SpaethCoDigital Plumber MVM join:2001-04-21 Minneapolis, MN
1 recommendation |
to wbynum
Re: Is Comcast throttling Usenet downloads suddenly? What gives?said by wbynum:Anyways, short of someone from Comcast actually analyzing my connection/packet flow, not much point going around in a circle on this topic. You've already received a response in this thread from jlivingood who works for Comcast NE&TO. So you have a statement that Comcast doesn't, by policy, shape any traffic for Usenet. You also have a problem that only a handful of folks have reported (and have later resolved to a problem on their side), even though Usenet access is widespread. I don't think anyone is arguing that you're not seeing an issue, the only point is that it's highly likely it is something very specific to your connection path / hardware and not the result of a larger Comcast policy. |
|
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
to wbynum
said by wbynum:Anyways, short of someone from Comcast actually analyzing my connection/packet flow, not much point going around in a circle on this topic. You should start your own thread and include: - traceroute from you to the news service - packet capture of a transfer - testing of an alternative news service (you can probably find one with a trial?) Without that, you're just ranting here and not really contributing to the OP, so if you want to rant by all means do so in your own thread. The OP may be interested in resolving the issue, not just accusing Comcast (without any good evidence) of something they point blank said they don't do. |
|
|
wbynum
Member
2013-Jun-13 5:01 pm
"- testing of an alternative news service (you can probably find one with a trial?)"
I give up. Some of you need to read more closely. How many times have I said that this issue happens with multiple Usenet providers, even with different backends (i.e. not all Highwinds resellers for example)? At least read my posts instead of just repeating the same cut and paste thing about it being a route problem, Usenet provider throttling, hardware problem on my end, etc. |
|
wbynum |
wbynum
Member
2013-Jun-13 5:07 pm
BTW, here is the trace route to my current Usenet provider from the my Comcast connection:
C:\>tracert news.sunnyusenet.com
Tracing route to news.sunnyusenet.com [195.200.85.41] over a maximum of 30 hops:
1 1 ms 1 ms 1 ms unknown [192.168.11.1] 2 7 ms 7 ms 9 ms 73.208.0.1 3 8 ms 7 ms 7 ms xe-9-0-0-32767-sur01.alief.tx.houston.comcast.net [68.85.255.213] 4 30 ms 9 ms 8 ms po-13-ur01.alief.tx.houston.comcast.net [68.85.245.33] 5 16 ms 15 ms 15 ms pos-3-9-0-0-cr01.dallas.tx.ibone.comcast.net [68.86.91.249] 6 16 ms 16 ms 14 ms pos-0-0-0-0-pe01.1950stemmons.tx.ibone.comcast.net [68.86.86.90] 7 65 ms 54 ms 51 ms 173-167-56-166-static.hfc.comcastbusiness.net [173.167.56.166] 8 50 ms 49 ms 47 ms be2031.ccr21.dfw01.atlas.cogentco.com [154.54.7.45] 9 58 ms 62 ms 64 ms te0-2-0-1.mpd21.mci01.atlas.cogentco.com [154.54.5.218] 10 82 ms 85 ms 83 ms te0-3-0-7.mpd21.ord01.atlas.cogentco.com [154.54.84.82] 11 102 ms 97 ms 104 ms te0-4-0-6.ccr21.bos01.atlas.cogentco.com [154.54.45.246] 12 175 ms 172 ms 170 ms te0-3-0-5.ccr22.lpl01.atlas.cogentco.com [154.54.85.222] 13 170 ms 174 ms 177 ms te0-3-0-6.mpd22.ams03.atlas.cogentco.com [154.54.58.41] 14 184 ms 174 ms 177 ms te0-7-0-10.mag21.ams03.atlas.cogentco.com [154.54.76.209] 15 141 ms 136 ms 139 ms 10ge-2-3.cr1.ams3.baseip.com [149.6.128.198] 16 190 ms 183 ms 188 ms 10ge-3-1.cr1.fra4.baseip.com [91.148.255.66] 17 197 ms 191 ms 191 ms 195.200.85.41
Trace complete.
