3 ms·
Looking at the reported patterns of the spikes (just a few widely seperated sensors having their stored aggregated data routinely being corrupted at similar tim
by defrost 3y ago
Looking at the reported patterns of the spikes (just a few widely seperated sensors having their stored aggregated data routinely being corrupted at similar times of day) there's another highly probable cause.
Brittle software and data management.
According to the linked third party piece and the Wired reporting things are going wrong at the first aggregate server that collates all the sensor data.
The first port of call for the regular data admin HN crowd would be bad cron jobs, sloppy scripting, glitchy deamon whacking memory, .. a thousand possible causes all of which have happened before in the history of remote sensor data aggregation.
It's likely persisting as ground zero is in a war torn country at a place that likely faces time, resource, and staffing challenges; and it's likely being "ignored" at the higher action levels as the data has been looked at and random spikes are being filtered away and there's a big yellow post it note somewhere to look into the root causes as soon as possible (which may take some time).
- mannykannot 3y agoAny such investigation would also want to look for an upgrade or reconfiguration at the time of the invasion, to explain the prior absence of the phenomena. Some sort of reconfiguration to work around problems caused by the invasion, maybe?
- defrost 3y agoNo shortage of suspects - see the linked article - gear was ripped out by the Russians left right and centre, very likely data routes | repeaters have been cut | bombed | replaced. Even with a distributed network of sensors largely remaining untouched there's still been sufficient kicking of the system to make "supposedly robust" data pipelines fail in interesting ways. On the central aggregate server there's a good chance some well written never failed script is encountering a never before experienced failure mode on one or two checks, causing a cascading issue that spills the coffee at 11:11 pm every third day. Heiseinbugs are the worst.
- nullc 3y agoHm. if the sensors are free running counters and interruptions to the collection infrastructure cause the counts from multiple time intervals to be combined this would result in a false rate spike.