1:17:13 x100 dbms kernel also supports range indexes (called MinMax) as well as Netezza, there's a nice paper from Guido Moerkette: "Small materialized aggregates: A light weight index structure for data warehousing"
@foxl91952 жыл бұрын
1:12:00 Interesting bit about the future of custom hardware for database systems: "RISC-V stuff looks super interesting because you can start fabbing all sorts of specialized accelerators because there's enough space on a die".
@hatxu4062 Жыл бұрын
In block nested join, would it be better to use some buffers for inner table?
@jonkemp92112 жыл бұрын
Is the larger table the one with most rows, most columns or most data?
@moazmohamedhassanbayoumi4722 Жыл бұрын
I think the point here is most rows, because it loops on rows not on columns
@vaibhaves21119 ай бұрын
most no of pages
@hdrkn52472 жыл бұрын
How is the cost of partitioned hash join only 3(M+N)? For the first building phase, when hashing the key of each tuple and write it to the corresponding bucket (each bucket backed by a page I suppose), the writes to buckets would be random which would lead to random disk I/O. What am I missing?
@hanzhoutang9235 Жыл бұрын
I think professor mentioned at the beginning, when count the time complexity, we ignore the cost of writing.