Constantine A. Murenin
Posts tagged ‘DSLReports’
AT&T U-verse on BroadbandMap.Gov.

We’ve all heard that the fcc.gov et al has spent 350 million USD on the National Broadband Map. Has anyone actually bothered to visit it recently?

http://www.broadbandmap.gov/about-provider/at&t-inc./nationwide/ — T
http://www.broadbandmap.gov/about-provider/verizon-communications-inc./nationwide/ — VZ
http://www.broadbandmap.gov/about-provider/surewest-communications/nationwide/ — SureWest in Sacramento, CA
http://www.broadbandmap.gov/about-provider/cincinnati-bell-inc./nationwide/ — Cincinnati Bell in Ohio

I’ve looked at it repeatedly just now, and I have no idea what those numbers are, or how they could be useful to anyone who is interested in broadband. Keep in mind, I’m an engineer, fascinated with math and numbers. I have no idea what a regular person would be doing with any of those numbers. Numbers by each provider simply make no sense. SureWest has all zero-dot-something percentages, so does Cincinnati Bell. I found no option on the site to get the numbers to make any kind of sense. Was BroadbandMap.Gov simply designed to be the map of AT&T and Verizon coverage?

Yet, apparently, according to the map, even AT&T doesn’t offer FTTP to anyone! If you’ve heard people discussing their FTTP, they must simply be confused by the marketing, AT&T doesn’t actually offer, as BroadbandMap.gov unambiguously puts it, “Optical Carrier - Fiber to the End User”. Also, Sonic.net and Paxio.net are myths, they don’t exist, and don’t offer anything to anyone, let alone any Gigabit speeds for mere pennies on the Mbps! So is Webpass.net, they don’t offer 200/200 speeds for 45$/month in San Francisco Bay Area, either.

Also, apparently, AT&T doesn’t even use VDSL2. Note that VDSL2 is not “Asymmetric xDSL” technology; VDSL2 is symmetric and capable of 100/100 speeds at 0.5km loop lengths. Surprise, surprise!

Another note is that AT&T does offer 6Mbps upload speeds… Hmm… Note that there’s a single page for wired and wireless divisions… Yeap, you’ve guessed it — 6Mbps uploads is the artefact of the wireless networks. :-) No 6Mbps upload luck for any U-verse users!

Written for, and discussion at, http://www.dslreports.com/forum/r26762551-AT-T-U-verse-on-BroadbandMap.Gov.


Any consumer routers that can do routing? Or, the router that isn’t.

I’m looking for a robust consumer router that can do the simplest routing function of all — simply forward packets between the WAN and LAN interfaces. The option of NAT’ing the RFC 1918 addresses would be awesome, too.

Spoiler alert: after several very long conversations with ZyXEL NA tech support (including the managers; by the way, ZyXEL’s tech support is outsourced to Anaheim, CA), I was told that noone makes such devices for the consumer market at all. Is that really true?

I have a 99.124.xxx.xxx/27 Static IP address allocation from AT&T U-verse FTTP; however, the way it worked with 2Wire is that you still get a single regular “dynamic” IP address via DHCP from their common and shared 76.220.xx.xx/22 pool, through which all your traffic to your static IP addresses (in a totally different subnet, as you may have noticed) is then routed. The 2Wire 3800HGV-B then has a setting called “Public Networks” → “User Defined Supplemental Networks”, where the user has to manually specify the allocation they have received; subsequently, for each individual device on the LAN (as well as in the default options for the LAN DHCP server itself) you can either assign a public address from the public pool, or a private address from the private pool (with the option of specifying which public address the private address will be NAT’ed to). However, I’m getting rid of 2Wire PoS due to the unlimited number of bugs, stability issues, as well as unacceptable power consumption (2× to 3× higher than the devices below, without even supporting GigE or 802.11n).

Prior to buying the routers as below, I’ve tried connecting my OpenBSD netbook to the Ethernet port on the SBC ONT directly, to see if I can indeed ditch 2Wire 3800HGV-B PoS, and after some playing with `ifconfig` and `route`, indeed was I getting all the packets for the static block from the internet without any problems!

