S6E10: "You Can't Scale a Fish"
Posted
Modified
Comments
0
Previous Episode | Next Episode |
---|---|
S6E9: "The Web of Tomorrow" | S6E11: "Ransom? Where?" |
Recorded (UTC) | Aired (UTC) | Editor |
---|---|---|
2021-06-25 03:51:23 | 2021-07-04 17:22:08 | "Edita" |
Format | SHA256 | GPG | Audio File |
---|---|---|---|
MP3 | 57e794adbc95eac357dc864991c46978eca215c3c14d951f7cd5feb8986faa4f | click | click | OGG | ddf9825c6af56771a9a7e9d0b5040291865c80a113e73d82e82890573a38c71e | click | click |
We talk about scaling up (AND down) systems.
Just the Tip
- John McAfee was found dead in his cell…
- But it would seem that he was Epsteined.
- He actually did literally fuck a whale.
- Paden talks about
seq
in bash.- I mention it’s deprecated.
Notes
Starts at 13m39s.
I was drinking a Founder’s Breakfast Stout. Paden was drinking PBR. Jthan was drinking Boulder Spirits High-Rye Bourbon.
- Scaling up (and down)
- We talk about scaling up here and here.
- Something we didn’t touch upon in the episode but should have is lately, the general application of scaling is not a long-term planning implementation but an attempt at automating a near-instantaneous process.
- Because of “the Cloud”.
- This will only end in tears.
- Scaling down is a good solution for budget concerns, underutilization, increasing lifespan of hardware.
- Scaling down to improve your budget consumption may have negative consequences – the next period’s budget allocation may then be decreased to match, and you’re in trouble if you need to refresh or grow your hardware deployment.
- Jthan refers to the “newer” theory of scaling, which he claims is different (but it isn’t, really; it’s just a smaller timescale and lacks any sort of direct planning stage).
- He thinks downscaling is an issue because you can’t determine how or when to scale down. It’s as much of an issue as upscaling automatically, you just apply the metrics in reverse. It’s the same exact methodology. You deal with the same problem with the same unknowns and use the same methodology.
- The entire reason you need to scale up is because you don’t have a crystal ball.
- What Jthan thinks is a big limiting factor here is e.g. bringing servers out of rotation. If you’re using a balancer/reverse proxy worth its shit, this is a very solved problem.
- He thinks downscaling is an issue because you can’t determine how or when to scale down. It’s as much of an issue as upscaling automatically, you just apply the metrics in reverse. It’s the same exact methodology. You deal with the same problem with the same unknowns and use the same methodology.
15 Clams
In this segment, Jthan shares with you a little slice of life. The title is a reference to this video. (2m16s in)
Starts at 49m33s.
Jthan doesn’t know the difference between a canoe and kayak. He linked to this page but those absolutely are canoes, not kayaks. They’re SMALL canoes, but they’re still canoes.
He saw a bunch of trout or something on vacation. And some ducks. I think. It was hard to follow him.
Errata
- Yes, Jthan, you can use homeostasis in the context of computer systems. Notice it says “ESPECIALLY the physiological system of higher animals”, not “ONLY…”
Music
Track | Title | Artist | Link | Copyright/License |
---|---|---|---|---|
Intro | TSP Meets El Primo | MUTE | click | CC-BY-NC-ND 4.0 | Outro | A Short, Destructive Ballad | Noi2er | click | CC-BY-NC-ND 4.0 |
Author
r00t^2
Categories
Season Six
Comments
There are currently no comments on this article.
Comment...