CMU Database Systems - 03 Database Storage I (Fall 2018)

[@alfin3644]

Log-structured have a serious flaw: data ends up spread out wide and hard to cache. It's easy to write log-structured storage so it's very appealing. But only niche database use cases of dominant writes would make it competitive.

[@scarface548]

So Just watched influxdb talk, it uses log-structured and robinhood hashing both of which he said were not that great :\.

[@pabloacuna2176]

Are you talking only about databases? Because Kafka is log-structured and it can handle reads/writes really well

[@DenisG631]

I can not express my respect enough towards you for uploading this videos and making it free. Not mentioning the HIGH QUALITY of teaching! The real MVP! Thanks a lot!

[@rajaths4860]

CMU database group is truly making a great difference to students across the globe by posting such high quality videos! THANKS A LOT!!!

[@KyleRichter23]

This is awesome and easy to understand. I am so glad that this is on KZbin, Andy you are truly inspiring and please keep doing what you're doing!

[@yl6193]

This is really a very high quality study material. Thanks to Prof. Andy and CMU.

[@정동호-i1r]

Admire for sharing knowledge to the people. Thanks a lot.

[@anmolkhemuka5796]

A big salute to you for 58:27, respect++

[@fxrcode7923]

Why there's beep 31:10?

[@andrew_1204]

maybe there is a F word.

[@izharkhankhattak]

*_Excellent job, man!_*

[@dataaholic]

Slides PDF and Lecture Notes PDF links are not working. Please help

[@КириллПальцев-ю1о]

If most databases don't use mmap, do they use O_DIRECT (=disable page cache in OS for reading/writing file) and rely on their own buffer subsystem?

[@FF-ne2qz]

yes

[@bhavikpatel1930]

while doing tuple lookup by tuple id, how do we find the file which contains the page having the tuple?

[@andypavlo]

TupleId is a pair (PageId, Offset).The DBMS's page directory maps PageId to file location.

[@bhavikpatel1930]

Andy Pavlo Thanks for the reply, from the slides it looks like directory map is part of file along with data pages, so my understanding is that if you want to find tuple with ID T1 then you have to go through all the files and check in the directory map if that tuple is part of the data pages held by the file.

[@ankitsablok952]

Is the new Intel technology you are talking about this one : www.intel.com/content/www/us/en/architecture-and-technology/intel-optane-technology.html?

[@scarface548]

which textbook does this course use?

[@andypavlo]

15445.courses.cs.cmu.edu/fall2018/syllabus.html

[@Originalimoc]

43:50 "How do store pages in a tuple"😂

[@therealjayz8036]

41:44

[@therealjayz8036]

1:05:30