November was another month of ResNet’s continual instability, clocking up 100 periods when the network was out in some form or another. This is down from 109 in October and 64 for part of September. There was only one major outage affecting only part of the network in addition to three extended periods of heightened instability.
November saw slightly fewer minor outages than October, with 100 outages of between about a minute (or less, the granularity of logging in only one minute) and around fifteen minutes.
To get the month off to a poor start, ResNet delivered very slow speeds during the evening of the first. Additionally, there were also seven outages on the first, the most for any day of the month.
There was also intermittent amounts of packet loss intermittently for seven hours starting in the early evening of the fifteenth which was a precursor of a major outage, similar to behaviour observed last month. Latency was very high on the evening of the sixteenth in part of the network, and was out for around nine hours on some segments of the network the day after, the sixteenth.
The 23rd was the last day with high latency and packet loss in the late evening.
Interestingly, the nine hour outage on the sixteenth was the same pattern of high latency and packet loss the day before as was the major outage in October. Once again, it appears that the ISP did not respond to the warning signs their network stability was declining.
I was unable to get an answer on why the internet was out for about half of Residence for so long, but if it was anything like previous issues, it was likely fully or at least partially preventable.
It is somewhat concerning that these major issues seem to continue to keep cropping up and that there has been little if any improvement in the network situation so far this semester (over three months so far).
On the plus side, the issue with slow routing from Residence to the SFU network might actually get taken care of as the ISP has finally (after eight months) figured out that the issue is at their end.
I was informed that this was due to the upstream traffic shaping to keep people from uploading too much (due to the low transfer of connection the ISP uses to get to the internet). Inexplicably, though possibly due to hardware or software limitation or misconfiguration, this is also being applied to routes over the dedicated link to SFU, which should not happen. It makes using remote desktop software virtually impossible, and file transfers are painfully slow.
As of this post, it was still not working properly.
The internet was up slightly more than last month, with an uptime of a (still) poor ~97.3% (or ~98.2% in the parts that didn’t experience the nine hour outage part way through the month). Average downstream speed was lower, around 12.5Mb/s (versus ~14Mb/s last month and ~17Mb/s in September) though upstream speed stayed about the same at about .75Mb/s (though lower than September’s .85Mb/s).
Average ping to the gateway was only slightly higher (around 2%).
Once again, the network seemed to struggle under the load that it was being put under, falling flat on its face partially once, highlighting the need for an infrastructure upgrade. (Last I checked, the network was running end-of-life Extreme Networks switches.)
Additionally, the upload and download speeds are still lacking, especially compared to the price we are paying for the service compared to the price of services from other ISPs.
Unfortunately, it does not look like the quality provided by ResNet is going to improve any time soon.