I’ve got a ZyXEL NBG4615 to replace 2Wire, then subsequently NETGEAR WNR3500L to replace ZyXEL. Both were (and still are) marketed as routers. When setting up each, I’ve changed the MAC-address to the one used by 2Wire, and set up my /27 subnet to be used for their LAN interfaces. Apparently, both ZyXEL and NETGEAR happily do NAT of publicly routable IP addresses instead of passing it straight, and neither one can do packet forwarding (also known as “routing”, surprise!) between the WAN and LAN interfaces without the NAT.

The ZyXEL does have an option of disabling NAT, so, according to their interface, it’s all supposed to work just dandy. However, apparently, in practice it doesn’t do any routing between the two interfaces once the NAT is disabled (I presume they erroneously also do something like `sysctl net.inet.ip.forwarding=0` or `sysctl net.ipv4.ip_forward=0` when you disable NAT), so my internet simply stops working immediately and as soon as I disable NAT within their interface. I’ve contacted the ZyXEL tech support, and they seem to misunderstand what routing is all about, they also claim that no consumer-oriented router can do routing without [also] doing NAT. Is that really true?

In any case, I tell them they have a clear bug with their user interface not functioning the way anyone would expect it to, yet they repeatedly conclude that they’ll only address the problem if other comparable products on the market also have the feature (“have implemented their own feature set correctly”, they mean?). Pardon me, but how are the obvious bugs in one’s interface are related to any other products by any other manufacturer? Especially if all that’s concerned is literally a one-byte change (0 to 1, that’s merely a bit even!); strike that, most likely is merely a matter of actually removing one or more lines of code that disables ip forwarding through sysctl when NAT is disabled through the interface. After all, this GigE router is based on Linux 2.6, from what I gather and based on nmap.

The NETGEAR doesn’t have any options to disable NAT in their default firmware. Although, to be fair, I would argue that having a default of doing NAT of non-RFC1918 addresses is a major bug in and of itself, and any NAT-disable options in any interface are only really meant to apply to the RFC1918 addresses in the first place.

So, just out of curiosity, any consumer routers that can actually do the simple routing, please?

Is AT&T’s setup of two different subnets (as explained above) really so uncommon in the ISP world to not get any attention of third-party consumer router manufacturers?

Am I actually doing something wrong, and is this whole thing supposed to be configured some other way? Or is this really too advanced and is not supposed to work with consumer off-the-shelf routers at all?

Any firmwares to recommend for WNR3500L that were actually thought out to be a great fit for packet forwarding and multiple routable IP addresses, over two subnets as above? I just want my subnet to work, nothing too fancy, really. That said, it would be disappointing to actually have fewer features than what was available back with 2Wire, e.g. it would be nice to continue having the ability to have two IP-address pools for my LAN, one public and one private. A SIP registration server, HE’s IPv6 TunnelBroker.net support and authoritative DNS would be a plus, too, though. SNMP won’t hurt, either. (-: Looking for something stable that I could install with uptime of months, and which would not break when I need to make simple changes of adding new LAN devices etc.

P.S. BTW, apparently, the ZyXEL tech support guys in Anaheim quite misunderstand what routing between two interfaces is all about. They claim that I want some kind of “advanced router”, whereas their product only offers NAT routing (what is “NAT routing” anyways? do they mean “routing + NAT”?), disregarding the fact that they explicitly have the option of disabling NAT in their interface, where the router is still advertised to be in the Router mode (they have a separate option to select the Mode between Router Mode, Access Point Mode etc). I assume that their NAT-disable option not only disables NAT, but also sets `sysctl net.ipv4.ip_forward` to 0. ZyXEL tech support suggested all sorts of things, from using the router in bridge mode, and configuring my host computers to be on my /27, yet somehow have me specify the AT&T gateway from the shared /22 (I’m, like, really?).

Written for, and discussion at, http://www.dslreports.com/forum/r26754312-Any-consumer-routers-that-can-do-routing-


dslreports: How does CenturyLink offer 40/20, when AT&T max is 24/3?

