1

What's a good way to benchmark a network connection? I'm looking for something like the Windows Experience Index, but for my internet connection.

WEI

For example, one place I visit has what seems to be a terrible connection. Pages load slow, etc. but file downloads are insanely fast once they get started (bandwidth good, latency bad?).

Alternatively, while tethering through my phone, browsing the web is nice and brisk, but downloads are excruciatingly slow (bandwidth bad, latency good?).

What's a good way to evaluate these variables (bandwidth and latency?) and are there others (load?) I should consider?

Michael Haren
  • 1,301
  • 6
  • 17
  • 31
  • bandwidth and latency are the only two meaningful generic measures of an Internet connection these days. Packet loss might be an issue in certain situations however (particularly wireless) – Chris S Jul 18 '11 at 20:55
  • @Chris are there tools that benchmark performance by simulating actual browsing (i.e. port 80)? I want internal filters, etc. factored in to my results. I use fiddler but it's not really convenient for amassing lots of test results – Michael Haren Jul 18 '11 at 21:00
  • Your internal filtering and caching would screw with results like that. A 100% cache hit rate could make a 300 baud modem look like the world's fastest Internet connection. Similarly the worlds fastest Internet connection could be brought to it's knees by running the test through an 8086 powered firewall. What use would numbers like that be? – Chris S Jul 18 '11 at 21:04
  • 1
    If the numbers represent real-world experience, I don't care if I have cache hits. Of course building this into the benchmark would be necessary – Michael Haren Jul 18 '11 at 22:41

2 Answers2

6

If you're looking for a quick and dirty bandwidth/latency test then http://www.speedtest.net/ works. It will give you latency measurements and up/down bandwidth. Loss and Jitter are other variables to measure, but less commonly an issue.

Evaluating overall experience depends on application. Some uses are latency sensitive, others are bandwidth sensitive. FTP file transfer experience is mostly about bandwidth, and a VoIP call is all about latency. Personally I consider latency to be one of the most important factors. Lower latency is always better, and bandwidth only matters when you don't have enough.

Edit: From your comments it appears that your focus is web browsing behavior from the endpoint. In many cases your web browsing may be slowed even when your bandwidth/latency are fine (You could have a slow DNS response for instance). I don't know of a tool which could accurately test for all cases of "browsing is slow". I wouldn't trust such a tool even if it existed. There are too many caches involved that you'll never know about as an endpoint. Bandwidth/latency will be the culprit most often, and trying to create a tool which would test for everything else would be difficult. If I were trying to generate information like that I'd probably use a sniffer.

Ryan
  • 912
  • 6
  • 12
4

http://speedtest.net/ and http://pingtest.net are good starting places -- they'll give you a rough idea of bandwidth and latency/jitter.
Make sure you run your tests to more than one endpoint, and at different times of the day. This is especially important if you're on a shared connection (Cable).

When conducting tests make sure you are eliminating intermediate sources of problems (e.g. don't test over wireless - Ideally you should be plugged directly into your network interface device (modem or router) to test your ISP's connection, and if that's OK start looking at the other internal components as possible trouble spots).

In terms of evaluating your connection, your perceived bandwidth is a function of several factors: Your real bandwidth (how much can be pushed through the pipe), your end-to-end latency, and the percentage of packet loss are the big three. Your provider's oversubscription rate (how much more bandwidth they've promised than they really have available to deliver) is the other factor (which usually manifests itself as packet loss when their routers start dropping queued data that they can't get out the door quickly enough).

Ultimately the evaluation is based on "Does the connection suck too much to be usable", and that's a subjective/qualitative call based on what you're using it for.

voretaq7
  • 79,345
  • 17
  • 128
  • 213
  • I think a big factor I might be hitting is internal filters. i.e. my pings and speeds are amazing, but browsing is terrible. Maybe this company is filtering me through an overburdened proxy – Michael Haren Jul 18 '11 at 20:58
  • distinctly possible, especially in larger corporate environments. – voretaq7 Jul 18 '11 at 21:13