My ISP has a poor service record. I made a tool to test how shitty it is
iscableoneshitty.comEDIT: Well, this got more popular than I ever expected. Resizing the linode now. Should be back up soon.
After months of fighting with CableOne about speed and service problems, I decided to finally spice up my website dedicated to their poor service: www.iscableoneshitty.com
Now, instead of just saying "YES," it gives an actual representation of the quality of the connection to my house. The server pings my modem 5 times at the top of every minute. It takes the response times and averages them.
If the average response time is <60ms, it reads "NOT CURRENTLY."
If the average response time is between 60 and 100ms, it reads "MOSTLY."
If the average response time is >100ms and <150ms, it reads "YES."
If the average response time is >150ms, it reads "VERY YES."
I've distributed to friends and neighbors in the area, because almost every one in my town is on CableOne for lack of other choices, so they can check and see if at least someone else is having trouble.
I have thoughts of actually tracking the data and compiling monthly reports of downtime and slowness, I just haven't gotten around to that yet.
This exact thing is now becoming law in Canada for all ISPs.
I work at an ISP.. and ... ah.. they're shitting bricks about it, because now they'll have to live up to the speeds and reliability they've sold to customers :)
This could be expanded to include more companies. I think having a real time snapshot of the reliability of all companies would be cool.
They have one. It's called www.internetpulse.com :-)
I was about to ask about a chart of the historical shittiness.
I just went through a 2 month ordeal with my ISP trying to get basic service. It's working now so I think it's resolved for the moment but I don't have much faith in them. Unfortunately I'm expecting further outages.
I thought about making a tool like this as well as crazier ideas like getting my lawyer to sue them for their incompetence - as it happens they are my only option for broadband.
What I'd like and I think alot of other customers would like is a little app to run on my computers that tracks when my internet connection goes down and for exactly how long. I pay a fixed amount monthly. As I'm concerned that pro-rata amount of lost connection time should be refunded to me from the ISP. There is no incentive for them to get better unless they lose money from their shit service.
I think it'd be a good adware product to cover basic expenses, and maybe some money could be made by taking a peice of crowd funded lawsuits vs ISPs. With my ISPs customer base, I don't think it'd be too hard to find contributors...
It's more dramatic, and historically useful, to pipe the output from ping into Graphite, and keep a few months worth of data. Really shuts up the techs who come over and say everything's fine.
I dont suppose you have a tool that does this? I've been pondering building one, but I'd rather just skip to the part where I can start complaining about my service :p
It's not on a machine I have access to right now, but I'm fairly certain it went something like:
while [ "true" ] do printf 'google_ping %f %s' $(ping -c1 www.google.com | awk '/ms/ {print $7};' | awk -F '=' '{print $2};') $(date +%s) | nc graphite_host 2003 sleep 10 doneAwesome, thanks :)
fwiw I use
ping -n www.google.com | awk '/ms/ {split($7,p,"="); print "ping.google " p[2] " " strftime("%s")}; fflush()' | nc -q0 127.0.0.1 2003 &I bow before your awk-fu. :)
It's not pretty, but I use Routers2, which is a layer on top of MRTG. We have TimeWarner's commercial cable internet service, and our latency spikes to ~800ms at a regular and consistent interval. I've been 10 rounds with TimeWarner over the issue, but they refuse to acknowledge that it's an issue because they have no latency SLA. We ended up bringing in a second Windstream Ethernet over copper circuit for our VoIP application because the TW circuit latency was so bad.
I have a local VM with Routers2 installed, which pings our LAN router, the TW modem, the first TW hop, and one of our VMs at Linode. The latency spikes start at the first TW hop. Routers2 provides great insight in to how TW is performing.
You can get Routers2 here: http://www.steveshipway.org/software/
Any recent version of collectd has the write_graphite plugin. Mix that with the ping plugin and you're all set. About five minutes of work (assuming you have a graphite stack already). My raspberry pi runs graphite / statsd and my personal servers write all sorts of useful system / network metrics to it via collectd.
