🎉🎉🎉🎉

Thanks a bunch for this presentation and very well made video. And for the recommendation for the Buffer Pool lecture

kzbin.info/www/bejne/epjZqJiXhKxjjKcsi=E7hqRytLSn0Vgc9r

kzbin.info/www/bejne/qJWYc3d_q5Ktb6ssi=ZwnFnINpLNRphOyg

膜拜膜拜

Impressive.

I‘m here to admire Xinjing Zhou and his presentation

Think your alternative titels have more poetry. Anyway, good stuff! Remember back-in-the-days when I was looking for a decent way to do IPC with windows32 and people pointed me to file mapping. Brand new feature back then. And my intuitive reaction “WTF, why solve a simple problem with a complicated solution that even misses the point?”

very useful! Thanks

audio really garbled

Would it be possible to try SystemD?

不知道周神什么时候开始好好学英语的，这个口音是很多高中就过来的小留都比不了的。 佩服佩服

I find the TROD fascinating. It seems like a game-changer for debugging distributed web apps. I'm curious, though, how does TROD handle scalability and performance when dealing with large-scale applications? Is there a specific trade-off or optimization strategy in place?

牛！

前来膜拜周佬

Good!

Thanks, your intervention saved a soul at exactly the right moment!

The great thing about this talk is you can replace DBMS with just about any real application that interacts with data and it's still true.

nice talk, really inspiring!!

This keynote is an inspiring look at developing systems in academia. A must-watch for anyone interested in the topic!

Very useful information and I appreciate it very much!! As for the title options you had, I think that the final one is the best. You have a very good article and something less agressive attracts more old guys like me.

It seems great. How could I test beta?

Filesystem exists for a reason. mmap only useful for accessing device file. If you want to handle real file, use fopen and friends instead!

Very interesting paper

as awesome as I expected

Very interesting, I would have liked to see a benchmark with VictoriaMetrics :-)

BTW, the letter 'x' is not an operator.

1 Parquet etc. are the file formats, while Delta Lake etc. are sth. in an upper-level abstraction. Maybe we can call it table format. SDC is just another table format - a self-evolving format. 2. Since the file formats are not optimized for performance, there is more room for research. 3. Given a workload, it is possible to optimize the data layout to get the best performance. We can do it for a traditional DB, e.g. partition a table horizontal or vertical. If most of the data is read-only, we may store it in multiple copies in different layouts. 4. How to evaluate the cost of data reorg? It is decided by the price model by the Cloud vendors.

Awesome presentation

Great presentation, Matei! Curious to see more about the Streaming engine

By far my favorite billionaire 🙂

I love your presentation. Compact and quite funny. Thank you for this. Interesting and useful information. Also, can't stress enough how much I appreciate that it's 10 minutes and not 1 hour long.

Build and run, then fix it if the product is still alive, which is the best practice of engineers. Make sure everything is under control and implemented in DBMS, which is the tradition of DB folks.

1:45 "maybe you're focusing on one of the new markets Andy is excited about, like web3 or blockchain" 😂😂🤣

So, who's going to fix mongodb and influxdb?

I'm sorry, but unless I'm missing something I thought Photon was not for machine learning and data science workloads but only for traditional workloads, as it is the engine behind Databricks SQL.

Great talk, Joe. Unlike Chapel, which is a full parallel programing language, MPI is a set of specifications to fully and specifically address the massage-passing parallel programing model.

The histogram is a step-wise fit of the true distribution so it is absolutely a learned index. The choice of variable bin widths ala Bayesian Blocks or with random averaged shifted histograms (RASH) would be an improvement to the fit.

Thanks for sharing this keynote! It's a great way to discover the vision that made Snowflake happen, straight from the founders.

This is awesome...

here is the link to our project: chainifydb.com/