a maximum of 30 hops: 1 1 ms 1 ms <1 ms firewall.ibsutah.net [192.168.1.1] 2 13 ms 12 ms 9 ms 67.186.246.1 3 12 ms 13 ms 9 ms te-0-6-0-7-sur02.sandy.ut.utah.comcast.net [68.87.220.221] 4 10 ms 9 ms 12 ms te-0-7-0-0-sur01.sandy.ut.utah.comcast.net [69.139.247.17] 5 25 ms 17 ms 15 ms te-0-9-0-1-ar04.saltlakecity.ut.utah.comcast.net [68.87.170.5] 6 31 ms 28 ms 31 ms he-0-1-0-0-cr01.denver.co.ibone.comcast.net [68.86.90.225] 7 27 ms 30 ms 29 ms be-10717-cr02.denver.co.ibone.comcast.net [68.86.84.226] 8 44 ms 38 ms 47 ms be-11317-cr01.dallas.tx.ibone.comcast.net [68.86.84.229] 9 45 ms 45 ms 42 ms be-14-pe02.1950stemmons.tx.ibone.comcast.net [68.86.83.110] 10 38 ms 41 ms 44 ms xe-0-3-0.0.dlls0.tr-cps.internet2.edu [64.57.20.93] 11 53 ms 53 ms 54 ms xe-4-3-0.206.chic0.tr-cps.internet2.edu [64.57.20.207] 12 53 ms 56 ms 56 ms 64.57.20.1 13 60 ms 60 ms 71 ms 192.5.170.226 14 65 ms 62 ms 54 ms ciscore-v911-221c1hsrp.net.anl.gov [130.202.212.141] 15 56 ms 58 ms 55 ms mirror.anl.gov [146.137.96.7] Trace complete. 2) Launch Wireshark and enter a capture filter for host 146.137.96.7 on your Ethernet interface and start a capture. 3) Download http://mirror.anl.gov/pub/centos/7/isos/x86_64/CentOS-7.0-1406-x86_64-NetInstall.iso I can download this 362MB file in under 60 seconds at over 50Mbps. 4) Stop the capture, and then look at your expert infos. I had two retransmitted packets for the download.

https://testmy.net/ipb/topic/32071-upload-speed-issue-comcast/page-3?hl=%2Bpacket+%2Bcapture#entry345372

I think the key is convincing them to send out a tech who will do more than just do the Xfinity speed test like any monkey can. If they don't believe the TMN tests, fine - there are plenty of other ways to demonstrate download performance is poor. Give them a list of ten distinct download URLs to show that they are all slow...and I still think you should do a packet capture to look for packet loss, as I'm pretty sure you'll have it. You can also try quoting the classic line from the old Heaven Can Wait movie: The likelihood of one individual being right increases in direct proportion to the intensity with which others are trying to prove him wrong.

Which area of the country are you in, and are you still getting results from your first post - paying for 60/10, but only getting ~ 11 down most of the time? At my residence, I'm paying for 50/10 through Comcast, and consistently get it (my downloads are actually closer to 60Mbps most of the time), and TMN consistently verifies I'm getting what I'm paying for. It also verifies EBrown's Comcast 50/10, as well as a colleague of mine who lives about 30 miles North. If TMN is measuring less, I'd believe it, but I also believe you need to validate that with other real-world transfers to show the Comcast people they're FOS. At our office (which is about 7 miles from my house, in another suburb), we just got the Comcast Business 150/20 installed about two weeks ago. On that, the 20Mbit upload is consistent. Downloads, on the other hand, are more like 50Mbps. TMN verifies the "real world" results with other transfer tests, and jives with our NMS monitoring. The Xfinity speed test measures it at 107/20. When we first looked into Comcast Business several months ago, 100/20 was the fastest plan available, and it was only recently that 150/20 was available, although the sales guy said our building would need a new line drop to support it. That never happened, and I believe that to be the reason of the slower speeds. I haven't yet tested from their modem (bypassing our equipment), but will likely take it up with them in the Sping, when weather would allow a new line drop. Even the 50Mbps for us is sooo much faster than what we had before, so we're happy for now, but still not getting what we should. Last but not least, have you performed a packet capture during a transfer test, to see if there is packet loss happening?

Nice. I tried the BBB approach with our ISP and it got us nowhere.

Nope. But I'm also a transplant; moved to the area from Colorado in '97, which I consider my "real" home

Do you know how far it is from the line tap on your house to the Comcast box on the street? We're fortunate in that their box is right in front of our house (making my wife's flower arrangements less pretty), and they ran a new cable from that to our house many years ago, I guess because the existing one wasn't "up to spec," which they discovered when we used to have cable TV + phone rather than just Internet.

I would totally leverage that...you could take your laptop over to his house and wire in, run a selection of upload tests with captures to different endpoints, and compare those with the results at your place using the same laptop. If the results are dramatically different, I don't see how the Comcast techs could deny there's something up with your last mile. I get that ICMP traffic isn't going to be prioritized, but that prioritization only comes into play when there is congestion, and I wouldn't think there would be congestion for services internal to their network. As we've seen from our office experience, packet loss issues are much more apparent for transfers with endpoints far away.

Ha! We're in Sandy; my wife's parents live in Murray. My oldest son graduated from Bountiful High...still working on getting the other two graduated . Oddly enough, none of them seem to want to be a computer nerd like me.

