1. Distributed Database Management Systems
Lecture 25

2. Distributed Database Management Systems
Virtual University of Pakistan

3. In Previous Lectures: time 1:40-230
Reasons for Fragmentation
Maximizes local access
Reduces table size, etc.
PHF using the SQL Server on same machine

4. In Previous Lectures--2:00
DDBS design is a team effort
Involves
Domain Experts
DDBS Experts
Network Experts

5. In Previous Lectures
Implemented PHF in a Banking Environment (around 3: 40)
DDBS layer is superimposed on the client sites (4:00)
Actual Data resides with the local sites (4:35)

6. Today’s Lecture
Derived Horizontal Fragmentation

7. Derived Horizontal Fragmentation
Fragmenting/ Partitioning a table based on the constraints defined on another table.
Both tables are linked with each other through Owner-Member relation

8. Time: 5:36
a, b, c, d
p, q, r, s, a
TABLE1
TABLE2
Link
Owner
Member

9. Scenario (around 7:25) Link Owner titleId, titleName, sal Member TITLE
empId, empName, empAdres, titleId
titleId, titleName, sal
EMP
TITLE
Link
Owner
Member

10. Why DHF Here (8: 50)
Employee and salary record is split in two tables due to Normalization
Storing all data in EMP table introduces Transitive Dependency
That causes Anomalies

11. PHF of TITLE table, around 11:00
Predicates defined on the sal attribute of TITLE table
p1 = sal > and sal <= 20000
p2 = sal > and sal <= 50000
p3 = sal > 50000

12. Conditions for the TITLE Table (11:58)
TITLE1 =  (sal > and SAL ≤30000) (SAL)
TITLE2 =  (sal > and SAL ≤50000) (SAL)
TITLE3 =  (sal > 50000) (SAL)

13. Tables created with constraints: 12:20
create table TITLE1 (titleID char(3) primary key, titleName char (15), sal int check (SAL between and 20000))
create table TITLE2 (titleID char(3) primary key, titleName char (15), sal int check (SAL between and 50000))
create table TITLE3 (titleID char(3) primary key, titleName char (15), sal int check (SAL > 50000))

14. TITLE TITLE3 TITLE1 TITLE2 titleID titleName Sal T01 Elect. Eng 42000
Sys Analyst
64000
T03
Mech. Eng
27000
T04
Programmer
19000
TITLE3
TITLE1
titleID
titleName
Sal
T02
Sys Analyst
64000
titleID
titleName
Sal
T04
Programmer
19000
TITLE2
titleID
titleName
Sal
T01
Elect. Eng
42000
T03
Mech. Eng
27000

15. EMP table at local sites (14:45)
create table EMP1 (empId char(5) primary key, empName char(25), empAdres char (30), titleId char(3) foreign key references TITLE1(titleID))

16. Natural Join with Owner Fragments
Time 15:40
PHF on
Owner
titleId, titleName, sal
TITLE
Link
empId, empName, empAdres, titleId
EMP
Member
Natural Join with Owner Fragments

17. Referential Integrity Constraint (18:30)
Null value in the EMP1.titleId is allowed
This violates the correctness requirement of the Fragmentation, i.e., it will violating the completeness critetion

18. Tighten Up the Constraint Around 20:00
Further we need to impose the “NOT NULL” constraint on the EMP1.titleID
Now the records in EMP1 will strictly adhere to the DHF

19. Revised EMP1 Definition(21::00)
create table EMP1 (empId char(5) primary key, empName char(25), empAdres char (30), titleId char(3) foreign key references TITLE1(titleID) not NULL)

20. Defining all three EMP tables:21:20
create table EMP1 (empId char(5) primary key, empName char(25), empAdres char (30), titleId char(3) foreign key references TITLE1(titleID) not NULL)
create table EMP2 (empId char(5) primary key, empName char(25), empAdres char (30), titleId char(3) foreign key references TITLE2(titleID) not NULL)
create table EMP3 (empId char(5) primary key, empName char(25), empAdres char (30), titleId char(3) foreign key references TITLE3(titleID) not NULL)

21. Implementing same scenario as PHF 22:10

22. PHF of EMP at different sites (22:30)
create table EMP1 (empId char(5) primary key check (empId in ('Programmer')), empName char(25), empAdres char (30), titleId char(3))

23. create table EMP2 (empId char(5) primary key check (empId in (‘Elect
create table EMP2 (empId char(5) primary key check (empId in (‘Elect. Engr’,’Mech. Engr’)), empName char(25), empAdres char (30), titleId char(3))
create table EMP3 (empId char(5) primary key check (empId in (' Sys Analyst ')), empName char(25), empAdres char (30), titleId char(3))

24. Adding a new record in TITLE
(23:10)
TITLE
TITLE
titleID
titleName
Sal
T01
Elect. Eng
42000
T02
Sys Analyst
64000
T03
Mech. Eng
27000
T04
Programmer
19000
titleID
titleName
Sal
T01
Elect. Eng
42000
T02
Sys Analyst
64000
T03
Mech. Eng
27000
T04
Programmer
19000
T05
Assist Supr
16000
New Record

25. Time 24:00
All three predicates of PHF defined in the two slide a couple of slides ago

26. May be: 24:40
create table EMP1 (empId char(5) primary key check (empId in ('Programmer‘, ‘Assist Supr’)), empName char(25), empAdres char (30), titleId char(3))

27. Original EMP Table EMP empId empName empAdres titleId E1 T Khan Multan
W Shah
Islamabad
T02 E3
R Dar
T03 E4
K Muhammad
Lahore
T04 E5
F Sahbai E6
A Haq E7
S Farhana E8
M Daud
Jhelum
EMP

28. DHFs of EMP Table

29. EMP1 EMP3 EMP2 empId empName empAdres titleId E4 K Muhammad Lahore T04
W Shah
Islamabad
T02 E5
F Sahbai
Lahore E8
M Daud
Jhelum
EMP2
empId
empName
empAdres
titleId E1
T Khan
Multan
T01 E3
R Dar
Islamabad
T03 E6
A Haq E7
S Farhana
Lahore

30. Transactional Replication (30:30)

31. Data replicated as the transaction executes
Preferred in higher bandwidth and lower latency
Transaction is replicated as it is executed

32. Begins with a snapshot, changes sent at subscribers as they occur or at timed intervals
A special Type allows changes at subscribers

33. Lets do it
From the Enterprise Manager, select Replication, after a couple of nexts, we get this screen

40. Publication has been created that can be viewed from Replication Monitor or from Replication, like this

42. It has also created snapshot and log reader agents, which won’t work until we create a subscription.
For this, we select the publication from replication monitor, right click it, then select Push new subscription

45. You select the particular database where you want to subscribe, we have created a new one

47. On next slide is asks the schedule for Distributor

48. A couple of more nexts and then

49. After this we run the snapshot agent, that creates a snapshot, you can verify this from snapshot agent history, or you can go to subscriber database and have a look, like this

51. We delete a record from our publication and we see that it is expressed in

53. This will be automatically be transferred to Subscription.
If this activity could not be performed on subscriber, then the replication monitor will generate an error. You have to trap it and tune your application

54. Merge Replication

55. From replication, start a new publication, a few nexts, and once again our old familiar screen

58. After a few nexts, the merge publication will be created

60. Now you execute the snapshot agent of this replication
It will create the snapshot, then you subscribe a database
We find the original data in the subscriber
After this if we make any change on either side, it will be reflected on the other side

61. In case of merge replication, we have to be careful about the constraints, like Primary Key, or other constraints