C:\>
|
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
to wbynum
As I understand it, there are only a few main providers of the content but MANY resellers which would all have different routes/connections/etc. So unless you have tried ALL of them here you can't conclude that. |
|
|
wbynum
Member
2013-Jun-13 5:30 pm
Never said it happened with all Usenet providers. Just the 3 or so that I have used since subscribing to Comcast.
BTW, you are correct. There are only a few main providers of Usenet now a days. Vast majority are simply resellers. Some resellers provide their own servers while a lot of others simply resolve to the back end provider's servers (Highwinds, etc). |
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
to wbynum
Looks like a server in Europe? Are all the ones you're trying in Europe? Have you tried something on the east and west coast? |
|
SpaethCoDigital Plumber MVM join:2001-04-21 Minneapolis, MN 1 edit |
to wbynum
Comcast/Cogent congestion |
said by wbynum:BTW, here is the trace route to my current Usenet provider from the my Comcast connection Comcast / Cogent peering has been saturated for weeks in most regions. Performance will be abysmal on this path during peak hours due to the packet loss. That's not usenet specific -- that's everyone that has to transverse Cogent getting screwed. Edit: Example latency / loss graph attached for traffic that has to cross Cogent in Chicago. |
|
|
to pflog
Yep. Looking back over my emails since I got Comcast, I have been with Supernews (one of the biggest tier 1 US based Usenet providers), Frugal Usenet (cheap US based Highwinds reseller I think) and a couple Euro (Netherlands) providers.
Note that the VPS I tunneled through is in an Atlanta data center. |
|
tshirt Premium Member join:2004-07-11 Snohomish, WA |
tshirt
Premium Member
2013-Jun-13 5:46 pm
even the Cogent looking glass has a hard time with that IP traceroute to 195.200.85.41 (195.200.85.41), 30 hops max, 60 byte packets 1 vl3.mag01.iah01.atlas.cogentco.com (66.250.250.33) 0.336 ms 0.341 ms 2 te0-2-1-0.mpd21.iah01.atlas.cogentco.com (154.54.3.238) 0.598 ms 0.604 ms 3 te0-1-0-6.mpd21.atl01.atlas.cogentco.com (154.54.2.145) 14.866 ms te0-2-0-2.mpd21.atl01.atlas.cogentco.com (154.54.42.218) 14.968 ms 4 te0-3-0-2.mpd21.dca01.atlas.cogentco.com (154.54.28.17) 25.833 ms te0-1-0-2.mpd21.dca01.atlas.cogentco.com (154.54.28.226) 25.884 ms 5 te0-2-0-2.mpd21.jfk02.atlas.cogentco.com (66.28.4.17) 32.503 ms te0-0-0-2.mpd21.jfk02.atlas.cogentco.com (154.54.41.18) 32.436 ms 6 te0-3-0-2.ccr21.bos01.atlas.cogentco.com (154.54.44.14) 38.233 ms 38.235 ms 7 te0-1-0-1.ccr21.lpl01.atlas.cogentco.com (154.54.31.182) 103.798 ms te0-5-0-1.ccr21.lpl01.atlas.cogentco.com (154.54.44.194) 103.676 ms 8 te0-6-0-2.ccr21.ams03.atlas.cogentco.com (130.117.48.42) 115.914 ms te0-2-0-6.ccr21.ams03.atlas.cogentco.com (154.54.37.89) 116.150 ms 9 te0-7-0-0.mag21.ams03.atlas.cogentco.com (154.54.76.170) 116.447 ms te0-7-0-4.mag21.ams03.atlas.cogentco.com (154.54.76.138) 116.349 ms 10 10ge-2-3.cr1.ams3.baseip.com (149.6.128.198) 116.429 ms 116.343 ms 11 10ge-3-1.cr1.fra4.baseip.com (91.148.255.66) 120.261 ms 120.346 ms 12 * * 13 * * 14 * * 15 * * 16 * * 17 * * 18 * * 19 * * 20 * * 21 * * 22 * * 23 * * 24 * * 25 * * 26 * * 27 * * 28 * * 29 * * 30 * * in case you want to try it yourself » www.cogentco.com/en/netw ··· ng-glass |
|
|
wbynum
Member
2013-Jun-13 5:53 pm
Here is the trace route from my VPS provider in Atlanta to my Usenet provider. I'm not sure why it does not complete but Usenet transfers do work.
