News for hackers. Stuff that matters.
Re:Introducing “Database Performance at Scale”: A Free, Open... by VirusNewbie on Thursday October 05, 2023 @01:53PM (#37781736)
I think it depends on the problem. I've worked at a tiny telco (<100 people) where MySql absolutely wouldn't have worked. We ended up using Scylla DB. We had billions of rows in single tables.

If you're launching an e-commerce site that sells shoes or something, yeah, you probably aren't going to sell a billion shoes every year so MySql is probably fine.

Finally, I use this example a lot: Github thought they could scale out MySql by writing their own Operator system that sharded MySql. They put thousands of engineering hours in it, and they still had global database outages that lost customer data.

You get that shit for free using Scylla/Cassandra (the ability to replicate across data centers and still fine grained control to do things like enforce local quorum to not impact write speed etc.)

They probably made the wrong choice.

Re:Introducing “Database Performance at Scale”: A Free, Open... by krembo on Thursday October 05, 2023 @02:12PM (#37781985)
Or just use RDS Aurora for same benefits out of the box at scale.
Re:Introducing “Database Performance at Scale”: A Free, Open... by pc86 on Thursday October 05, 2023 @04:29PM (#37783891)
You know as I get closer and closer to official graybeard status I start having more of a hair trigger when someone says "just use __________" as if that solves everything and doesn't just create a list of a dozen or two other problems.

Everything is tradeoffs, you need to look at what you're trying to do and see which problems you are okay with. Some of these will be show-stoppers at larger scale. Some might be show-stoppers now.

But "just use __________" is never the right answer for any serious discussion.

Re:Introducing “Database Performance at Scale”: A Free, Open... by VirusNewbie on Friday October 06, 2023 @01:03PM (#37793272)
Aurora won't work near as well for write heavy/time series type workloads. It's probably a good middle ground for a lot of applications though.
Re:Introducing “Database Performance at Scale”: A Free, Open... by tynorf on Thursday October 05, 2023 @03:34PM (#37783202)
Cassandra does not make you immune to database issues: https://monzo.com/blog/2019/09/08/why-monzo-wasnt-working-on...
Re:Introducing “Database Performance at Scale”: A Free, Open... by pmcf on Thursday October 05, 2023 @04:01PM (#37783563)
Yes. You still have to read the docs before randomly setting “auto_bootstrap” to false.

Or just use Astra and not worry about scaling your own cluster.

Re:Introducing “Database Performance at Scale”: A Free, Open... by tomcam on Thursday October 05, 2023 @04:29PM (#37783894)
Did GP make such a claim? I didn’t see it, but maybe the comment was edited.
Re:Introducing “Database Performance at Scale”: A Free, Open... by sciurus on Thursday October 05, 2023 @08:48PM (#37786050)
Are you referring to their use of Vitess?

https://github.blog/2021-09-27-partitioning-githubs-relation...

GitHub didn't write it, Google/YouTube did.