Verizon FIOS Scam - Bottlenecking Bandwidth

[unknown66]

I had Verizon FIOS 50/25 service installed recently and I've been having issues with their 'advertised' speeds and the actual speeds I notice during downstream and upstream usage. I am only able to achieve around 2 mbytes/sec maximum downstream from any server in the world. However, if I use multiple downstream threads at once then I can achieve up to around 7 mbytes/sec total on my downstream, basically by using 5 threads or so. And for upstream I can only get around 200 kbytes/sec per thread so I'd need quite a bit threads to reach the 25mbit upstream. The speedtests to the Verizon servers are obviously fine and show speeds as advertised, but this is not accurate when you apply it to actual bandwidth usage. I had my service reprovisioned from COAX to Ethernet so that I could test any difference in performance. I was initially using their supplied MOCA router to connect but now I am connected directly to the modem box unit with my laptop directly. When I made this switch I noticed the bandwidth speeds changed drastically as I was now getting up to around 7 mbytes/sec on any single thread downstream. Upstream however did not change. After a few hours I changed from Work Network to Home Network in Windows network and sharing center and it strangly reset my LAN Card. The reason I say it is strange is because it also reverted the bandwidth speeds back to what they were before, which is 2 mbytes/sec maximum per thread downstream and 200 kbytes/sec per thread upstream.

I've been fighting with Verizon tech support to resolve this and I've explained to them that there must be some sort of throttling and/or bottlenecking going where my bandwidth is being compressed into several threads rather then being utilized as it should in one single thread for each stream. They keep repeating the same rebuttal over and over saying once you leave the Verizon local network 'we aren't responsible anymore for your speeds'. As if all Verizon customers only download content from Verizon servers. I've done extensive speedtesting and traceroutes across the world and my findings are constintent with my bottlenecking theory. This is not just an issue with a few server locations, it's everywhere. Yet, I could not get any tech support to openly admit to such throttiling activities going on though some suggested they were 'unsure' if it's occuring at Verizon. That's close enough I suppose.

I've used other ISPs in the past in other areas with these servers that I download from and speedtest/traceroute and always got appropriate speeds according to the bandwidth allotted to me. And no, for you negative nancies out there, I'm not using any illegal services like torrents, p2p, etc. I feel that even if I were to upgrade to 75mbit for example with FIOS, my speeds would remain CAPPED as they are now or just adjust to a slightly higher throttle like 3 megabytes/sec downstream cap per thread. I'm considering trying out a 30mbit or 50mbit package from Time Warner Cable, as they are the only other provider in my area. If I am able to reach my desired downstream speeds with TWC then it will prove that FIOS in fact does bottleneck their bandwidth usage to customers and disgustingly lie about it.

Does anyone have any ideas how to resolve these throttle/bottleneck issues with FIOS that I am having so that I can achieve the INTENDED advertised speeds. Verizon only goes by their meaningless speedtest which only measures latency on local Verizon servers. They refuse to do anything outside of that and alleviate any and all responsibility elsewhere. People with similar testimonials with FIOS feel free to post reply as well as those who can testify to the quality of TWC's Broadband speeds. Thanks.

 

[bill001g]

You can get the rated speed if you run multiple streams when added together but you cannot get the rated speed on a single stream ?

Then I suspect there is no cap it is related to how TCP works. There are many well written articles that discuss the issues with
" latency and tcp window size ".

The only way you can get a single session to use the whole bandwidth many times is to use something other than TCP....like UDP. The end devices would need this special software to do this.

This problem is one of the reasons bit torrent is so much faster than almost any other protocol.

 

[AntiZig]

other thing to keep in mind that the server you're downloading from might have a limit on connection speed dedicated to single connection

Some hosting service servers have bandwidth of something like 250kbps max, the only other way to get it to go faster is to get a download manager than can establish multiple "streams" (pieces) to be downloaded

 

[unknown66]

i've tested with 10-20 servers from several servers spread across many countries. all of them 100mbit or higher. How can I download using the UDP method you mentioned above? It's really a hassle to have to use 4-5 threads rather than 1 simple thread. A download manager will not help unless you are downloading segments, but that is not the case here. I'm simply downloading single normal size files 4-5x from the same server in order to reach 7 mbytes/sec total downstream.

 

[bill001g]

You can't change download method to UDP unless you have some control of both ends. It is normally only a solution in a corporate environment where you control the servers and the clients.

I deal with this issue every day. It pretty much can't be fixed it is a fundamental "flaw?" in TCP. Because TCP acknowledges all data sent to confirm it was correctly transmitted and there is a limit to how much can be sent before waiting for a acknowledgement you have delays waiting for this acknowledgement.

I can't tell you the number of meeting I have been in trying to explain why buying more bandwidth will not increase the speed transferring large file between the US and India when I already have huge amounts of unused bandwidth available.

This is the example that is normally used to explain the problem.

You have a huge pile of dirt to move. You also have the maximum size legal truck you can buy and it can run at the speed limits. So if the truck can hold 1/10 the pile it will take 10 trips to move it. So if the one way time is 1 hour it will take 20 hours to move the pile.