traceroute to news.sunnyusenet.com (195.200.85.42), 30 hops max, 60 byte packets 1 atl-vps1.chicagovps.net (198.23.229.130) 0.023 ms 0.013 ms 0.012 ms 2 host.colocrossing.com (69.31.134.225) 0.427 ms 0.680 ms 0.902 ms 3 74.119.147.145 (74.119.147.145) 0.291 ms ae6-107.cr2.atl1.us.nlayer.net (69.31.135.93) 0.168 ms 0.153 ms 4 vlan-74.ar1.atl1.us.nlayer.net (69.31.135.166) 5.711 ms xe1-4.edge-ab.atl01.coloat.com (184.170.251.253) 0.248 ms vlan-74.ar1.atl1.us.nlayer.net (69.31.135.166) 3.643 ms 5 78.152.46.193 (78.152.46.193) 0.154 ms abovenet.tieatl.telxgroup.net (198.32.132.79) 0.285 ms 0.246 ms 6 eth3-2.r1.chi1.us.atrato.net (78.152.34.197) 21.106 ms 19.104 ms xe-7-3-0.cr1.dca2.us.above.net (64.125.21.222) 14.240 ms 7 xe-2-2-0.cr1.lga5.us.above.net (64.125.26.98) 32.658 ms eth1-4.r1.nyc1.us.atrato.net (78.152.34.149) 44.417 ms 42.405 ms 8 xe-4-3-0.cr2.lga5.us.above.net (64.125.22.170) 17.406 ms 17.384 ms 17.364 ms 9 eth7-8.core1.ams2.nl.atrato.net (78.152.44.156) 109.386 ms so-0-1-0.mpr2.ams5.nl.above.net (64.125.31.181) 97.254 ms eth7-8.core1.ams2.nl.atrato.net (78.152.44.156) 110.761 ms 10 xe-5-2-0.er1.ams5.nl.above.net (64.125.22.62) 99.992 ms 100.654 ms eth4-1.r1.ams2.nl.atrato.net (78.152.44.76) 117.997 ms 11 30ge-1-2.cr1.ams2.baseip.com (195.69.145.174) 110.161 ms xe-2-2-0.er1.ams1.nl.above.net (64.125.28.102) 97.019 ms 30ge-1-2.cr1.ams2.baseip.com (195.69.145.174) 115.944 ms 12 94.31.42.58 (94.31.42.58) 103.570 ms 10ge-3-1.cr1.fra4.baseip.com (91.148.255.66) 118.054 ms 94.31.42.58 (94.31.42.58) 105.255 ms 13 10ge-3-1.cr1.fra4.baseip.com (91.148.255.66) 109.235 ms * * 14 * * * 15 * * * 16 * * * 17 * * * 18 * * * 19 * * * 20 * * * 21 * * * 22 * * * 23 * * * 24 * * * 25 * * * 26 * * * 27 * * * 28 * * * 29 * * * 30 * * * |
|
wbynum |
wbynum
Member
2013-Jun-13 5:55 pm
Can someone recommend a US based Usenet provider that should not have route issues from Comcast Houston? If so I'd be glad to try out that provider and report back trace route, download speed, etc.
If possible though, please make it a cheaper one. I don't want to spend something like $20+ on Giganews for a month. Of course if the provider offers a free trial that would be fine. |
|
graysonf MVM join:1999-07-16 Fort Lauderdale, FL |
I can't speak to network conditions in Houston.
