FiOS – Living with it for 3 weeks…

I have been living with a fiber optic based broadband connection (FiOS) to my house for 3 weeks now. Here are some of my impressions.

Downloads are faster but…

Network throughput is largely dependent on the size of the file being downloaded. This is because the two endpoint systems (your computer and the server you are accessing) progressively negotiate larger and larger packets of data if they find that the connection (i.e. the Internet) can handle them reliably. If the connection starts dropping packets, the systems will back off and reduce the packet size, thus reducing throughput. As a result, it takes time before two connected systems achieve a high steady-state throughput.

To put my connection through its paces, I tested my FiOS throughput by downloading some very large, publicly available files, from sources known for having huge pipelines to the Internet. Here are two examples:

• MSDN January 2006 Library – I downloaded the latest MSDN Library (2.8 GB)from Microsoft’s MSDN subscriptions web site using File Transfer Manager v5.0. The throughput was initially stable at 65 KB/sec for about 20 minutes. However, within about an hour, it had increased to 635 KB/sec or about 10 times faster. I don’t remember numbers even half this fast with my cable connection.
• Structure and Interpretation of Computer Programs lectures from MIT - This is the famous lecture given by Abelson and Sussman. You can download these fascinating 10 lectures from MIT’s servers as 20 MPEG files of about 1.5 GB each. You can find them here: http://swiss.ai.mit.edu/classes/6.001/abelson-sussman-lectures. I tested using the [download] option, not the [torrent] one. My throughput was about 593 KB/sec. Downloading two video files concurrently halved this speed. The speed was also halved when I downloaded the MSDN library along with one of the lecture videos from MIT.

My first conclusion is that I was able to max out the throughput of my FiOS connection to about 600+ KB/sec. Just a reminder that I have a 5 Mb/sec down and 2 Mb/sec up connection. Had I not maxed out my fiber connection, I would have seen no degradation in speed as I added downloads from a different source of servers. In other words, I should not have seen any drop in throughput, or at least a limited one, when I downloaded files from both MIT and Microsoft concurrently.

This conclusion begs another question: when will the content providers reach a limit (technical or financial) in their ability to deliver content at high-speed to more and more individuals who have fatter and fatter pipelines to the Internet. After all, many of us have a pipeline to the Internet that is fatter than the one used by our employer. NOTE: I am talking pure bandwidth here. Digital leased lines such as T-1s and T-3s have other desirable characteristics over home-based connection lines.

My second conclusion is that you probably should not pay for high bandwidth if you are mostly going to be downloading small files, say in the order of tens of megabytes. This is because the systems take a significant amount of time (in the order of an hour) to saturate a relatively large pipeline like a FiOS connection. It would have been very interesting to repeat this test with, say, a 15 or even 30 MB/sec down FiOS connection. If someone from Verizon FiOS is reading this, I would be VERY interested in a free upgrade to the 30/5 plan to test this and further report on my experience.

FiOS broadband is sensitive to power loss

I have found that my FiOS connection is more sensitive to losses of power than DSL and cable. This is obviously hard to verify because losses of power happen unpredictably in my area. Additionally, conducting a controlled experiment would be rather difficult. However, I distinctly remember being able to surf the net with my cable and DSL connections while experiencing power outages at my residence. This wasn’t the case with FiOS even though I have more battery powered UPSes than before. As I mentioned in a previous posting on FiOS, the UPS attached to the FiOS hub provides power preferentially to the phone system over Internet connectivity in the case of loss of power. This is important to remember for those of you who have selected a VoIP vendor as your primary phone system carrier. Hence the importance of having a fully charged cell phone around.

Can you hear me now? - Odd dropped connections

I have also experienced odd outages with my instant messaging software. My connections to both, the MSN and Yahoo servers sometimes drop over the course of a day. Although usually not often noticeable, it can be aggravating when you are in the middle of a conversation and the drop lasts a matter of a minute or more. Can you hear me now? Can you hear me now?

To try to resolve this problem, I changed slightly the configuration on my router. I went from ‘Connect on Demand with Max Idle of 5 min.’ to ‘Keep Alive: Redial Period 30 sec.’ on my PPPoE settings.

I am having trouble reconciling these intermittent outages with the fact that I’ve experienced fewer dropped PING packets over FiOS than over cable. PING packets are used to verify the end-to-end connectivity between two network devices or systems. NOTE: many network administrators disable PING responses for security reasons.

I tested the quality of the connections using the following command at a DOS prompt:

C:\>ping www.cisco.com –t > c:\temp\pingcisco.log

This command continuously pings a Cisco web server and redirects ( ‘>’ ) the results from the screen to a text file named ‘pingcisco.log’ that will be created in the Temp directory on drive C. This command will run until you interrupt it by pressing Ctrl-C. I chose to run this test for a couple of hours but you could easily do so for days on end. Now open the log file with a text editor like ‘UltraEdit-32’ (Thanks IDM for a great tool!) or using ‘Notepad’. Jump to the end of the file and one of the statistics you will get is the number of dropped packets (% loss). Usually, the fewer the packets lost, the better the quality of the connection. On the other hand, remember that the problem could be at Cisco’s or Yahoo’s end too, however unlikely this may be. NOTE: other parameters, such as latency time in particular, also reflect the quality of a connection. Latency is the time, usually expressed in milliseconds that it takes for a PING packet to reach the remote server and travel back: the lower the latency, the better the connection.

The table below compares the % of dropped PING packets between my cable and FiOS connections when pinging Cisco and Yahoo servers.

• Cable => Cisco (71% loss) => Yahoo (36% loss)
• FiOS => Cisco (53% loss) => Yahoo (0% loss)

I conclude from these data that the quality of the connection is better using FiOS than cable. This conclusion seems at odds with the experience I had had with my instant messengers.

Can any reader offer an explanation and a fix for this?