51–60 of 66 posts
This sounds more like an attack on the AWS hardware (Nitro?) . Does Nitro not have RDMA? I'm very curious if someone can discuss this. (Currently studying for AWS Solutions architect exam)
https://aws.amazon.com/about-aws/whats-new/2020/11/efa-suppo...
They don't even bother trying to make this sound believable. You claim it's 1.5x-2x faster? Sure I might believe that despite your prohibition against benchmarks. 10x faster? You better show how your engineers pulled out that miracle. 1000x? You've either picked out a very weird edge case, used something in a way it's not meant to be used, or just made that up. Anyone who's done any serious benchmarking in a competit…
Python vs C/C++/Rust/Zig has even more startling performance differences. If you’re looking for benchmarks of the same classes of software, among techempower benchmarks, the fastest on the charts are often well over 1000x faster. Being 1000x slower than shit needs to be is, very literally, the default state of the vast majority of software today. I don’t even blink at 1000x claims cause that’s just the standard gains…
There's no way you're seeing a >1000x speed difference in Python/Ruby/JS and C/C++/Rust/Zig/D without building some heavily contrived comparisons. That is not a normal differential.
The "high-level" languages are slower, certainly, but usually only on a single order of magnitude.
They don't even bother trying to make this sound believable. You claim it's 1.5x-2x faster? Sure I might believe that despite your prohibition against benchmarks. 10x faster? You better show how your engineers pulled out that miracle. 1000x? You've either picked out a very weird edge case, used something in a way it's not meant to be used, or just made that up. Anyone who's done any serious benchmarking in a competit…
Title is a bit misleading, since Oracle is the brand and well as the a specific database product, he is claiming that Oracles cloud hosted version of mysql is faster than AWS Aurora, not claiming that Oracle DB is faster than mysql.
That is _probably_ what he means, but there is nothing in the quotes in the article to confirm that is _actually_ the case. Also, the comment at the end of the article "... RDMA and other key drivers behind Oracle’s ascent to being the world’s hottest major cloud vendor" demonstrates that the person who wrote the article is not interested in asking hard questions. Oracle Cloud is quite a way from being the "hottest c…
> "We’ve announced a new database — a new version of MySQL with a fast query processor called HeatWave — and we have customers moving from Amazon Aurora where they’re experiencing a 1,000x speedup versus Aurora."
> we have customers moving from Amazon Aurora where they’re experiencing a 1,000x speedup versus Aurora. > So we’re a thousand times faster in query processing than Amazon’s version of MySQL No, “some customers are experiencing 1,000 speedup”, even if true, does not imply that, in general, your db offering is 1,000 faster.
Earlier quoted context omitted.
In that case the fairer comparison should have been to Redshift, though I will concede that Amazon walked into this one themselves by overhyping Aurora's OLAP capabilities. OLTP != OLAP, and the sooner database vendors learn to stay in their lane and focus on one use case, the better.
Oracle claims it works for both OLTP and OLAP. I mean on the fundamentals Heatwave adds a giant memory cache / query engine in front. That alone makes it faster than anything disk based. As to Redshift, yes, they claim it's faster than Redshift too by quite a bit. (I use claim as it's against the ToS to benchmark it independently)
Being in-memory does not make OLAP databases faster. And typically, "in-memory" is a disadvantage and a limiting factor in OLAP database architecture.
I have a presentation on this topic: https://presentations.clickhouse.com/meetup53/optimizations/
It is similar to how frequently someone says "lock-free" to highlight imaginary performance advantages. If someone says a database engine is fast because it is in-memory and lock-free, it smells wrong.
Perhaps it's better now?
Earlier quoted context omitted.
No, he’s not complaining about the marketing. He’s basically saying cloud computing is nothing, just a redefinition of existing computing. He’s not complaining about the marketing. He’s saying that cloud computing is only marketing. Clearly he was wrong.
Was he clearly wrong though? I listened through the whole clip and it seems like he was spot on to me. Steve Jobs made similar comments about cloud in the 90s - it’s really just mainframe 2.0.
Earlier quoted context omitted.
No, he’s not complaining about the marketing. He’s basically saying cloud computing is nothing, just a redefinition of existing computing. He’s not complaining about the marketing. He’s saying that cloud computing is only marketing. Clearly he was wrong.
He's wrong like those "There is no cloud, it's just someone else's computer" t-shirts are wrong. IE: not wrong, but too literal and short sighted.
> IE: not wrong
Private cloud is cloud, and its not someone else’s computer, so, “not wrong” is not not wrong.
Earlier quoted context omitted.
Or that it's easy to contrive a benchmark to show anything you want, and they don't want the negative press.
Then why doesn't every other piece of software in the world ban benchmarks? Is Oracle really leading the charge into the future?
Lots of commercial software does. SQL Server does (or did in the past), Atlassian does, etc.