Try www.altopia.com
Never had any problems with them, and I have been a customer since 1995.
Last I checked you could sign up and if you were not happy you could ask for a refund during the first week or so. |
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
to wbynum
I asked my friend at work and he uses newsguy and they have a free 20GB (14 day) trial: » acc.newsguy.com/cgi-bin/ ··· ial_form |
|
whfsdude Premium Member join:2003-04-05 Washington, DC |
to wbynum
said by wbynum:Can someone recommend a US based Usenet provider that should not have route issues from Comcast Houston? If so I'd be glad to try out that provider and report back trace route, download speed, etc. I can hit over 250mbit/s on Astraweb (20 connections) from Comcast when using a machine with a SSD. I'd suggest you look at your router and make sure it can handle the firewall throughput. Also make sure you're not maxing out I/O on the box. |
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
pflog
MVM
2013-Jun-13 6:08 pm
said by whfsdude:I'd suggest you look at your router and make sure it can handle the firewall throughput. Also make sure you're not maxing out I/O on the box. firewall is a slight possibility, but they said it is fast over a VPN so I would think the local I/O isn't an issue and a firewall issue is unlikely. |
|
|
to whfsdude
Shouldn't be a router issue since the same transfer tunneling through my VPS maxes out the Comcast 50 mb connection. Same router, windows box running Sabnzbd, same number of concurrent Usenet connections, etc. |
|
|
to wbynum
said by wbynum:I appreciate the offers for help. To be honest though I'm not sure what would be proven by testing the usenet connection from a different part of the country. Just as it could be a route problem from my end (unlikely IMO), Comcast could have throttling on in some markets but not others. For what it's worth I've never seen any signs of throttling. This coming from someone in the PDX market, connecting to Newshosting (Highwinds). |
|
|
Yaright to SpaethCo
Anon
2013-Jun-13 8:04 pm
to SpaethCo
said by SpaethCo:said by wbynum:BTW, here is the trace route to my current Usenet provider from the my Comcast connection Comcast / Cogent peering has been saturated for weeks in most regions. Performance will be abysmal on this path during peak hours due to the packet loss. That's not usenet specific -- that's everyone that has to transverse Cogent getting screwed. Edit: Example latency / loss graph attached for traffic that has to cross Cogent in Chicago. I heard that Netflix shifted all their traffic to Cogent which congested all these paths so Comcast would peer with them. They seem to be doing this to many ISPs as part of their master plan. Explains their ranking system as well |
|
whfsdude Premium Member join:2003-04-05 Washington, DC |
whfsdude
Premium Member
2013-Jun-13 8:43 pm
said by Yaright :I heard that Netflix shifted all their traffic to Cogent which congested all these paths so Comcast would peer with them. They seem to be doing this to many ISPs as part of their master plan. Explains their ranking system as well Nope. If you hit their OpenConnect streaming servers (IPv6 even without being an OpenConnect partner), there is a chance you'll go over Cogent but that has more to do w/BGP than anything else. » bgp.he.net/AS2906v4, they push over Level3 or Limelight on Comcast. What's more annoying is they seem to favor v4 on Comcast than v6. Which is really kind of infuriating to push v4 traffic when something's dual-stack. |
|
|
yaright
Anon
2013-Jun-13 9:18 pm
When I said "all their traffic" I meant the OpenConnect served traffic. After Cogent is 100% the rest must need to spill to other CDNs. Maybe they should rename this DoS CDN to "CloseConnections" |
|
|
to Papageno
Signed up for a free trial at Tweaknews. Yes, still a euro provider but the route is different from that of Sunnyusenet. Haven't been able to find a US provider that I don't have to give out CC details.