So if I add more lanes to the freeway does it cut the time if the freeway was not congested and I could run at the speed limit before you added the lanes? Hopefully everyone understands the answer is NO.

The only way to fix this is to split the pile and add more trucks which is the same as adding more streams in a data network.

Adding more bandwidth is the same as adding lanes it does no good you can only use so much. The truck size is similar to the TCP window size and the speed limit on a network is pretty much related to the speed of light.

There is no solution on a public network to fix this really. Bit torrent is the closest you have and all this really is a variant of the chop the filing into pieces and run multiple streams.

 

[unknown66]

I modified the TCP/IP Buffer size for upstream and downstream to 1024 (the max). My upstream is now maxed out at 8 mbytes/sec on a single thread. My downstream however remains unchanged and is stuck at around 2.8 mbytes/sec. From what I've been reading there are some expensive UDP transfer tools available and they are vastly superior to TCP perhaps even 300x as fast in some cases. So I'm currently looking for some UDP apps that may work on Windows, as most work soley on unix.

 

[psgordon]

I sell FiOS in the NYC area and rarely hear about these kind of problems. Generally people are exceeded the promised up- and download speeds.

However, when I do get calls about speed issues I refer customers to tech support (800-VERIZON). My experience is that they are able to solve the issue remotely or, occasionally, via a site visit.

 

[AntiZig]

in other words, you have no idea what this thread is talking about. Nice talking

 

[psgordon]

In other words you're a geek who'd rather tinker for hours then get to the core of the problem with advice from real experts.

Mot nice talking.

 

[AntiZig]

lmao, yeah calling that number will get you the REAL experts, like maggie, who's got a high school diploma and a 2 day crash course in their customer service training.

but to answer your ad hominem, you're contradicting yourself, because a geek doesn't need hours to fix a simple problem. Don't they teach you anything in those good-for-nothing marketing seminars?

 

[psgordon]

This is pretty silly and I'm sorry I continued it. I apologize.

Again, my experience has been pretty good with the reps in Tech Support (although I really have a very limited knowledge).
They're generally very patient and very knowledgable.

The seminars are pretty much good-for-nothing. Most of what we learn is on the job. We're also sales people, no technicians.

I'm no Verizon shill. I work for a sales organization that is hired by Verizon - we're independent contractors, eating what we kill. I have NO loyalty to these corporate animals. However, I've probably sold over 1000 Triple Plays and people rarely complain about the internet speed. It's just the opposite - they're thrilled to be rid of Time Warner and their slower speeds.

If I go to speedtest.net will I get an accurate value? If so, my customers are usually getting what they're paying for.

Happy Holidays!

 

[michaelbrenden]

EXACT same thing happens here in Washington, DC while using 75/35 "business class" Verizon FiOS.

Transferring single large file between two overpowerful linux servers:

# axel -n 1 http://somesite.dom/100MB_file.zip
xfer speed = ~ 600kbps

# axel -n 10 http://somesite.dom/100MB_file.zip
xfer speed = ~ 4Mbps

# axel -n 50 http://somesite.dom/100MB_file.zip
xfer speed = ~6Mbps

# axel -n 100 http://somesite.dom/100MB_file.zip
xfer speed = ~5Mbps
(note increasing load of 100 parallel downloads slightly slows total speed)

One server is on 10Gbps connection at co-location site.
Second server is on Verizon FiOS, Washington DC.

This speed is nowhere near the advertised 75Mbps.

For 12 months so far, test transfers to or from anywhere using 75/35 Verizon FiOS have NEVER exceeded 10Mbps and are often only 1/10th to 1/100th and even (frequently) down to 1/1,000th of 75Mbps - as in 70kbps during "rush hour" transfers between evil / overloaded peers.

Tests using UDP (instead of TCP) are generally faster (unless they are encrypted OpenVPN, which is limited by single-CPU-bound openSSL encryption library), but reaching the mythical 75Mbps on one connection between two machines has so far eluded all minds, all machines, all attempts.


I'll add that, when all is said and done, the true "bottom-line" monthly bill for Verizon FiOS "business class" 75/35 + basic phone + basic television + HBO + all taxes, etc. is about US$350 - about $10/day, or 50 cents an hour, or a penny per minute...or so. I say "about" and "or so" because, to date, we have never received the same billing amount, over 12 months -- it seems there is always some change we're charged for, and the overall trend has always been up. However, this is still a 20% (or so) savings over the old solution that (masochistically) involved Comcast cable "high-speed internet", which was, bar none and by quite a large margin, the very worst "service" I have ever encountered or endured.

I'll also add that the reliability of FiOS has been incredible, almost legendary (time will tell if/how it matches the POTS phone service), and it starkly reveals our ten-year (monopoly) comcast fiasco for what it was.

 

[the great randini]

pathping is a good way to see were the slowdown is, takes about five minutes to complete


http://technet.microsoft.com/en-us/library/cc958876.aspx still works in windows 7

 

[bill001g]

If you have control of both ends try using IPERF and set it stream UDP packets at whatever rate you want. You can then see the loss rate on the other end. You do need to be a little careful with this tool some ISP will detect it as a denial of service attack if your run it for too long.

We use this all the time to load our internal corporate circuits and make sure the QoS is working as we purchased from the vendor.