Creative Misconfiguration

December was another month of ResNet’s continual instability and, despite lower load due to people not being around for the winter break, it still managed to have more outages than any of the previous three months.
Due to a power outage scheduled with insufficient notice, I was unable to log around four days worth of data. But even with fewer days, ResNet was out in some form or another 126 separate times, which is more than any other month that detailed logs were kept for.

December 13th was notable for thirteen separate outages, which is among the highest I have seen. There was high latency (~600ms to about 3s) to the gateway for about two hours on the 7th.
Additionally, there was a rogue DHCP server causing some people connectivity issues, and somehow the ISP left it going for nearly a week. Whether or not they took it off or the owner fixed it is unknown.

There were no major outages, save the internet going down when the power was out. Why there are not have battery backups on their equipment is rather strange, but given the track record of the equipment, rather unsurprising. It would have been nice to have some notice that the internet was going to be interrupted when the power was out, but the ISP and Residence do not seem to make any attempts to keep their customers informed.

Interestingly, there were more minor outages this month than any month prior, but unlike the previous two months, no major outages occurred. I was actually hoping to see a few days where the internet worked properly,over the winter break when most residents were at home, but true to form, ResNet couldn’t even handle the reduced load placed upon it.

Unfortunately, I can’t give an overall uptime due to a technical glitch caused by the power outage, but I do have enough data to give average speeds that the network performed at.
Downstream speed continued being poor at roughly 12Mb/s (or roughly the same as last month’s), while upstream was also a poor at around .72Mb/s.
Average ping to the gateway, on the other hand, was up roughly 15%, which on a roughly static network, is a little odd. It does seem to indicate poor network management practices.

All in all, I wish I had something better to write about, but the network continued to fail at anything other than basic service. Problems such as poor network configuration (manifesting itself as traffic shaping on the outbound link to SFU, which should not be shaped at all) still plague the network, even though some have existed for quite some time.
All in all, another month of lame service for too much money.

I just went through my logs to see when ResNet up here was down or otherwise having issues. Since I moved in September 6th (and started keeping track), there wasn’t a single day where the internet wasn’t out.

Between September 6th and 30th, the service was down briefly 64 times. September 27th was notable for having seven different small outages, while the 17th and 28th saw five each.
Also notable was a general instability for two and a half hours in the late afternoon of the 8th, a few hours of connectivity issues to the internet (but not internally) early on the 10th, two hours of slowness on the 17th and DNS issues on the 27th.
For comparison purposes, over the summer there were many fewer outages (I logged less than eighty minor ones and a handful of larger ones for the whole four months I was here for summer semester) and stability seemed much higher.

Another interesting note was that average ping on the internal network nearly tripled between my logs for the summer and the average ping times for September, with latency to the rest of the internet about doubling. Connection speed between the summer and September’s average (as measured by scheduled connection speed tests) was about halved from around 39Mb/s to around 17Mb/s. There is also much more variation show over the day, with speeds getting up to the 65Mb/s range in the early morning hours, and dropping to 7-12Mb/s during peak hours.
Upload speeds are even more abysmal than they were in the summer, averaging around 850kb/s, down from an average of around 1.5Mb/s in the summer.

The conclusion that I can draw from this is that the network is just not meant for the load it is being subjected to. Over the summer, when residence was at half-capacity or less, the network was generally fast and quite stable. One cause of the increased latency and decreased transfer rates could be that the outgoing connection is not fast enough for the load it is under, which is likely given was (last I had heard at least, it may have changed) a paltry 200Mb/s.