DSL woes iii

Every couple of years something goes wonky with our DSL connection and it requires major intervention. A couple of years ago, I had two Frontier technicians in my office talking to a third who was at the local switch.

Yesterday, I had THREE Frontier technicians in my office. One of them was the first guy who came out over two weeks ago.

Restricting them to just the iMac connected via Ethernet to the modem sped things up a bit — no futzing with the Airport wifi settings.

After all three tested and tried their ideas, they wound up doing what I thought they’d have to do: switch out the modem for a newer model. In this case, an Arris NVG443B modem, which is a brand I’ve never heard of before.

So far, so good. There were some teething problems early on as reported by Network Logger Pro, mainly Domain Name Server (DNS) outages. Network Logger has not reported any total outages since before the technicians came.

After an evening of iPad-surfing, Liz reported pages taking a long time to load, but they eventually did load and there were no outages. So the connection seems stable, though slower.

I ran the macOS-based Speedtest app and was discouraged to see our 3 mbps provisioned download speed drop to 1 mbps or less on several tests. Even though we experienced outages before, we got close to 3 mbps when the connection was up.

Update: I got 2.8 or 2.9 mbps late last night, but am barely scraping 1.5 this morning.

According to the modem’s built-in xDSL stats page, it’s seeing a downstream rate of 3360 kbps (so, about 3 mbps) and an upstream rate of 863 kbps. 

Hm. The modem seems to be receiving data at the provisioned rate, but on my Ethernet connection, I’m seeing 1 mbps or less. 

Hm. Network Logger Pro is showing now only DNS outages since the Frontier guys left. Fewer than before they arrived, but still...should I be seeing DNS outages at all?

On the old modem, for the last few years, I’d replace the default ISP DNS addresses with the Google DNS addresses or OpenDNS addresses. Supposedly, those server addresses resolve the domain names faster than the ISP’s DNS servers do. (There’s also a new entrant to the DNS space called Cloudflare.)

After some Googling around, here’s what I am doing, using Kim Komando’s DNS article as my guide. 

  1. Removed the Google DNS numbers from Airport Utility and the macOS Network System Preferences pane. They automatically reverted to the ISP’s DNS servers. It seems you can set the new DNS addresses at the modem, router, or device level. I'm keeping the modem's DNS settings untouched for now; I don't want to give Frontier ammunition for saying I mucked up their modem settings.
  2. Flushing the DNS cache (for both computers and browsers) was a new idea for me. Flushing the cache on High Sierra required a special syntax. This required me to log in as the administrator on my Mac to run the Terminal command, since having a separate administrator account is considered good practice.
  3. I downloaded Kim's recommendation of namebench, whose purpose is to discover the fastest DNS servers for your geographic area. I ran it but I would not recommend using it. It was last updated in 2010 and there are few recent references to anyone using it. Namebench recommended Ultradns-2 as the fastest DNS address server, but I will likely go with Cloudflare as primary and Google as secondary, since they're more recent and up to date.
  4. I added 1.1.1.1 (Cloudflare) and 8.8.8.8 (Google) as DNS addresses to the Airport. So our wifi devices should be OK. 
  5. For my iMac, I'm connected to the modem via Ethernet, so I'm using Frontier's default DNS resolvers. 
  6. Liz and I will do our normal surfing to see if speed and stability are better. 
  7. If the speeds are still slow, place another call to Tech Support.

I’m not a tech guru by any means, and network stuff brings me to my knees. What I’ve been documenting in this series is what I know to do, given my paltry knowledge and Googling around. How do normal people deal with this stuff?

Update: Komando wrote about Netflix's Fast.com site. I'm checking its results against Speedtest to see if they agree.