Collectd is pretty awesome for tracking the state of a machine. I use it all over the place.
I modified a munin plugin that pings google and charts it: http://imgur.com/jaaLaJP
smokeping does this.
came here to post about smokeping. here's their demo: http://oss.oetiker.ch/smokeping-demo/?target=Customers.OP
I worked at a company that made electricity usage monitors. The little boxes would send data to us with energy usage every minute.
We could see how often internet went down, based on how many times the boxes sent us "stored" data.
We didn't do any analysis on the data, but we did note that the residential internet while good isn't as always on as you think it should be. We were wondering how much of a problem this was going to be for home automation. (It would be frustrating not to be able to connect)
Also of note in the three years I was there before I left, its seemed to be getting better. It might be that the business switched more to corporate customers and there internet was better. We ended up doing some thermostat control and it worked for the most part (Most wifi thermostats are kinda crappy, thus the rise of the nest)
The question is, is it that corporate internet is that much better, or that they're just more diligent about making sure that their internet never goes down?
In other words: a residential customer is more likely to accidentally unplug their modem, or maybe flip a lightswitch that controls an outlet that the modem is plugged into, or accidentally flip the switch on their surge protector, or turn their modems off at night to save electricity, or to have old/bad cabling running from the telephone poles that need replacing.
>or that they're just more diligent about making sure that their internet never goes down?
I think that has something to do with it. Business too have blocks of time (nights) when the internet can go down and nobody is going to call for service. I wasn't involved with the network provisioning but often we were put on the same network as the "guest wifi" as opposed to the business network or the network with the cash registers (for security reasons)
It would be helpful to see this graphed over time. Also, how do you rule out service quality problems at the server's end?
Yeah, graphing is a thing I'll probably implement. It would be useful to use as a tool when fighting with them.
I currently don't account for that. I guess I could ping something always stable like 8.8.8.8 or something as a baseline to compare against.
Is the site up? I can't seem to get in. I really want a tool to have data to complain to my ISP. Shitty leased line service i am getting.
Cant open your website but it seems that you're creating something like IhateRogers.ca ---> a candian ISP :)
Hey, I'd recommend changing your font-size from 200pt to 16vw(ish). This should make it fully responsive.
Points are fully responsive. They correspond to the on-screen physical size of the text. (See http://en.wikipedia.org/wiki/Point_%28typography%29 )
If the text doesn't fit on your screen -as it fails to on mine-, then either your screen is physically too small (as mine is), or your user-agent (or windowing system, or display, or video card drivers) is broken.
While you're totally correct, the point implementation fails to display the content correctly across all devices, hence it's not really responsive in the standard 'responsive design' definition.
If your intent is to display letters that are X inches high, then points correctly display textual content on all devices that:
* Know how large their display is
* Know the display's DPI
* Can correctly communicate this information to the underlying windowing system and applications
(Note that it is highly unlikely that any high-end "mobile" device fails to perform any of the items in the bulleted list.)
The designer wanted letters that were five inches high. It's not the designer's fault that your screen is too small. :)
Do you disagree that it is important to have a display-density-independent method of specifying the physical size of text or (for that matter) an image?
Thanks! I'll look into that.
How did you implement this? I would like to use this to monitor my ISP as well.
Hmm, right now, your website has a poor service record as well.
That's unfortunate. My little 512Mb Linode apparently couldn't handle it. It's back up for the moment.
Edit: Scratch that. Resizing to a larger linode now. This got far more traffic than I ever expected it would.
Linode has a maximum packet per second limit set by default on all nodes. You'll need to email them to have them remove the limit for this particular box. They're very responsive and will increase it within a few minutes. No matter how big your instance is, there's probably way too many packets hitting it right now.
Ah. Good to know. This is just running on a personal test box. I'll email them, though I think the wave is over anyway.
Thanks :)