I recently upgraded from Verizon DSL to FiOS but couldn’t find the answers to my questions online (and sales reps weren’t much help). So I tried to document my experience.
I’m not really concerned with the extra bandwidth. If I play online games, I just want to ensure that my latency doesn’t increase. Likewise for web browsing, though some sites are media-heavy and can benefit from the bandwidth. Reliability is another concern. But companies only advertise bandwidth.
The other issue is price. It’s not strictly required to have local phone service to get DSL or FiOS Internet, but the pricing structure really pushes it. My motivation in this process is that the FiOS Internet/Phone package was cheaper per-month than DSL Internet with traditional local/regional phone (no long distance).
Price-wise, DSL and traditional local/regional phone was $68.27 a month (including $8.34 fees/taxes). The original FiOS offer was something like $50/mo for 6 months then $60/month for the rest of the first two years. Under the assumption that fees remain the same (yet to be seen), it’s a short term savings of about $60 and (hopefully) better service.
I was on the fence about the offer, but then they sent me an extra $5 off per month code in the mail, which adds up to a nice long-term saving. They also waived activation fees and installation/router fees. Though I don’t think I’d even consider changing services without that.
So I was looking at going from a 3Mb down, 768Kb up DSL connection to 15 Mb down, 5 Mb up FiOS connection. I ran a few benchmarks on DSL before the switch, which I’ll compare to FiOS after describing the installation process.
The installation isn’t quite what I was expecting. Years ago Verizon installed boxes in everyone’s apartment in my complex. They put it in the bedroom closet in a corner against the ceiling.
I was expecting that a guy would show up during the allotted time (8am-5pm…), would throw some switches in the basement/crawlspace/access points, connect the router to the box, and use an extension cord to get power from the bedroom (no plugs in the closet). I knew there was a battery backup (UPS) involved, but I figured it just had a long cord.
Boy was I wrong! The installation was a 4 hour experience. The tech had to replace the box in my closet. I didn’t watch, but there’s spackle above it now. The output of that box is a fibre line going to a larger box, which appears to be the same box as my office. Then four cables come out of that. Two are routed to the power “system”, one routed for the nearest telephone jack, and one coax line routed for where he wanted to put the router (where my DSL router was).
Each cable was routed and secured to the wall around door frames and the baseboard molding. To get to the power outlet, he drilled straight through the wall. The UPS/power system was attached on the other side to the lower plug of the outlet. He drilled in a different spot for the phone/coax lines, routed around the baseboard and door frames, then drilled a hole to get to the living room underneath the bedroom telephone jack.
When I say drilled, I mean an unadorned hole. Fortunately it’s small enough to spackle when I move out (if I even need to).
When I say the wires were routed around edging, I really want to use the word tacking, but usually I reserve that for small nails. This is instead screws with plastic holders for groups of wires.
After all that and verifying the Internet connection, he had to work on my kitchen phone jack for a while, though I don’t really know what he was doing.
The default router configuration is pretty good for the average user. Router access is the same (192.168.1.1 or dslrouter), but now the default account isn’t admin/admin. The username is admin but the password is some random string or the order number, something like that.
They also setup the wireless config a little better. The default network name is a short random string and they enable WEP with some random-looking password (probably something related to order number). WEP is pretty insecure, but some devices don’t support WPA or WPA2. I switched to WPA2 and changed things around, but the default will probably work well for uninformed home users.
Also, the configuration info is written on the router for you. That’s a nice touch cause it’s easy to forget. They also add a link on your desktop with all the info in it. I was uncomfortable with someone else using my computer but I’ll let it slide.
No much to say: I get 15/5 based on SpeedTest results. Actually, I got a little more than 15 down (15.29-16.03 Mbps) and a little more than 5 up (4.8-5.5 Mbps).
In rare practical situations I’ve seen up to maybe 1.8 MB/s, which is close to the theoretical max. With such a fast connection, it can be difficult to tell whether speeds below the max are due to your ISP or whether it’s the server side of the connection.
Originally I was concerned that latency might degrade, but it improved with FiOS. I ran tests with Google using ping -n 50 http://www.google.com. FiOS reduced the latency from 31 ms to 14 ms (55%). Although the nearest Google IP address may differ under DSL vs FiOS, in web browsing we only care how long it takes to load and render.
Then I used PingTest for the other two latency tests. The ping tests hosted by Ookla (Washington, D.C.) show 49% reduction in latency (35 ms to 18 ms) and the tests against Comcast (Philadelphia) show 18% reduction in latency (37 ms to 30.5 ms).
