In Memory Storage

Question 1

What is Buffer Pool Overhead?

Answer 1

• accessing page through page table (lookup cost)
• calculating memory pointer to tuple

Question 2

3 Advantages of Column Stores for OLAP

Answer 2

• higher read efficiency if only few columns need to be accessed
• better compression schemes
• enables vector processing

Question 3

Advantage of Heavyweight compression

Answer 3

Data can be heavily compressed

Question 4

Advantage of lightweight compression

Answer 4

• queries can often directly operate on compressed data

Question 5

Basic idea of dictionary compression

Answer 5

Encode values in any column as integers

Question 6

What property does dictionary compression have to have?

Answer 6

• Order preserving

Question 7

Main problem with order preserving dictionaries?

Answer 7

Updates

Question 8

2 Solutions for updates on order preserving dictionaries

Answer 8

• store new values in separate partition and merge in regular intervals
• leave gaps between codes

Question 9

Basic idea of bit -packing

Answer 9

• Map a wide and sparse domain into a dense domain
• pack multiple codes in a processor word

Question 10

Run-Length Encoding

Answer 10

• runs of values are encoded as (value, start, length)

Question 11

Frame of reference encoding

Answer 11

• encodes values as offset from frame using a fixed number of bits for offset
• escape code indicates that next value cannot be represented as offset

Question 12

Differential encoding

Answer 12

• encode values as offset from previous value using fixed number of bits
• escape code can be used to mark exceptions

Question 13

Early Materialization

Answer 13

Decompress columns and reconstruct tuples as part of your table scan

Question 14

Late materialization

Answer 14

Wait as long as possible with decompression and tuple reconstruction

Question 15

Heavyweight compression use case

Answer 15

Accelerate I/O path from SSD using GPu