1 / 67

Distributed Data Management

Distributed Data Management. Distributed Systems Department of Computer Science UC Irvine. Application SQL Front End Query Processor Transaction Proc. File Access. Centralized DB systems. P. Software:. M. Simplifications: single front end one place to keep data, locks

efrancisco
Download Presentation

Distributed Data Management

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Distributed Data Management Distributed Systems Department of Computer Science UC Irvine

  2. Application SQL Front End Query Processor Transaction Proc. File Access Centralized DB systems P Software: M ... • Simplifications: • single front end • one place to keep data, locks • if processor fails, system fails, ...

  3. Distributed Database Systems • Multiple processors ( + memories) • Heterogeneity and autonomy of “components”

  4. Why do we need Distributed Databases? • Example: IBM has offices in London, New York, and Hong Kong. • Employee data: • EMP(ENO, NAME, TITLE, SALARY, …) • Where should the employee data table reside?

  5. IBM Data Access Pattern • Mostly, employee data is managed at the office where the employee works • E.g., payroll, benefits, hire and fire • Periodically, IBM needs consolidated access to employee data • E.g., IBM changes benefit plans and that affects all employees. • E.g., Annual bonus depends on global net profit.

  6. Internet New York Payroll app London Payroll app EMP London New York Hong Kong Payroll app Problem: NY and HK payroll apps run very slowly! Hong Kong

  7. Internet New York Payroll app London Payroll app London Emp NY Emp London New York Hong Kong Payroll app Much better!! Hong Kong HK Emp

  8. Internet New York Payroll app London Payroll app Annual Bonus app London Emp NY Emp London New York Hong Kong Payroll app Distribution provides opportunities for parallel execution Hong Kong HK Emp

  9. Internet New York Payroll app London Payroll app Annual Bonus app London Emp NY Emp London New York Hong Kong Payroll app Hong Kong HK Emp

  10. Internet New York Payroll app London Payroll app Annual Bonus app Lon, NY Emp NY, HK Emp London New York Hong Kong Payroll app Replication improves availability Hong Kong HK, Lon Emp

  11. Homogeneous Distributed Databases • In a homogeneous distributed database • All sites have identical software • Are aware of each other and agree to cooperate in processing user requests. • Each site surrenders part of its autonomy in terms of right to change schemas or software • Appears to user as a single system • In a heterogeneous distributed database • Different sites may use different schemas and software • Difference in schema is a major problem for query processing • Difference in software is a major problem for transaction processing • Sites may not be aware of each other and may provide only limited facilities for cooperation in transaction processing

  12. DB architectures (1) Shared memory P P P ... M

  13. P P P M DB architectures (2) Shared disk ... M M ...

  14. P P P M M M DB architectures (3) Shared nothing ...

  15. ... ... P P P P P P M M DB architectures (4) Hybrid example – Hierarchical or Clustered

  16. Issues for selecting architecture • Reliability • Scalability • Geographic distribution of data • Performance • Cost

  17. Parallel or distributed DB system? • More similarities than differences!

  18. Typically, parallel DBs: • Fast interconnect • Homogeneous software • High performance is goal • Transparency is goal

  19. Typically, distributed DBs: • Geographically distributed • Data sharing is goal (may run into heterogeneity, autonomy) • Disconnected operation possible

  20. Distributed Database Challenges • Distributed Database Design • Deciding what data goes where • Depends on data access patterns of major applications • Two subproblems: • Fragmentation: partition tables into fragments • Allocation: allocate fragments to nodes

  21. Assume relational data model Replication System maintains multiple copies of data, stored in different sites, for faster retrieval and fault tolerance. Fragmentation Relation is partitioned into several fragments stored in distinct sites Replication and fragmentation can be combined Relation is partitioned into several fragments: system maintains several identical replicas of each such fragment. Distributed Data Storage

  22. A relation or fragment of a relation is replicated if it is stored redundantly in two or more sites. Full replication of a relation is the case where the relation is stored at all sites. Fully redundant databases are those in which every site contains a copy of the entire database. Data Replication

  23. Advantages of Replication Availability: failure of site containing relation r does not result in unavailability of r is replicas exist. Parallelism: queries on r may be processed by several nodes in parallel. Reduced data transfer: relation r is available locally at each site containing a replica of r. Disadvantages of Replication Increased cost of updates: each replica of relation r must be updated. Increased complexity of concurrency control: concurrent updates to distinct replicas may lead to inconsistent data unless special concurrency control mechanisms are implemented. One solution: choose one copy as primary copy and apply concurrency control operations on primary copy Data Replication (Cont.)

  24. Division of relation r into fragments r1, r2, …, rnwhich contain sufficient information to reconstruct relation r. Horizontal fragmentation: each tuple of r is assigned to one or more fragments Vertical fragmentation: the schema for relation r is split into several smaller schemas All schemas must contain a common candidate key (or superkey) to ensure lossless join property. A special attribute, the tuple-id attribute may be added to each schema to serve as a candidate key. Example : relation account with following schema Account = (branch_name, account_number, balance ) Data Fragmentation

  25. Horizontal Fragmentation of account Relation account_number branch_name balance Hillside Hillside Hillside A-305 A-226 A-155 500 336 62 account1 =branch_name=“Hillside” (account ) account_number branch_name balance Valleyview Valleyview Valleyview Valleyview A-177 A-402 A-408 A-639 205 10000 1123 750 account2 =branch_name=“Valleyview” (account )

