Mobile and Heterogeneous databases

Size: px

Start display at page:

Download "Mobile and Heterogeneous databases"

Blaze Howard Knight
5 years ago
Views:

1 Mobile and Heterogeneous databases Heterogeneous Distributed Databases Transaction Processing A.R. Hurson Computer Science Missouri Science & Technology 1

2 Note, this unit will be covered in two lectures. In case you finish it earlier, then you have the following options: 1) Take the early test and start CS6302.module9 2) Study the supplement module (supplement CS6302.module8) 3) Act as a helper to help other students in studying CS6302.module8 Note, options 2 and 3 have extra credits as noted in course outline. 2

3 Enforcement of background Current Module { Glossary of prerequisite topics No Familiar with the topics? Yes Take Test No Pass? Remedial action Yes { Glossary of topics No Familiar with the topics? Take the Module Yes Take Test No Pass? Yes Options Study next module? Review CS6302 module8background Lead a group of students in this module (extra credits)? Study more advanced related topics (extra credits)? At the end: take exam, record the score, impose remedial action if not successful Extra Curricular activities 3

4 You are expected to be familiar with: Heterogeneous Distributed Databases, If not, you need to study CS6302.module3, and CS6302.module7 4

5 Access and manipulate data across multiple heterogeneous pre-existing data sources in sometimes, somewhere and anytime anywhere, access environment: Multidatabase System (MDBS) Mobile DataAccess System (MDAS) Need an efficient global transaction management scheme with an eye on issues such as: Heterogeneity, autonomy, network bandwidth, frequent disconnections, and limited processing power and resources. 5

6 The Summary Schemas Model has been developed as an extension to multidatabase systems to provide linguistic support to automatically identify semantically similar entities with different access terms. 6

7 Summary Schemas Model Advantages The meta-data is by orders of magnitude smaller than the meta-data generated by the Global-schema approach. Preserves local DBMS autonomy. Provides good system scalability. Reduces average search time. Resolves imprecise queries. 7

8 Summary Schemas Model Prototyped a Client/Server based SSM Access Control and Security in SSM Transaction management in SSM Query processing in SSM 8

9 Summary Schemas Model Lessons learned Lack of portability, Lack of stability, Relying on network connectivity. 9

10 Summary Schemas Model Current distributed and multidatabase systems are designed to allow timely and reliable access to large amounts of autonomous and heterogeneous data sources Sometimes, somewhere access environment. 10

11 Summary Schemas Model Adding mobility and wireless connection to the traditional multidatabase environments allow anytime, anywhere access to the information sources. However, this advantage comes at the expense of additional complexities due to the network bandwidth, frequent disconnections, and limited processing power and resources. 11

12 Extended Summary Schemas Model Add mobility to the Summary Schemas Model (MDAS) Transaction management in MDAS Power management Application of Software Agent Multimedia support Community of local ad-hoc nodes/sensor networks 12

13 It was decided to use agent technology to overcome technological constraints. Mobile agent-based computation paradigm responds to the: Limited resources of mobile devices by migrating tasks to more powerful servers on the network, Intermittent connectivity by supporting disconnected operations, and Reduced Bandwidth by moving computation to the data. 13

subtransaction LDBS Local Transaction Manager LDBS Local Transaction Manager LDBS Local

14 Local Transaction Global Transaction Global subtransaction Global Transaction Global Mobile Multidatabase Management System Global Transaction Manager Global Transaction Global subtransaction LDBS Local Transaction Manager LDBS Local Transaction Manager LDBS Local Transaction Manager Local databases residing on wired network receiving transactions from both fixed and mobile clients. 14

15 Autonomy and heterogeneity Direct and Indirect Conflicts Intermittent and unreliable network connectivity Excessive transaction aborts Long-lived Transactions (LLT) Overcommitted resources Power and processing constraints Lower bandwidth 15

16 Efficient access and manipulation of data across multiple heterogeneous and autonomous databases via wireless connections should: Preserve ACID properties. Preserve autonomy of the local databases as much as possible. Accommodate frequent disconnections. 16

17 Non-blocking scheme at the global level. Parallel processing of global sub-transactions. Reduced Communication (Agents can make local decision, avoiding acknowledgements). No cascading Abort. Mobility Support. 17

18 Summary Schema Model (SSM) Global Transaction Agent Global Sub-Transaction Agent NodeManager Global Order Table 18

SSN Level 1 SSN Level 2 1.A NodeManager Global order GT1 table 2.A NodeManager Global order table 2.B NodeManager Global order table 2.C NodeManager Global order table 3.

