Answer the typical os uses variants of lru which are

Answer: The typical OS uses variants of LRU, which are cheaper to
implement than LRU, for buffer replacement. LRU and its variants are
often a bad strategy for databases. As explained in Section 10.8.2 of the
text, MRU is the best strategy for nested loop join. In general no single
strategy handles all scenarios well, and the database system should be
able to manage its own buffer cache for which the replacement policy
takes into account all the performance related issues.
Many operating systems provide mechanisms to lock pages in memory,
which can be used to ensure buffer pages stay in memory. However,
operating systems today generally do not allow any other control on page
replacement.
10.17 List two advantages and two disadvantages of each of the following strate-
gies for storing a relational database:
a. Store each relation in one file.
b. Store multiple relations (perhaps even the entire database) in one
file.
Answer:
a. Advantages of storing a relation as a file include using the file sys-
tem provided by the OS , thus simplifying the
DBMS
, but incurs
the disadvantage of restricting the ability of the
DBMS
to increase
performance by using more sophisticated storage structures.
Downloaded by Harsh Rawat ([email protected])
lOMoARcPSD|16890397

94
Chapter 10 Storage and File Structure
b. By using one file for the entire database, these complex structures
can be implemented through the
DBMS
, but this increases the size
and complexity of the
DBMS
.
10.18 In the sequential file organization, why is an overflow block used even if
there is, at the moment, only one overflow record?
Answer: An overflow block is used in sequential file organization be-
cause a block is the smallest space which can be read from a disk. There-
fore, using any smaller region would not be useful from a performance
standpoint. The space saved by allocating disk storage in record units
would be overshadowed by the performance cost of allowing blocks to
contain records of multiple files.
10.19 Give a normalized version of the Index metadata relation, and explain why
using the normalized version would result in worse performance.
Answer: The Index metadata relation can be normalized as follows
Index metadata(index name , relation name , index type)
Index Attrib metadata (index-name , position , attribute-name)
The normalized version will require extra disk accesses to read Index
Attrib metadata everytime an index has to be accessed. Thus, it will lead
to worse performance.
10.20 If you have data that should not be lost on disk failure, and the data are
write intensive, how would you store the data?
Answer: A RAID array can handle the failure of a single drive (two drives
in the case of RAID 6) without data loss, and is relatively inexpensive.
There are several RAID alternatives, each with different performance and
cost implications. For write intensive data with mostly sequential writes,
RAID 1 and RAID 5 will both perform well, but with less storage overhead
for RAID 5. If writes are random, RAID 1 is preferred, since a random block
write requires multiple reads and writes in RAID 5.