Isn’t it surprisingly shocking that AT&T with all the same technologies as its competitors in other regions, delivers significantly lower speeds than the said competitors, roughly using about the exact same technologies as the competitors are using?

Right now:

BPON @ Verizon FiOS: 15/5, 25/25 and 50/20. (real speeds are actually higher than advertised)
BPON @ AT&T U-verse: …, 12/1.5 and 18/1.5. (yes, there’s no 24Mbps package on AT&T FTTH)

GPON @ Verizon FiOS: 150/35 in print, 150/75 in reality. :-)
GPON @ AT&T: trials? still 18/1.5?

VDSL2 @ CenturyLink: 40/20.
VDSL2 @ AT&T U-verse: 24/3 for single pair, or 18/1.5 for bonded pair (yes, in that order).

Threads about future plans:

CenturyLink:
http://www.dslreports.com/forum/r26533877-CenturyLink-offers-100mpbs-
http://www.dslreports.com/forum/r26643724-Qwest-Pictures-of-100MBps-and-PRICES-Look-Here
Summary: 100Mbps at rather affordable prices (competitive with FiOS), coming soon, official prices and materials leaked from multiple sources.

AT&T U-verse:
http://www.dslreports.com/forum/r26433752-45-MB-internet-service
Summary: 30/3 and 36/6 for FTTH customers only (http://www.dslreports.com/forum/r26644272-45-MB-internet-service). Still merely an unconfirmed rumour, replaced some other prior unconfirmed rumour about entirely different plans, no solid proof whatsoever for neither the old nor the new info.

Will AT&T ever catch up to its ILEC competitors? Why AT&T’s future plans are lower than the existing plans from its competition? Not even talking about Verizon FiOS 150/75, how could CenturyLink be delivering 40/20 over copper, whereas here in this forum it is claimed that AT&T would have problems with the 3.0 upstream on those bonded VDSL2 lines that are seemingly easily capable of much higher speeds than that?

For the record, although fibre is obviously the future, I think copper still has a lot of life in it left. Just one more proof that it’s not what you have, it’s how you use it, AT&T.

Written for, and discussion at, http://www.dslreports.com/forum/r26671109-How-does-CenturyLink-offer-40-20-when-AT-T-max-is-24-3-


dslreports: [FTTH] 4ms+ ping to anywhere outside of the AT&T border

Come to think about it, I’m confused why my ping times are so bad. I have a latency of noticeably well under 2ms to several first hops into the AT&T U-verse network (hops 2, 3*, 4* and 5), yet recently, anything anywhere other than the AT&T routers is always at least 4ms away.

What’s adding those extra 2 to 3 ms of delay between AT&T serving me pretty much in the middle of San Jose, and the other networks within the city limits of, well, you guessed it, still San Jose? If it matters, my ZIP code is 95126 (a 10-minute bike ride to downtown SJ), and, in case anyone missed the prior discussions, my U-verse is powered by Alcatel HONT-C (4 POTS, 1 Ethernet; “155.52 Mbps upstream and 622.08 Mbps downstream”, most certainly 1:32), with the 18Mbps (18/1.5) HSI package (highest available from AT&T over fibre right now, with 24/3 being limited to copper only).

From prior discussion and recollection (http://www.dslreports.com/forum/r26124366-What-has-better-latency-from-a-residential-area-), the best latency I ever got was this summer, at 2.1ms to a USPS site through akamai.net (see a link to my tumblr from the above thread, http://tu.cnst.su/post/8423170265/ping-to-usps-in-under-3ms-from-u-verse-ftth), so I know it should be entirely possible, but no longer happens in practice.

