transaction ,concurrency , and DDBMS
- 1. Name Zahid Hussain Roll no. 16717 Subject DBMS Topic Concurrency & Distributed DBMS
- 2. Overview 1. Transaction 2. Concurrency 3. Resource locking 4. Deadlock 5. Types of Distributed DBMS
- 3. 1.1 Transaction logical unit of work it consist of one or more operations that must be completed together.It either be successfully complete or fail. 1.1.1 Commit A transaction is committed if it is completely successful and makes changes in data. 1.1.2 Rollback if transaction failed and leave the data unchanged it is said the data has rolled back. 1.2 Transaction properties(ACID properties) 1.2.1 Atomicity A transaction must completely succeed or completely fail. If any statement fail the entire transaction fails completely. 1.2.2 Consistency After completion the transaction leaves the data in a consistent state. For example money should never be “created “ or “deleted” without an deposit or withdrawal.
- 4. 1.2.3 Isolation All transactions that modify the data are isolated from each other. A modifying transaction can access the data only after the completion of other transaction 1.2.4 Durability Durability means that the modifications made by a transaction are permanent and persistent.
- 5. 1.3 Concurrency Concurrency is a situation in which two users access the same piece of data at the same time. It occurs very commonly but sometimes create some problems. 1.3.1 Lost update problem Problem arises when two or more transactions update the same data. Each transaction unaware of other transactions . The last update overwrites. TIME TEAM A STOCK TABLE TEAM B 9:00 Qty=100 10:30 Retrieve Qty Qty=100 Retrieve Qty Qty=100 10:31 Update=Qty-90 10:32 Qty=10 10:33 Update=Qty+30 Qty=130
- 6. 1.3.2. Uncommitted Dependency Problem In above example team B update the quantity 100 to 150 at time T1 and team A retrieve this update value at t2.Team B rollbacks the action while making quantity 100 again .So, the value of Team A becomes wrong at T3. TIME TEAM A TEAM B T1 - Update qty T2 Retrieve qty - T3 - Rollback
- 7. 1.3.3 Inconsistent Analysis Problems problem occurs when one transaction reading the several values from the DB but another transaction update these some of them during the executing of the first transaction. TIME TEAM A TEAM B T1 Retrieve acc-1:Sum=40 - T2 Retrieve acc-2:Sum=90 - T3 Retrieve acc-3 T4 Update acc-3=0 T5 Retrieve acc-1 T6 Update acc-1=50 T7 Commit T8 Retrieve acc-3:Sum=90 Three accounts Acc-1:40 Acc-2: 50 Acc-3: 30
- 8. 1.4 Resource locking When a user access the data the second user has to wait until the first user has finishes his work. For example USER 1 USER 2 1 Lock the item 2 retrieve the item 3 reduce item by 10 4 update the item 5 unlock the item 1 Lock the item 2 retrieve the item 3 reduce item by 10 4 update the item 5 unlock the item Sequence of processing commands by CPU 1 Lock the item user 1 2 retrieve the item user 1 3 lock the item for user 2. already locked so user 2 has to wait until user 1 unlock 4 reduce item by 10 for user 1 5 update the item for user 1 6 unlock the item for user 1 7 lock the item for user 2 8 retrieve item for user 2 9 reduce item by 20 for user 2 10 update the item for user 2 11 Unlock the item for user 2
- 9. 1.4.1 Lock terminology 1. Implicit lock Type of lock placed automatically by DBMS. 2. Explicit lock Placed by application program by issuing a command. 3. Exclusive lock Type of lock that locks item for any type of access no other can read or update. 4. Shared lock Locks an item from change but not from read is called share lock. 5. Lock Granularity The level of lock applied on an item is known as lock granularity Levels a. Row level(smallest lock granularity) b. Page level c. Table level d. Database level(largest lock granularity)
- 10. 5. Deadlock Deadlock is a situation in which two transaction wait for completion of each other. USER 1 USER 2 1 Lock item A 2 Retrieve item A 3 Lock the item B 1 Lock item B 2 Retrieve item B 3 Lock the item A Sequence processing by command in CPU 1 Lock the item A for user 1. 2 Lock the item B for user 2. 3 Retrieve A for user 1. 4 Retrieve item B for user 2. 5 Put user 1 in wait for user 2. 6 Put user 2 in wait for user 1
- 11. 5.1 Solution of deadlock 1 Deadlock prevention In this type users allow to issue only one lock request at a time.If user A and B at once in the first line deadlock will not occurs. 2 Deadlock detection & recovery If one transaction of user 2 is aborted then the item will be released from the lock and the user 1 will be able to complete his work. When a transaction is aborted then all changes made by that transaction should be undone.
- 12. 6. Types of Distributed DBMS 6.1 Homogeneous DDBMS Means that the DB technology is the same or at least compatible to each . It simplify the data sharing among different users. Characteristics of Homogeneous DDBMS i. Data distributed across all the nodes. ii. The same DBMS used. iii. All date manage by the DDBMS there is no exclusively local data. iv. All users access the DB through database definition. v. The global schema is the union of local DB schemas.
- 13. 6.2 Heterogeneous DDBMS Different DBMS used at each node. Much more difficult to manage. Characteristics of Heterogeneous DDBMS i. Data distributed across all the nodes. ii. Different DBMS used at each location. iii. A global schema exist which allow local user to access the data iv. Some users require only local access to databases which cab be accomplished using only local data DBMS and schema.
- 14. Any Question?
- 15. Thank you