Direct connect to Tweaknews vs tunneling through my VPS to Tweaknews shows the same behavior:
Direct connect: ~200 KB/s Through VPS: ~1.9 MB/s
Here are the trace routes:
Directly to Tweaknes:
C:\>tracert news.tweaknews.eu
Tracing route to news.tweak-news.eu [82.197.196.44] over a maximum of 30 hops:
1 1 ms 1 ms 1 ms unknown [192.168.11.1] 2 8 ms 6 ms 7 ms 73.208.0.1 3 9 ms 9 ms 11 ms xe-9-0-0-32767-sur01.alief.tx.houston.comcast.net [68.85.255.213] 4 9 ms 12 ms 10 ms po-13-ur01.alief.tx.houston.comcast.net [68.85.245.33] 5 20 ms 14 ms 18 ms pos-3-8-0-0-cr01.dallas.tx.ibone.comcast.net [68.86.94.121] 6 17 ms 14 ms 18 ms pos-0-4-0-0-pe01.1950stemmons.tx.ibone.comcast.net [68.86.87.218] 7 18 ms 14 ms 15 ms 173.167.57.114 8 138 ms 137 ms 143 ms xe-5-3-0.ams12.ip4.tinet.net [89.149.185.93] 9 142 ms 140 ms 145 ms a2b-internet-gw.ip4.tinet.net [141.136.97.10] 10 222 ms * 212 ms 217.19.17.66 11 151 ms 154 ms 151 ms 82.197.196.44
Through VPS:
$ traceroute news.tweaknews.eu traceroute to news.tweaknews.eu (82.197.196.44), 30 hops max, 60 byte packets 1 atl-vps1.chicagovps.net (198.23.229.130) 0.045 ms 0.011 ms 0.008 ms 2 host.colocrossing.com (69.31.134.225) 0.414 ms 0.654 ms 0.886 ms 3 ae6-107.cr2.atl1.us.nlayer.net (69.31.135.93) 0.165 ms 0.154 ms 74.119.147.145 (74.119.147.145) 0.248 ms 4 vlan-74.ar1.atl1.us.nlayer.net (69.31.135.166) 5.039 ms xe1-4.edge-ab.atl01.coloat.com (184.170.251.253) 0.273 ms 0.350 ms 5 78.152.46.193 (78.152.46.193) 17.226 ms ae3-113.atl11.ip4.tinet.net (173.241.128.245) 0.208 ms 78.152.46.193 (78.152.46.193) 17.260 ms 6 xe-5-3-0.ams12.ip4.tinet.net (89.149.185.93) 106.837 ms eth3-2.r1.chi1.us.atrato.net (78.152.34.197) 19.054 ms 19.165 ms 7 eth1-4.r1.nyc1.us.atrato.net (78.152.34.149) 42.426 ms a2b-internet-gw.ip4.tinet.net (141.136.97.10) 113.020 ms 112.756 ms 8 eth1-5.core1.lon1.uk.atrato.net (78.152.44.134) 106.301 ms 217.19.17.66 (217.19.17.66) 112.790 ms 123.687 ms 9 82.197.196.44 (82.197.196.44) 112.220 ms eth7-8.core1.ams2.nl.atrato.net (78.152.44.156) 113.991 ms 82.197.196.44 (82.197.196.44) 101.308 ms
Thoughts? Still bad routing? |
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
pflog
MVM
2013-Jun-13 10:37 pm
said by wbynum:Thoughts? Still bad routing? Use a server that isn't in Europe. Seriously. All you're proving is the connection between you on Comcast and EU is bad. Find a server in the US and post results. newshosting.com has a trial, too, I was informed. And at least for me the trace takes me to a DC/VA datacenter. That's like saying "traffic is slow to place xyz" when in reality, there's a traffic jam between exits 50 and 51 and all routes you've tried take you on that highway. Take the back roads my friend and see how it goes |
|
SpaethCoDigital Plumber MVM join:2001-04-21 Minneapolis, MN |
to wbynum
said by wbynum:Thoughts? Still bad routing? Does your NIC support Large segment offload or checksum offloading? If so, try disabling them. There is a known issue with many NICs (particularly common with Intel and Broadcom) that creates issues with large TCP window sizes associated with high latency transfers, such as the one you are initiating to Europe. Once you encapsulate that in a VPN tunnel, it would no longer able to be offloaded to the NIC, and hence you'd see the better performance if this interaction was the cause. The cliffs notes version on the background is that your OS has way more memory available to it for packet buffering than the NIC, so when you try to offload it to the NIC you actually force yourself into smaller window sizes that end up throttling your maximum TCP throughput. |
