Great lecture :) I missed why the combination of No-Steal+No-Force and Steal+Force is not discussed. The paper "Principles of Transaction-Oriented Database Recovery" briefly elaborates on them. Is it the case that modern systems don't fall in those categories anymore?
@AshishNegi1618
4 жыл бұрын
In steal + force => multiple concurrent transactions will modify but on commit for each tx, we need to flush. Since we are also flushing uncommitted changes of other txs, to keep track of them, we also note them down (like in WAL).. but then it is double write penalty or amplification. :( In no steal + no force => only 1 tx can change a page at a time and on commit we don't flush it => we use WAL for durability => so it is like non concurrent version of steal + no force (WAL) :( seems like both are non-optimal versions.
@abdelrhmanahmed1378
Жыл бұрын
when the changes we made and recorded inside the wal get reflected to the actual pages of the table ?
@xiaohutang5666
3 жыл бұрын
WAL starts from: 41:34
@anarionzuo1425
3 жыл бұрын
So is the baby yours?
@pb25193
4 жыл бұрын
Today I realized that DJ DROP TABLE has to be written on all caps, the way you will write it in SQL while deleting tables. Am I right Andy?
@andypavlo
4 жыл бұрын
Yes
@tanzilurrahman1999
4 жыл бұрын
@@andypavlo I'd always found it a peculiar DJ name. Makes so much more sense now
@mazenezzeddine2188
4 жыл бұрын
Congratulations! Like Father Like Son. Long live for you all.
@cydiater
2 жыл бұрын
the start is so funny🤣
@kechenghuang845
3 жыл бұрын
are there any possible no-steal+no-force solutions?
Пікірлер: 14