PingTest also reports jitter, which is analogous to standard deviation. Jitter is usually lower in my tests, but sometimes a test would report enormous jitter, meaning that some packets were very delayed. In using the connection for a while, I can’t say I’ve noticed anything strange. Sometimes I notice long delays with gmail, but that could be any number of things.
Both tests were run on the same computer last Saturday – the DSL tests in the morning and the FiOS tests in the afternoon. If anything, the time of day would probably give an advantage to DSL.
From a subjective perspective, when I started browsing webpages with FiOS I could tell the improvement right away.
The improvement in latency could be due to lower latency in communicating with the local Verizon office or it could be lower latency elsewhere in the chain. If it’s lower latency outside of the local office, that could mean either different uplinks for FiOS customers or something like traffic prioritization for FiOS customers over DSL customers. I’m not a networking specialist by any means, but I ran a traceroute to Google with DSL and again with FiOS.
I traced the same IP addresses to try to match the routes as closely as possible. On the FiOS route, I’ve annotated the route with colors to indicate whether the route was identical (green), close (mustard), or different (red). The close cases have different IP addresses but maybe the same or similar DNS addresses.
2 24 ms 24 ms 24 ms 10.7.62.1
3 25 ms 24 ms 24 ms ge-2-2-1-0.PHIL-CORE-RTR1.verizon-gni.net [220.127.116.11]
4 24 ms 24 ms 24 ms so-7-2-0-0.PHIL-BB-RTR1.verizon-gni.net [18.104.22.168]
5 58 ms 86 ms 30 ms 0.xe-3-1-0.XL3.IAD8.ALTER.NET [22.214.171.124]
6 30 ms 29 ms 30 ms TenGigE0-6-1-0.GW7.IAD8.ALTER.NET [126.96.36.199]
7 56 ms 56 ms 57 ms google-gw.customer.alter.net [188.8.131.52]
8 32 ms 31 ms 45 ms 184.108.40.206
9 37 ms 33 ms 37 ms 220.127.116.11
10 31 ms 32 ms 31 ms iad04s01-in-f103.1e100.net [18.104.22.168]
2 5 ms 4 ms 6 ms L100.PHLAPA-VFTTP-90.verizon-gni.net [22.214.171.124]
3 7 ms 8 ms 6 ms G0-3-3-5.PHLAPA-LCR-21.verizon-gni.net [126.96.36.199]
4 7 ms 7 ms 8 ms so-9-0-0-0.PHIL-BB-RTR1.verizon-gni.net [188.8.131.52]
5 62 ms 67 ms 12 ms 0.xe-3-0-1.XL3.IAD8.ALTER.NET [184.108.40.206]
6 14 ms 13 ms 13 ms TenGigE0-6-2-0.GW7.IAD8.ALTER.NET [220.127.116.11]
7 47 ms 45 ms 46 ms google-gw.customer.alter.net [18.104.22.168]
8 15 ms 16 ms 15 ms 22.214.171.124
9 17 ms 18 ms 16 ms 126.96.36.199
10 14 ms 15 ms 15 ms iad04s01-in-f103.1e100.net [188.8.131.52]
As far as I know, some routers de-prioritize the traceroute responses, so that’s why it looks like it takes a long time to get to some of the servers in the middle. But really it’s only taking a long time when that specific router sends an ICMP timeout message (which is what traceroute relies on). In a successful message it takes much less time to reach those routers.
Of course the first hop (to my router) is quick. Beyond that the next few hops differ. It looks like the second hop with FiOS is to a Philadelphia-area fibre-to-the-premises network (guessing from DNS). From there it looks like both make it to a similar Philly-area Verizon connection (maybe BB stands for backbone?). The rest is close enough that the routes seem comparable.
Based on the timings, it appears that the latency of the second hop is the biggest change in overall latency, which is 24ms for DSL and 4-6ms for FiOS. If I assume that these values are correct and the final values are correct, that means that going from #2 to the end is 7-8ms for DSL and 9-11ms for FiOS.
So why is there such a difference between the router and the second hop? It could be the time it takes to encode and decode the signal going over the telephone line. Or it could be something completely different. All I can do is speculate due to my lack of experience and knowledge.
In my brief testing I’ve found that FiOS is better in both bandwidth and latency. The phone service has some value-added features like emails when you get a message and also long distance. All for a lower monthly fee.
The installation was a pain and I hope my apartment complex doesn’t mind. For all I know, my complex may require reversible installations, leading to this process. Though if I’d known how long it would take and the work involved, I might not have gotten FiOS or I may have deliberated longer.
Blog note: I’d love to analyse IBM’s Watson more, but there are hundreds of articles out there now. Not sure what I’ll write next; maybe an update about title capitalization.