  26. Vertical Fragmentation of employee_info Relation tuple_id branch_name customer_name Lowman Camp Camp Kahn Kahn Kahn Green 1 2 3 4 5 6 7 Hillside Hillside Valleyview Valleyview Hillside Valleyview Valleyview deposit1 =branch_name, customer_name, tuple_id (employee_info ) account_number tuple_id balance 500 336 205 10000 62 1123 750 A-305 A-226 A-177 A-402 A-155 A-408 A-639 1 2 3 4 5 6 7 deposit2 =account_number, balance, tuple_id (employee_info )

  27. Horizontal: allows parallel processing on fragments of a relation allows a relation to be split so that tuples are located where they are most frequently accessed Vertical: allows tuples to be split so that each part of the tuple is stored where it is most frequently accessed tuple-id attribute allows efficient joining of vertical fragments allows parallel processing on a relation Vertical and horizontal fragmentation can be mixed. Fragments may be successively fragmented to an arbitrary depth. Advantages of Fragmentation

  28. Data transparency: Degree to which system user may remain unaware of the details of how and where the data items are stored in a distributed system Consider transparency issues in relation to: Fragmentation transparency Replication transparency Location transparency Data Transparency

  29. 1. Every data item must have a system-wide unique name. 2. It should be possible to find the location of data items efficiently. 3. It should be possible to change the location of data items transparently. 4. Each site should be able to create new data items autonomously. Naming of Data Items - Criteria

  30. Structure: name server assigns all names each site maintains a record of local data items sites ask name server to locate non-local data items Advantages: satisfies naming criteria 1-3 Disadvantages: does not satisfy naming criterion 4 name server is a potential performance bottleneck name server is a single point of failure Centralized Scheme - Name Server

  31. Alternative to centralized scheme: each site prefixes its own site identifier to any name that it generates i.e., site 17.account. Fulfills having a unique identifier, and avoids problems associated with central control. However, fails to achieve network transparency. Use of Aliases

  32. What is a Transaction? • A set of steps completed by a DBMS to accomplish a single user task. • Must be either entirely completed or aborted • No intermediate states are acceptable

  33. Transaction may access data at several sites. Each site has a local transaction manager responsible for: Maintaining a log for recovery purposes Participating in coordinating the concurrent execution of the transactions executing at that site. Each site has a transaction coordinator, which is responsible for: Starting the execution of transactions that originate at the site. Distributing subtransactions at appropriate sites for execution. Coordinating the termination of each transaction that originates at the site, which may result in the transaction being committed at all sites or aborted at all sites. Distributed Transactions

  34. Transaction System Architecture

  35. Failures unique to distributed systems: Failure of a site. Loss of massages Handled by network transmission control protocols such as TCP-IP Failure of a communication link Handled by network protocols, by routing messages via alternative links Network partition A network is said to be partitionedwhen it has been split into two or more subsystems that lack any connection between them Note: a subsystem may consist of a single node Network partitioning and site failures are generally indistinguishable. System Failure Modes

  36. Distributed commit problem . Transaction T Action: a1,a2 Action: a3 Action: a4,a5 Commit must be atomic

  37. Example of a Distributed Transaction How a distributed transaction that has components at several sites can execute atomically? T1 Store 2 should ship 500 toothbrushes to store 5 FAILs N := N - 500 T2 T0 Headquarters N := N + 500 T5 Tn Stores Each Ti is executed atomically, but T0 itself is not atomic.

  38. Distributed commit problem • Commit must be atomic • Solution: Two-phase commit (2PC) • Centralized 2PC • Distributed 2PC • Linear 2PC • Many other variants…

  39. TWO-PHASE COMMIT (2PC) - OK

  40. TWO-PHASE COMMIT (2PC) - ABORT ‘Global Abort’

  41. Terminology • Resource Managers (RMs) • Usually databases • Participants • RMs that did work on behalf of transaction • Coordinator • Component that runs two-phase commit on behalf of transaction

  42. REQUEST-TO-PREPARE PREPARED* COMMIT* Participant DONE Coordinator

  43. REQUEST-TO-PREPARE NO ABORT Participant DONE Coordinator

  44. Centralized two-phase commit Coordinator Participant I I request-prepare no request-prepare prepared commit-request request-prepare* no abort* W A W A abort done prepared* Commit* commit done done* C F C done*

  45. Notation: Incoming message Outgoing message ( * = everyone) • When participant enters “W” state: • it must have acquired all resources • it can only abort or commit if so instructed by a coordinator • Coordinator only enters “C” state if all participants are in “W”, i.e., it is certain that all will eventually commit

  46. After coordinator receives DONE message, it can forget about the transaction • E.g., cleanup control structures

  47. Next • Add timeouts to cope with messages lost during crashes

  48. any abort _t_ abort _t_ abort* any commit _t_ commit I commit-request request-prepare* no abort* Coordinator W A prepared* commit* done* - C F done* - t=timeout

  49. abort done _t_ ping equivalent to finish state commit done request-prepare no I Participant request-prepare prepared abort done W A commit done C

  50. For centralized systems, the primary criterion for measuring the cost of a particular strategy is the number of disk accesses. In a distributed system, other issues must be taken into account: The cost of a data transmission over the network. The potential gain in performance from having several sites process parts of the query in parallel. Distributed Query Processing

More Related