Glad you got it sorted; my suggestion was going to be EasyBCD.

So this saga essentially came to an end today; our account manager called me and said they are releasing us from the contract, with no penalties. They were unable to get XO to fix the last mile problem, and do not have other alternatives for us. They are sending that to us in writing for the lawyers, and we're in the process of provisioning Comcast Business service. There will surely be more bureaucracy with the latter, as it needs to go through our corporate sourcing, but at least we can see a light at the end of the tunnel now.

While dropping only eight packets looks insignificant on the surface, I think it might be. The pathping uses a small packet size which can't be made larger. BTW when I run a pathping to that same endpoint (since I'm on Comcast too, the traffic never leaves their network), I lose no packets, and I'm a lot further away: pathping.txt So here is something else to try, depending on which OS you're using: Windows: ping -l 1472 -t -f speed-upload-01.sanjose.ca.sanfran.comcast.net Mac: ping -s 1472 -D speed-upload-01.sanjose.ca.sanfran.comcast.net Just an ICMP ping with the packet size set at 1472 bytes (the max you can go with a 1500-byte MTU) with the do-not-fragment bit set. Let the ping run for ~ten minutes, and then ctrl+c for the results. Here are my results from: Los Angeles data center (ISP Zayo Group): Ping statistics for 68.87.76.56: Packets: Sent = 934, Received = 934, Lost = 0 (0% loss), Approximate round trip times in milli-seconds: Minimum = 14ms, Maximum = 16ms, Average = 14ms My house (ISP Comcast): Ping statistics for 68.87.76.56: Packets: Sent = 647, Received = 647, Lost = 0 (0% loss), Approximate round trip times in milli-seconds: Minimum = 39ms, Maximum = 58ms, Average = 40ms Our office (ISP Veracity + XO with known last mile problems): Ping statistics for 68.87.76.56: Packets: Sent = 629, Received = 556, Lost = 73 (11% loss), Approximate round trip times in milli-seconds: Minimum = 28ms, Maximum = 31ms, Average = 28ms You can also try this with the closer endpoints in your trace - I'm curious to see the results. There really should be virtually no packet loss with an endpoint that close (and on Comcast's network).

You might also like VisualRoute. It's not free, but does have a 15-day trial which could be used for analysis.

I think I'd ask for a replacement. Are they charging you a monthly fee for the gateway? I'll readily admit I'm old-school, but I much prefer to buy my own cable modem (something on the Cable provider's certified list, such as Motorola SB6121 or newer), and use a different device(s) for the router/firewall/wireless.

Let me try and summarize: ISP is Midcontinent, cable service as some sort of mash-up with Comcast. Your plan is 50Mbps down and 10Mbps up, but the tests here show you're getting 20Mbps down and less than 1Mbps up (max at 94KB/s = 749Kbs). You mentioned the new Hitron CGN gateway, and that the [outbound/upload] streaming problem started with that. Is this correct? What were you using prior to the Hitron, and why was that changed?

Halfway joking, halfway serious: tell them to read the Wikipedia article on packet loss...it's actually quite good.

I've read your post a few times, and am having a hard time following what you mean. Is the twitch.tv OBS thing an upload for you? You're using the little b (for bits) as opposed to the big B (for bytes), so we need to confirm that's really what you mean. In other words, when you say 1000kb/s and use a little b, that is [roughly] equivalent to 1Mb/s (one megabit per second). 100kb/s (kilobits) is very slow. I hope I'm not coming across as condescending - just trying to accurately follow your post.

Yes, agreed, your upload statistics look like the packet loss problem at our office (a local ISP, not Comcast), where there are hundreds/thousands of duplicate ACKs and retransmissions. My Comcast capture from my house is from 50/10 service that I consider "normal," in that I do consistently get those speeds. I share your frustration, albeit mine is not with Comcast, which has provided my residential service for nearly a decade with very few problems. When there is a problem, I think the challenge (with any provider) is convincing them of that problem, and that it is theirs to fix. With all of the potential variations on the customer side (from being wired or wireless, infinite equipment combinations, computer novice to expert, etc.) there's a threshold to cross before someone takes you seriously and will engage. It took us a long time to get to that threshold at our office, when the ISP's VP of Engineering came over with his laptop and did some packet captures of his own. Before he came over and saw the results himself, we got a lot of "sorry, we don't know what to tell you" mumbo-jumbo.

So it looks like you're on the plan we are going to subscribe to at our office, which is 150/20. I wish that was in place so that we could compare notes directly. I just tried what you did from my house, where the plan is 50/10. I'm getting a lot of retransmissions during the upload too, but not near as many duplicate ACKs as you. It is interesting that these are far greater for uploads vs. downloads...I'll have to play with this some more...we haven't been as concerned about uploads at the office, so I haven't spent much time analyzing those. Comcast_speed_test.mov

The problem with our office connection is excessive packet loss in the "last mile," which makes single-threaded transfers max out at around 8Mbps on our 30Mbps symmetrical circuit, unless we are transferring files over low-latency routes (directly from our ISP or from providers whom they peer with), in which case we can hit 27Mbps. It has been a long struggle, but we are in the process of ditching the last mile provider (XO Communications). Our [local] ISP is either going to resell us a Comcast Business 150Mbs connection, or we're going to get that directly. It's a long story in another thread; this is the short version.