Carnegie Mellon Univ. Dept. of Computer Science 15-415 - Database Applications

1
Carnegie Mellon Univ.Dept. of Computer
Science15-415 - Database Applications

• C. Faloutsos
• Advanced topics RAID, authorization

2
General Overview

• Relational model SQL db design
• Indexing Q-optTransaction processing
• Advanced topics
• Distributed Databases
• RAID
• Authorization / Stat. DB
• Spatial Access Methods
• Multimedia Indexing

3
RAID

• Redundant Array of Inexpensive Disks
• Motivation
• higher throughput
• cost
• But reliability?

4
RAID contd

• Main idea striping (bit- or block-level)
• Level 0 just striping
• Level 1 plain mirroring
• Level 2 E.C.C.

5
RAID contd

• Level 3 Parity disk bit-level striping
• Level 4 Parity disk distributed parity
• Level 5 Parity disk distr. parity block-level
  striping

6
RAID contd

• additional levels (not examined)
• goals
• small/large reads
• small/large writes
• smooth operation during a failure
• smooth catch-up

7
General Overview

• Relational model SQL db design
• Indexing Q-optTransaction processing
• Advanced topics
• Distributed Databases
• RAID
• Authorization / Stat. DB
• Spatial Access Methods
• Multimedia Indexing

8
Authorization

• grant ltpriv.-listgt on lttable-namegt to ltuser-listgt
• privileges for tuples read / insert / delete /
  update
• privileges for tables create, drop, index

9
Authorization contd

• variations
• with grant option
• revoke ltpriv.-listgt on ltt-namegt from ltuser_idsgt

10
General Overview

• Relational model SQL db design
• Indexing Q-optTransaction processing
• Advanced topics
• Distributed Databases
• RAID
• Authorization / Stat. DB
• Spatial Access Methods
• Multimedia Indexing

11
Stat. Databases

• goals allow
• aggregate queries (min, max, sum etc)
• accurate answers and
• preserving privacy
• In short impossible to have all three

12
Stat. Databases

• given a table of salaries and job-codes
• offender find the salary of the only PGM
• defender protect against offender

13
Stat. DB

• offender
• select sum(salary)
• where job-codePGM

defender - accept only queries with high of
qual. tuples (gtn)
14
Stat. DB

• offender
• sum(salary) of all
• sum(salary) of non-PGM

defender - accept only queries with small of
overlapping tuples (gtm)
15
Stat. DB

• offender
• - ask high-overlap queries far apart (or with
  diff. logins)

defender - add some noise to the response (!!)
16
Stat. DB

• offender
• - ask same query many times, and compute average

defender refuse to answer same query many times
17
Stat. Databases - conclusions

• goals allow
• aggregate queries (min, max, sum etc)
• accurate answers and
• preserving privacy
• In short impossible to have all three
• but can make intrusion too expensive