% traceroute -I www.sanjoseca.gov ; date
traceroute to www.sanjoseca.gov (156.39.1.45), 32 hops max, 60 byte packets
 2  76-220-32-3.lightspeed.sntcca.sbcglobal.net (76.220.32.3)  2.985 ms  1.667 ms  1.624 ms
 3  71.145.0.104 (71.145.0.104)  5.641 ms  1.659 ms  1.673 ms
 4  * * *
 5  12.83.39.137 (12.83.39.137)  1.685 ms  1.630 ms  1.588 ms
 6  12.122.200.9 (12.122.200.9)  2.916 ms  2.764 ms  2.830 ms
 7  208.178.58.185 (208.178.58.185)  115.682 ms 192.205.32.50 (192.205.32.50)  199.823 ms dcr2-so-3-0-0.washington.savvis.net (192.205.32.46)  183.584 ms
 8  po1-20G.ar3.SJC2.gblx.net (67.16.134.26)  128.934 ms  394.740 ms  233.524 ms
 9  Hurrican-Electric-LLC.Port-channel100.ar3.SJC2.gblx.net (64.214.174.246)  4.886 ms  4.904 ms  4.882 ms
10  10gigabitethernet1-4.core1.sjc1.he.net (72.52.92.117)  5.184 ms  5.421 ms  5.008 ms
11  city-of-san-jose.gigabitethernet2-8.core1.sjc1.he.net (64.71.176.98)  6.897 ms  6.839 ms  6.595 ms
12  156.39.0.251 (156.39.0.251)  5.316 ms  5.283 ms  4.951 ms
13  * * *
14  156.39.1.45 (156.39.1.45)  5.929 ms  5.690 ms  5.459 ms
Sat 17 Dec 2011 19:26:20 PST

% traceroute -I www.netflix.com ; date
traceroute to www.netflix.com (69.53.236.17), 32 hops max, 60 byte packets
 2  76-220-32-3.lightspeed.sntcca.sbcglobal.net (76.220.32.3)  4.704 ms  1.657 ms  1.627 ms
 3  * * *
 4  * * *
 5  12.83.39.137 (12.83.39.137)  1.612 ms  1.590 ms  1.854 ms
 6  ggr7.sffca.ip.att.net (12.122.114.13)  2.721 ms  2.562 ms  2.617 ms
 7  xe-1-2-0.mpr4.sjc7.us.above.net (64.125.12.117)  4.580 ms  4.315 ms  4.527 ms
 8  xe-3-1-0.er2.sjc2.us.above.net (64.125.27.94)  5.290 ms  5.437 ms  6.319 ms
 9  64.125.28.13.available.above.net (64.125.28.13)  5.183 ms  4.965 ms  5.081 ms
10  64.124.65.90.allocated.above.net (64.124.65.90)  5.305 ms  5.285 ms  5.162 ms
11  xe-2-3-0-945.jnrt-edge02.sv1.netflix.com (208.75.77.197)  6.222 ms  6.232 ms  6.252 ms
12  xe-2-2-0-955.jnrt-edge02.prod1.netflix.com (69.53.225.30)  7.432 ms  7.377 ms  7.253 ms
13  te1-8.csrt-agg02.prod1.netflix.com (69.53.225.10)  12.506 ms  7.671 ms  7.645 ms
14  netflix.co.uk (69.53.236.17)  7.220 ms  7.242 ms  10.942 ms
Sat 17 Dec 2011 19:27:47 PST