|
|
to pflog
I have not experienced any throttling on my connection to powerusenet (us hosted). Here is a trace-route. Goes completely within the Comcast network (at least according to the IP names). Here is a link to the trial, it does require billing info though (» www.powerusenet.com/signup/). I have used them since 2006 on 3 different ISP's and have never had a problem. I think there is a problem between you and Europe, I just pulled a download about three minutes ago and was pulling 47.6Mbps of a 50/10 connection. Tracing route to news.powerusenet.com [216.196.97.161] over a maximum of 30 hops: 1 1 ms 1 ms 1 ms [192.168.1.1] 2 7 ms 7 ms 7 ms 96.120.0.133 3 7 ms 7 ms 7 ms te-9-3-ur01.southvalley.nm.albuq.comcast.net [69.139.177.77] 4 11 ms 9 ms 17 ms te-9-2-ar01.albuquerque.nm.albuq.comcast.net [68.86.182.21] 5 17 ms 23 ms 24 ms te-0-0-0-13-cr01.denverqwest.co.ibone.comcast.net [68.86.90.185] 6 17 ms 23 ms 22 ms he-3-13-0-0-cr01.denver.co.ibone.comcast.net [68.86.89.73] 7 46 ms 39 ms 39 ms he-5-14-0-0-cr01.350ecermak.il.ibone.comcast.net [68.86.85.117] 8 60 ms 63 ms 59 ms he-4-7-0-0-cr01.newyork.ny.ibone.comcast.net [68.86.88.157] 9 71 ms 71 ms 71 ms he-4-12-0-0-cr01.ashburn.va.ibone.comcast.net [68.86.85.29] 10 65 ms 67 ms 63 ms be-16-pe04.ashburn.va.ibone.comcast.net [68.86.84.226] 11 65 ms 67 ms 64 ms as30094-2.ashburn.va.ibone.comcast.net [66.208.229.62] 12 65 ms 64 ms 63 ms news.powerusenet.com [216.196.97.161] Trace complete. |
|
|
yaright to whfsdude
Anon
2013-Jun-13 11:11 pm
to whfsdude
It's the Netflix client that chooses the CDN and path not BGP. Also the incoming path is selected by neflix. I have this on very good authority that neflix is purposely contesting paths to drive their OpenConnect strategy across many ISPs |
|
pflogBueller? Bueller? MVM join:2001-09-01 El Dorado Hills, CA |
pflog
MVM
2013-Jun-13 11:21 pm
said by yaright :It's the Netflix client that chooses the CDN and path not BGP. Also the incoming path is selected by neflix. I have this on very good authority that neflix is purposely contesting paths to drive their OpenConnect strategy across many ISPs Netflix has nothing to do with usenet so this is kind of out of left field. |
|
|
to Papageno
Tweaknews free trial only allows 4 or 8 connections iirc. I max my 25mb connection with Tweaknews using 20 connections. |
|
|
wbynum
Member
2013-Jun-14 10:58 am
Yep, 8 connections for their trial. Test was like for like though so it should not matter. Eight connections directly to Tweaknews vs eight connections tunneled to Tweaknews. Huge disparity in the download speed.
I'll do a free trial of Supernews this weekend. Will be interesting to see what results I get. With Charter as my ISP I got the expected bandwidth. I move to Houston, hook up exact same computer, router, etc and my Supernews downloads were ~500 KB/sec. |
|