Post #825465
2025-08-21 02:04 UTC
Replies (21)
-
@mhoye@mastodon.social 2025-08-21 12:42
“So, do you have a _database_ or do you have a table that happens to be stored in a database” is not a question that makes you friends with CTOs it turns out. (It’s a table that happens to be stored in a database.)
-
@dashdsrdash@tilde.zone 2025-08-21 02:22
@mhoye When "Big Data" can be processed entirely with sed and awk on last year's laptop...
-
@michaelcoyote@mastodon.social 2025-08-21 02:23
@mhoye wat?
-
@admin@mastodon.slightlycyberpunk.com 2025-08-21 02:28
@mhoye Mailing it? "Millions of records" is what, megabytes? Maybe a couple gigs? You could send that over the network faster than you could get the card into a damn mailbox... ...although you could probably shove the data AND the computer to process it into a standard envelope. Might need a bit of extra postage though.
-
@ErgonWolf@pawb.fun 2025-08-21 02:33
@mhoye What's the old canard? Never underestimate the data transfer rate of a station wagon loaded full with magnet tape driving from A to B?
-
@neal@social.typica.us 2025-08-21 04:57
@mhoye I've heard that one measurement per second is "big data." It's not. (I don't think they appreciated hearing from me that I had previously done a usability study on exactly what they were talking about, that their users would be happier with a few times more data than they were collecting, and the size of the data would still be tiny.)
-
@adrake@sfba.social 2025-08-21 09:20
@mhoye we had a vendor that our data team was using for replicating our databases. Super expensive, and with a pricing model clearly (to me) unrelated to their actual costs. This came to my attention about a month before we were supposed to renew our annual contract. In that month I replaced them with some Rust that was an exact drop-in, but faster, with a couple of new features we had long wanted, and literally *two orders of magnitude* less cost. Hundreds of thousands a year to less than two thousand. The compute now easily fits on a laptop, but it does benefit from a local network connection to the database 😉 Their CEO was pissed we were leaving (because they had already given us a big discount) and absolutely certain there was no way we could possibly do it cheaper, and that we'd come crawling back. Given the obvious insanity of that position, my only explanation is he was drinking his own SaaS pricing Kool-aid.
-
@claudius@darmstadt.social 2025-08-21 09:25
@mhoye client: "we want big data" me: "so how many billions of entries are we talking about" client: "several THOUSAND entries!" Cute. So, not even a large excel sheet?
-
@zaphodb@twitter.resolvt.net 2025-08-21 09:30
@mhoye @tante .oO( SQLite on pingfs consumed by CF worker - et voila serverless - nobody said anything about restore )
-
@haui@mastodon.giftedmc.com 2025-08-21 09:31
@mhoye I call this "diseducation" but its probably just systematic disinformation. Companies like deutsche telekom use fancy corpo speak and systematically change their descriptions to exclude transparent and comparable terms. They instead call everything an experience now. Its part of monopolization. Disgusting. Thanks for informing people. :)
-
@lispi314@udongein.xyz 2025-08-21 09:37
@mhoye @rysiek The main difficulty in a migration is really just converting/formatting the data appropriately, assuming it's migrating between different software stack rather than merely locations.
-
@GreenSkyOverMe@ohai.social 2025-08-21 09:43
@mhoye Millions of rows is kind of the lower limit of anything I deal with … as long as the migration is done with database tools, not with APIs that trigger validations, this should be done really quickly
-
@simon@augsburg.social 2025-08-21 09:45
@mhoye LOL
-
@whynothugo@fosstodon.org 2025-08-21 09:48
@mhoye Mailing the SD card (or implementing RFC1149) has much higher throughput but the latency might be prohibitive.
-
@m@martinh.net 2025-08-21 09:57
@mhoye This Kubernetes cluster could have been an #EMFCamp badge... http://badge.kaimac.org/w/about
-
@maswan@mastodon.acc.sunet.se 2025-08-21 10:32
@mhoye So.. That's smaller in all ways than our DB for the preprod system that sees light functionality testing. Which is a default installed postgresql on a random VM, neither of which needed any tuning (unlike prod where we at least count the transactions in kHz).
-
@myfear@mastodon.online 2025-08-21 11:12
@mhoye rofl 😆
-
@jerryorr@fosstodon.org 2025-08-21 11:35
@mhoye Reminds me of the (now defunct) AWS Snowmobile service, which transferred exabytes worth of data via tractor trailer https://aws.amazon.com/blogs/aws/category/aws-snowmobile/
-
@papageier@digitalcourage.social 2025-08-21 11:52
@mhoye fastest bulk data transfer protocol: dhl://
-
@neel@monoxide.network 2025-08-21 12:12
@mhoye I worked at Microsoft for five years. IT departments are as prone to consumerism as mere mortals are. MS 365 is incredibly slow and unreliable. My self-hosted Postfix/Dovecot is faster and more reliable by a large margin. Yet IT is told they "need" 365 to "save money" when all it does is enrich MS partners. Same with "Big Data". People were told they need Hadoop and DataBricks when all they need is Excel or sed/awk.
-
@Colman@mastodon.ie 2025-08-21 12:16
@mhoye I'm dealing with a system with a full AWS deployment costing hundreds per month for maybe 250 transactions a day. It's a booking system for four small sites.