% traceroute -I www.facebook.com ; date
traceroute to www.facebook.com (69.171.228.13), 32 hops max, 60 byte packets
 2  76-220-32-3.lightspeed.sntcca.sbcglobal.net (76.220.32.3)  2.506 ms  10.590 ms  1.785 ms
 3  * * *
 4  * * *
 5  12.83.39.137 (12.83.39.137)  2.205 ms  1.656 ms  1.584 ms
 6  151.164.101.206 (151.164.101.206)  3.870 ms  3.810 ms  3.724 ms
 7  xe-0-2-0-5.r07.snjsca04.us.bb.gin.ntt.net (129.250.9.121)  5.644 ms xe-0-2-0-6.r07.snjsca04.us.bb.gin.ntt.net (129.250.8.93)  5.565 ms xe-0-2-0-7.r07.snjsca04.us.bb.gin.ntt.net (129.250.9.141)  5.723 ms
 8  ae-7.r20.snjsca04.us.bb.gin.ntt.net (129.250.5.52)  5.359 ms  5.358 ms  5.213 ms
 9  ae-4.r05.plalca01.us.bb.gin.ntt.net (129.250.5.32)  6.740 ms  6.442 ms  6.229 ms
10  140.174.21.142 (140.174.21.142)  6.838 ms  6.068 ms  6.574 ms
11  ae1.bb01.pao1.tfbnw.net (74.119.76.134)  6.689 ms  6.445 ms  6.464 ms
12  ae9.bb01.prn1.tfbnw.net (204.15.20.51)  32.184 ms  31.861 ms  31.486 ms
13  ae0.dr04.prn1.tfbnw.net (204.15.23.87)  30.840 ms  31.005 ms  30.862 ms
14  po1022.csw05a.prn1.tfbnw.net (74.119.76.243)  31.777 ms  31.721 ms  31.610 ms
15  www-12-05-prn1.facebook.com (69.171.228.13)  29.830 ms  29.854 ms  30.804 ms
Sat 17 Dec 2011 19:29:29 PST
And just for completeness, the test below is back to that IP address to which I was getting 2.1ms in the summer; now merely 4.0ms. The only difference now on my part is that I’ve ditched 2Wire and I’m also not currently using the /27 IP-addresses (but that shouldn’t matter, unless the routing tables are misconfigured on AT&T’s side).
% traceroute -I 69.22.162.122; date
traceroute to 69.22.162.122 (69.22.162.122), 32 hops max, 60 byte packets
 2  76-220-32-3.lightspeed.sntcca.sbcglobal.net (76.220.32.3)  2.984 ms  1.835 ms  2.272 ms
 3  * * *
 4  71.145.0.80 (71.145.0.80)  4.298 ms * *
 5  12.83.39.137 (12.83.39.137)  1.867 ms  1.653 ms  1.550 ms
 6  ppp-151-164-52-233.rcsntx.swbell.net (151.164.52.233)  4.542 ms  3.660 ms  3.726 ms
 7  asn4436-nlayer.pxpaca.sbcglobal.net (151.164.46.70)  4.776 ms  4.858 ms  4.704 ms
 8  69.22.162.122 (69.22.162.122)  4.007 ms  4.041 ms  4.131 ms
Sat 17 Dec 2011 20:02:59 PST

% traceroute -I trkcnfrm1.smi.usps.com; date
traceroute: Warning: trkcnfrm1.smi.usps.com has multiple addresses; using 63.235.28.96
traceroute to a1834.b.akamai.net (63.235.28.96), 32 hops max, 60 byte packets
 2  76-220-32-3.lightspeed.sntcca.sbcglobal.net (76.220.32.3)  2.254 ms  1.912 ms  1.689 ms
 3  * * *
 4  * * *
 5  12.83.39.137 (12.83.39.137)  1.810 ms  1.505 ms  1.665 ms
 6  ggr7.sffca.ip.att.net (12.122.114.17)  2.846 ms  2.796 ms  2.620 ms
 7  192.205.36.2 (192.205.36.2)  4.413 ms  4.382 ms  4.236 ms
 8  * * *
 9  63-235-28-96.dia.static.qwest.net (63.235.28.96)  5.082 ms  6.201 ms  4.846 ms
Sat 17 Dec 2011 20:03:38 PST

Basically, from the original USPS example, it would seem like nlayer.net had the best latency, and turns out, it is indeed true to this day, too:

