Contributor
•
2 Messages
Bad first hop
I have latency and packet loss problems with my first hop, which is an sbcpacbell server that is owned by AT&T (AT&T bought SBC Pacific Bell). On occasion, these problems go away, but, usually, they are apparent. The trace route (shown below) shows that the first hop is the source of most of the incurred latency. I also have a static IP, which may contribute to my issue, as there is perhaps less flexiblity in my initial routing from the Central Office. I am paying for 6 Mbps down and 1.5 Mbps up; my speeds are currently ranging from 2.5 to 5.5 down and 0.7 to 0.08 up. I have had 3 inside techs in the last 2 weeks and 1 outside tech, none of which have located any problems (although I did have a useless filter removed and some wiring cleaned up). One tech replaced my NGV-510 router with a NGV-5031, which didn't help the internet connection at all, and actually made the inside wireless connections more problemmatic; I have since reverted to the 510. I have attached a pingplotter output as a nominal example that shows latency and lost packets in the bottom graph, with a trace route in the top half.
It appears to be impossible to get any U-verse assistance other than another useless inside or outside tech visit (anothe schedule for tomorrow). I figure after AT&T spends the money for 20 or more visits, somebody at AT&T might notice. I am really beginning to regret switching to U-verse from my regular AT&T phone line with ADSL.
JefferMC
ACE - Expert
•
36K Messages
9 years ago
At first I thought you were talking about the traffic loss, and I see you're not. For anyone reading this and seeing the packet loss and going "hey": ICMP is not considered "real" traffic. Routers dump ICMP all the time, especially ICMP errors. I see that quite a few hosts are responding to your ICMP packets beyond your first hop, which means that your first hop is transmitting ICMP packets that still have time to live. Note that hop 12 dropped 100% of the packets intended for it, but hop 13 returned 99.2 (which means at least 99.2 of the packets intended for hop 13 passed through both hop 1 and 12).
I see that what you're taking issue with is latency. First hop latency like that normally comes from one of two sources:
1) correctable errors
2) congested upstream
What traffic do you have going upstream during this period? Are you doing any QoS management to restrict upflowing traffic? If you fill your upstream pipe up, you will see this sort of latency.
But, while you're at it, why don't you collect your Gateway's Broadband Sync Rates. SNR, attenuation and error counts and post them here.
0
0
lkoepke
Contributor
•
2 Messages
9 years ago
Thanks for the prompt reply.
You're right, packet loss is not my concern here as I have already learned that routers may simply toss these queries. The latency is occurring at intervals; we can see good behavior for hours or days, then the kind of behavior that I am describing for a hours or days. It comes and goes. We don't activiely monitor it other than noticing annoying slow response or unacceptable streaming (Netflix, Roku, YouTube).
The AT&T inside tech today proclaimed everything OK on the inside, but that there is a severe problem on the outside. He showed be a 30 day history chart filled with red blocks and another with huge FEC values. He did not understand how the previous outside tech proclaimed everything to be OK. He scheduled an "emergency" outside tech visit. He mention changing out my DSL pair, but that has been done before, with all spurs eliminated. I have a direct line to the CO (~9,000'). It may be just coincidence, but our connection went from "really bad" to "horrible" at the same time that we got rain (yesterday). I'm not sure that there is another spare pair to connect.
The data that you requested:
Sync Rate (kbps) Down =6013,Up 766
SN Margin (dB) Down = 7.1, Up = 15.0
Line Attenuation Down 44.0, Up 27.2
FEC Errors (redundant term?) Down 17658, Up 0
CRC Errors Down 1, Up 0
This data comes from a period when our Internet connection is behaving. I will post an update when/if our connection goes bad again; should be any time, as it has been good all afternoon and evening (so far).
I'm a software guy, not an internet guy. How best should I monitor our internet traffic to get a handle on the load, both up and down? Can you advise me of some software that I may use? I see a number available that will run on each computer. I have a DD-WRT router behind the AT&T DSL router; maybe there is software that can query that for internet traffic?
We have very little upstream traffic that is consciously initiated outside of the occasional video game. I know nothing of QoS management other than the definition and the fact that our current QoS falls far short of AT&T's promise.
Again, Thanks
Larry
0
0
JefferMC
ACE - Expert
•
36K Messages
9 years ago
Not knowing the collection time frame, it's sort of hard to gauge the significance of that FEC error count, but it looks reasonably low assuming at least 3 hours of collection time. Your SNR rate is a bit low.
Yes, rain water could be finding its way into a splice anywhere between your home and the CO and wreaking havoc with your signal quality.
I'm pretty sure DD-WRT can show you traffic counts up and down on each interface, so you can see how much traffic is going out the WAN interface to the RG and make sure that there's not a bunch of outgoing traffic everytime you're seeing bad latency, though I don't know if DD-WRT will show you anything other that raw counts since last reset, i.e. you may have to do something towards collecting/graphing the data (e.g. MRTG, or something like this: http://www.dd-wrt.com/phpBB2/viewtopic.php?t=2563 ).
You may also be able to have DD-WRT limit upstream traffic to 80% of the 0.75 Mbps upstream you have if upstream turns out to be an issue. This could be considered a QoS tuning.
0
0