DSL Speed issues

ChedWick

Gawd
Joined
Sep 16, 2011
Messages
596
Evening everyone,

TL;DR Version
Internet speeds are awful intermittently
Basic web browsing appears to cause a spike in ping response time
External ping responses are in the 1000s of ms
Firewall was replaced but the issue persists
Had Verizon "test" the lines; claim no issue with speed
Ran internal ping test while the connection was terrible and all appeared fine
Network monitoring shows no apparent increase in traffic on any computer
All computers have been scanned with nearly every notable av and malware scanner one might recommend
Firewall does show an increase in traffic going out at the time of the issue; unsure about this metric tho
Traffic meters are showing nothing out of the ordinary


Longer Version

I'm quite frustrated. I'm having a network issue that I can't wrap my head around and I've been fighting with Verizon on whether or not its their problem. To be quite honest I have no idea if it really is but I've exhausted all options for internal testing.

Basically I have a site 2.5 hours away with a 3mb DSL connection. They have a netgear vpn firewall, 1 switch, 1 netgear nighthawk set as an AP, 1 network printer, a whopping 6 computers and their primary system is a sales/inventory system they access via an RDP session. So not a whole lot of traffic for mission critical stuff.

Over the past few weeks they've been complaining about horrible slow downs throughout the day. Initially the reported problem was that the internet would just go out. All VPNs would drop and websites would stop loading. I diagnosed the issue to their netgear vpn firewall rebooting at random times. After reading some reports on the same issue with the same model, I got it RMAd. The new one arrived and has been in place a few days but speed is still an issue.

The new version of the netgear firewall gave me a little more reporting than the previous one. Their traffic meters were nothing special. 4 gigs down over 12 hours and 1 gig sent. I do see that when their connection is bombing the inbound and outbound meters on the firewall are in-between 50-90 thousand bps where as they usually sit at around 1-3k bps. I honestly don't know if this is a snapshot of what is going through at the very second it refreshes or what needs to be processes and go through. I've pulled traffic logs and I'm not seeing anything out of the ordinary. I've been in all 6 machines at the same time and watched their network resource monitor go no higher than 10% at any given time.


Remotely when I continuously ping their IP I can see the response times randomly skyrocket to the 1-3k range. Then they'll hover back down around 42ms and occasionally back up. The high response times can last anywhere from 1-20 minutes. If I do a trace route the last hop from me to them I see a huge increase in the response time. If I do the trace route from the firewall out, the first hop is super high then the rest are fine. Something like this happened 2 years ago but the connection was slow ALL DAY. They ended up having a bug; that machine was redone. I've scanned all computers multiple times with multiple AV scanners and I haven't come back with so much as a bad cookie.



So I called Verizon to get some answers. They've been less than helpful. All their Indian call reps do is run a basic line test and say everything is ok while I'm trying to explain that this is an intermittent issue. On a few occasions I got lucky and when their connection was slowing, the Verizon rep claimed to see something too. I'm not sure if this is just because I was on the phone yelling at them for 40 minutes to understand what I was actually saying or if they actually saw something. They had a tech onsite and all they did was plug into the modem and do a speed test. They got their measly 3mb download and said all was clear. I finally recently got someone at a call center who could understand what I was saying so they claimed to have run a copper test. They said they did get a fault and issued a tech. Another tech came onsite and he tested the line, said it was OK but he actually called me to talk about it. This tech was willing to admit that it could still be on their end if it was intermittent but that was about all he gave me.

So I hit more internal testing. I'm now convinced its probably not some super bug hiding away because this evening only 3 of their machines are on and I've been doing some more ping testing. It seems that on any of their machines, as soon as I load of an internet browser and go to a website such as YouTube, their connection will start to struggle. Like all I did was open 1 tab and went to YouTube and I was seeing external pings in the thousands again. It wasn't as constantly in the 1000s but during the initial load of the page there were definitely issues and for the most part average ping responses where in the 400s. If I played a video I saw a constant lag in ping responses. As soon as I close the browser, everything settles and they see normal response times. I did this probably 20 times on the 3 different computers and it was the same result each time. So I started pinging across the lan while doing this. No issues what so ever. Every single ping was <1ms. I pinged the firewall directly and saw the same thing. Everything internall was flowing smoothly but response times to the internet and back were awful.

I'll probably be able to make it onsite Monday but does anyone have any ideas? Any similar experiences or DSL experience? Something I can throw at Verizon that I'm unaware of? Something I can request they test other than plugging into the modem for 5 minutes. Something I can look at internally?
 
Last edited:
Do a traceroute and see where the latency occurs. Is this a modem in bridge mode you're using or are you running double NAT and other fun things?
 
Their modem is in bridge. I have done trace routes during the periods that their connection is awful and its always the first hop when going from the firewall out to Verizons first node or the last hop when I trace route to the firewall from somewhere else.
 
Do note that incoming packets and outgoing doesn't necessarily use the same IP. So what you're saying is that the gateway IP (seen from the DSL connection) that Verizon operates goes wonky each time.

I had a very similar experience recently, packet loss would occur out of the blue continue for a few hours and then promptly stop (like at 4PM-2AM). Interestingly it always disappeared and occured during in periods of full hours. I had to test everything (as usual), disconnect routers, do ping tests for hours and they even replaced the modem. Recently they did some generic "network upgrades" and ever since this issue hasn't been seen. In the end it turned out be a capacity issue which they still today deny......

I'd suggest that you setup something like collectd or smokeping and graph a few test points, Verizon gateway, webfarm2.berkeley.edu (for instance), 66.250.250.105 (cogent) and
4.69.159.186 (level3).
 
Last edited:
Back
Top