% traceroute -I nlayer.net ; date
traceroute to nlayer.net (204.93.207.190), 32 hops max, 60 byte packets
 2  76-220-32-3.lightspeed.sntcca.sbcglobal.net (76.220.32.3)  22.426 ms  3.748 ms  2.055 ms
 3  * * 71.145.0.104 (71.145.0.104)  2.677 ms
 4  * * *
 5  12.83.39.137 (12.83.39.137)  1.768 ms  1.620 ms  1.509 ms
 6  ppp-151-164-52-133.rcsntx.swbell.net (151.164.52.133)  3.638 ms  3.564 ms  3.340 ms
 7  asn4436-nlayer.pxpaca.sbcglobal.net (151.164.251.34)  4.897 ms  4.993 ms  14.479 ms
 8  ae0-80g.cr1.pao1.us.nlayer.net (69.22.153.18)  3.806 ms  3.766 ms  4.081 ms
 9  ae1-60g.cr1.sfo1.us.nlayer.net (69.22.143.169)  10.091 ms  9.223 ms  9.687 ms
10  xe-0-0-3.cr1.slc1.us.nlayer.net (69.22.142.96)  24.748 ms  24.892 ms  25.809 ms
11  xe-5-2-0.cr1.ord1.us.nlayer.net (69.22.142.101)  53.535 ms  53.569 ms  53.602 ms
12  po6.ar1.ord1.us.scnet.net (69.31.111.58)  54.156 ms  53.814 ms  53.741 ms
13  61.po1.ar1.ord6.us.scnet.net (75.102.3.226)  54.864 ms  54.891 ms  54.795 ms
14  ge0-1.aggrGA139-1.ord6.us.scnet.net (75.102.0.146)  55.514 ms  56.915 ms  60.966 ms
15  web1.ord.servercentral.net (204.93.207.190)  54.756 ms  54.974 ms  54.642 ms
Sat 17 Dec 2011 20:25:13 PST

Any ideas why the latency is so bad between the AT&T U-verse network and the internet? Why am I routinely getting 4ms+ instead of that 2.1ms that was once sustainably achieved? Basically, it seems like there is an extra 2.5ms to 3ms delay (translating to 4.0ms and 4.5ms overall delay from my ONT) between the AT&T U-verse routers (hop 5 @ 1.5ms overall) and the actual internet (hop 7 or later @ 4ms+ overall). Why? Where are they losing the 2.5ms? Why did those 2.5ms used to be merely 0.5ms, but no more?

Also, where is my BPON coming from? Is it not the AT&T building at 95 S Almaden Ave, San Jose, CA 95113? I always thought that was their CO, and that BPON users are served right from the CO.

Written for discussion at http://www.dslreports.com/forum/r26669632-FTTH-4ms-ping-to-anywhere-outside-of-the-AT-T-border


Why is Cat5e in Bay Area’s large new residential buildings and highrises left unused?

I’m just trying to understand the phenomenon.

Here in the South Bay area, there are a very significant number of medium- and high-rise new apartment and condominium buildings, with 100+ units per building (i.e. per webpass.net’s pricing, they’d offer free building hookup), already having centralised Cat5e or even Cat6, yet there are basically no telecoms who’d steps up and offer any kind of VDSL or Ethernet connectivity at all. Why?

Heck, it seems like the community I’m in right now, is just about the only one that even has AT&T U-verse, thanks the building being pre-glassed with fibreoptics! Residents in most of these other new (2005+) and even newer (2010+) buildings would have to suffice themselves with either Comcast or xDSL from the CO! Seriously? Even AT&T can’t seem to collect their guts and put up a node within the 250 unit buildings, and offer outstanding VDSL2 experience. Why? Where if not within these high-rises (Skyline at Tamien Station, Axis San Jose, The 88 SJ, 360 Residences etc) could the population density be utilised to the fullest extent for the best VDSL results?

Why is noone stepping up? Webpass.net in San Francisco and Paxio.net in Santa Clara are about the only providers I have found so far that even offer this kind of service in principle (both happen to offer residential unlimited symmetric 100/100 under 100$, which is essentially what we’re after here). Both seem reasonably sound, stable and mature, yet there are still thousands of people in the Bay Area living in brand-new buildings with internet connections slower than you can have just about all across the world, including Eastern Europe and South Korea. What’s going on? Why’s noone stepping in? Why not even the most popular CLEC/DLEC in the area, Sonic.net?

In my search, I’m still yet to personally stumble across a building here in the South Bay that’d have Cat5e utilised! I mean, I do have the moral right, if not even an outright obligation, to be surprised here, right? (-:

http://www.dslreports.com/forum/r26345062-Cat5e-in-Bay-Area-large-new-res-buildings-left-unused-