GeekNights Monday - Computer Noise

Tonight on GeekNights, we talk about the noise that computers make. It used to be a lot worse, but most of you are probably too young to have experienced this era. In the news, Rym went to a Star Trek convention, Google supports passkeys, actually read this article and not just the headline, and Rym is running in the New York Marathon in November!

Things of the Day

Episode Links

On Patreon

The floppy drive sounds like a printer dying.

retvrn

3 Likes

The first science fiction convention as we would recognize them today (meeting guests of honor, costuming, etc) was before TV science fiction shows so was mainly for written fiction. That’s why “WorldCon” gets the name with no qualifications attached:

2 Likes

I know WorldCon, but I didn’t realize it was the first con.

So loud you can’t hear in the data center

The hard drives don’t like when you yell at them.

We should put bets on Rym’s marathon time.

1 Like

For reference, I can run a half without any trouble and without feeling tired in 2 hours.

I can run a half and be tired at the end reliably under 1:50.

My personal best half is 1:42.

The furthest I’ve ever run in one go without stopping is 17 miles.

This is a great video. I first saw it when it came out and I was working as an enterprise storage administrator. We had a lot of interesting discussions about things we could do to maximize/optimize performance and when we saw this it opened a whole new can of worms :sweat_smile:

I remember there was a storage array vendor that we bought hardware from, they mounted their disks in opposing pairs, top-loaded in a pull out shelf one facing forwards and the next backwards (back to back, instead of back to front as most other large disk arrays did, and still do I believe). The purported benefit (though hard to say how much benefit it had in the real world) was to cancel out vibration and oscillations in the chassis and especially if you had a whole server rack loaded up with these drive shelves.

image

Cancel out oscillations reminds me of that story about the IBM Black Team:

1 Like

That’s a great story, even if apocryphal!

Makes me think of my own “so weird I might not believe it if I hadn’t dealt with it personally” tech failure story.

So, around the same time as I was managing these SAN arrays I also had responsibility for some big clustered NAS boxes. One node of one of the NAS arrays would randomly-ish reboot itself once a week or so and otherwise showed no sign of ill-health, mismatched code, or suspected failing hardware. Yet the reboots persisted for weeks. Worked with the vendor and they couldn’t identify any reason or recommended remediation. In the data center we used these RS-232 to Cat5e serial console cables with patch panels distributed amongst the server racks allowing a remote admin (me) to connect to a device over its serial port via an Ethernet-to-serial go-between device. Normally they would be left disconnected until an issue occurred and we would create a ticket for someone on-site to go hook up the cable so we could remote connect to the server/storage array/NAS device and troubleshoot & fix it.

Someone had left this cable plugged into the patch panel and the serial port of this one NAS node and the patch panel port had its home-run Ethernet cable untwisted a little too much, such that random EM noise was being induced onto any wire plugged into that port which the serial console port was receiving. But since it was garbage the serial port couldn’t do anything with the small amount of data it was receiving constantly so it was just filling its buffer space (which was tiny compared to any then-current NIC) and when that buffer filled up it spilled onto the system bus which caused a kernel panic in the OS. When we watched how fast that buffer was filling up with data and extrapolated out the time required to overflow the available buffer memory it corresponded perfectly with the frequency of random reboots. Unplugged the serial cable and the node was perfectly stable.

So that is my story about how random EM noise caused me job headaches for weeks. As close as possible to a random technology issue being caused by solar flares as I’ve witnessed firsthand.

2 Likes

I will collect predictions, and keep them secret from Rym if you wish. DM me here or on the Geeknights discord.

Alternatively, post a hash of your prediction, like so:

sha256: 630dd468943fd3b26287355d1ebb015b24f5442b2c74b06ff3baa10bcb941656

I will reveal all the non-public predictions after Rym runs the race on November 5, 2023.

3 months until race day. You know you want to get in on the SaltNights action.