19 SSN Level 1 SSN Level 2 1.A NodeManager Global order GT1 table 2.A NodeManager Global order table 2.B NodeManager Global order table 2.C NodeManager Global order table 3.A NodeManager Global order table 3.B NodeManager Global order table 3.C NodeManager Global order table 3.D NodeManager Global order table LDBS TM LDBS TM LDBS TM LDBS TM LDB1 LDB2 LDB3 LDB4

20 Phase1: Transaction Resolution and selection of the Global Transaction Coordinator Phase2: Global Order AT3M uses pessimistic approach. Collaboration between NodeManagers and GSTAgent to perform a timestamp-based algorithm and determine a global execution order before the execution of the global sub-transactions. Global order is enforced through out the hierarchy. 20

C [GT1] Global Order Table [GT1, GT2] [GT1] [GT1] GT2LDB3AG GT1LDB2AG GT1LDB3AG [GT1] [GT1, GT2] GT1LDB4AG

21 [GT1] 1.A Global Order Table GT1AG GT1 [GT1] GT1LDB4AG GT2 2.A Global Order Table GT2AG [GT1, [GT1] GT2] 2.B Global Order Table [GT1] GT1LDB2AG GT1LDB3AG 2.C [GT1] Global Order Table [GT1, GT2] [GT1] [GT1] GT2LDB3AG GT1LDB2AG GT1LDB3AG [GT1] [GT1, GT2] GT1LDB4AG GT2LDB2AG [GT1, [GT1] GT2] [GT1, GT2] [GT1] 3.A 3.B 3.C 3.D Global Order Table Global Order Table Global Order Table Global Order Table LDBS LDBS LDBS LDBS LTM LTM LTM LTM LDB1 LDB2 LDB3 LDB4 21

22 Phase3: Global Order Enforcement NodeManagers at leaf node preserve global serialization. Phase4: Global Commit Each GSTAgent after execution enters in the prepare-to-commit stage and reports back to the GTAgent. The GTAgent after receiving all responses initiates global commit command. The GSTAgent receiving global commit command commits the global subtransaction and terminates. GTAgent receiving any abortrequest initiates global abort. 22

23 A simulator was developed using SimJava 2.0. The mobile multidatabase system consists of several local databases. Each local database contains 100 data items, 20 items are hot-spot (likely to be accessed). AT3M is simulated and compared against V- Locking and Pre-Serialization algorithms. 23

24 Average Processing Time(s) Average Processing Time VS. Probability of Read-Only Transactions 24

25 Message/Transaction Number of Communication Messages per Transaction VS. Probability of Read-Only Transactions 25

26 Processing Time(s) Average Processing Time for Different Levels of Local And Global Transactions Ratio 26

27 8 7 Throughput (T/s.) AT3M's GT V-Locking's GT AT3M's LT V-Locking's LT LT:GT Ratio Throughput for Different Levels of Local And Global Transactions Ratio 27

28 45 Average response time (s.) :10 5:10 10:10 20:10 30:10 High priority clients: Low priority clients High priority client Low priority client 28

29 An agent-based pessimistic non-locking approach that preserves the local autonomy and technological constraints. It is a promising approach as it provides better average processing time and lower number of communication messages. 29

30 Extension of AT3M to operate in pervasive computing environment. Extension of AT3M to operate in environment with replicated databases. 30

Mobile and Heterogeneous databases

Mobile and Heterogeneous databases Heterogeneous Distributed Databases Query Processing A.R. Hurson Computer Science Missouri Science & Technology 1 Note, this unit will be covered in two lectures. In