Disk based architecture
- the DBMS primarily stores stuff on non-violatle disk
- meaning: the DBMS’ job is to move stuff between violatle and non-volatle storage
Challenges
- how do you makue sure memory-writtten info is commited to disk?
Hierarchy model
there’s a tradeoff because as your system because faster, you have less of it and (in case its volatle) it stays around less long
Access Order
Key challenge: random-access on non-volatile storage is almost always slower than sequential access
Good DBMS algorithms tries to reduce writes to random pages so stuff is in continuous blocks.
Subchapters
Overall DBMS Goals
- We should manage databases that’s larger than memory available
- We should limit reading/writing to disk
- We should minimize random accesses
Lifetime of a request
- execution engine asks for a page, say page 2
- memory asks its directory for where page 2 is
- memory found it missing, so ask disk for page 2
- disk loads page 2 into memory page 2
- memory gives execution engine a pointer to memory