Transparent Management of Distributed Data

Distributed Database Systems Overview

• A distributed database system is a collection of multiple, logically interrelated databases distributed over a computer network.
• A distributed DBMS (DDBMS) is the software system that permits the management of the distributed database and makes the distribution transparent to the users.

Characteristics of Distributed Database Systems

• Data management at multiple sites: data is stored and managed at geographically multiple sites.
• Local requirements: each site storing data in a DDBS is called a local site, catering to local users.
• Global perspective: the DDBS fulfills global requirements in a transparent way.

Access to a Distributed Database System

• Three types of accesses:
  • Local access: access by users connected to a site, accessing data from the same site.
  • Remote access: a user connected to a site, accessing data from another site.
  • Global access: data is displayed after being collected from all locations, regardless of access location.

Application of Distributed Databases

• Candidates for a DDBS have two main characteristics:
  • Large number of users.
  • Users are physically spread across a large geographical area.
• Examples of candidates:
  • Banking applications.
  • Air ticketing.
  • Business at multiple locations.

Data Delivery Alternatives

• Three orthogonal dimensions/alternatives of DDBS exist:
  • Delivery modes (pull-only, push-only, hybrid).
  • Frequency (periodic, conditional, ad-hoc).
  • Communication methods (unicast, multicast).

Communication Methods

• Unicast: one-to-one communication from a server to a client.
• Multicast: one-to-many communication from a server to multiple clients.

Distributed Database System vs Centralized Client-Server System

• Centralized C/S System:
  • Data management is carried out on a single centralized system.
  • Data is accessed from different machines (clients) connected through a network.
  • Data storage and management is mainly done on the server.

Promises of Distributed Database Systems

• Four fundamental advantages:
  • Transparent management of distributed and replicated data.
  • Reliable access to data through distributed transactions.
  • Improved performance.
  • Easier system expansion.

Transparency in Distributed Database Systems

• Types of transparencies:
  • Data independence.
  • Network transparency or distribution transparency.
  • Replication transparency.
  • Fragmentation transparency.
• Responsibility of transparency:
  • Access layer (language/compiler).
  • Operating system level.
  • DBMS level.

Layers of Transparency

• Language/compiler provides transparency features.
• Operating system provides network transparency.
• DBMS provides fragmentation or replication transparency.### Distributed Database Systems (DDBS)

Critical Issues in DDBS

• Consistency of data across multiple sites is a serious issue
• Distributed transactions should be executed at multiple sites without fear of failure
• Reliability through distributed transactions ensures user applications do not need to worry about coordinating accesses to individual local databases or site/communication link failures

Advantages of DDBS

Improved Performance

• Data Localization: data is stored close to its point of use, reducing remote access delays and contention for CPU and I/O services
• Query Parallelism: a query can be executed in parallel, improving performance

Easier System Expansion

• Expansion can be handled by adding processing and storage power to the network
• More economical than establishing a single, centralized system

Complications Introduced by Distribution

• Selection of the Copy: choosing the right copy based on constraining variables (distance, load, site availability, etc.)
• Failure Recovery: synchronization of copies after failure is more complex than in centralized systems
• Complexity: overall system is more complex than centralized database systems
• Cost: involves more cost, as hardware and trained manpower are needed at multiple sites
• Distribution of Control: access to data must be carefully managed, with well-defined rights for local sites

Design Issues in DDBS

• Distributed Database Design: how to place the database and applications across sites
• Distributed Directory Management: global or local directories, centralized or distributed
• Distributed Query Processing: deciding on a strategy for executing queries over the network
• Distributed Concurrency Control: maintaining consistency of multiple copies of the database
• Distributed Deadlock Management: prevention, avoidance, and detection/recovery
• Reliability of Distributed DBMS: ensuring consistency and detecting/recovering from failures
• Replication: fully or partially replicating the database
• Relationship among Problems: design issues affect each other

Distributed Concurrency Control

• Ensuring integrity of the database despite concurrent access
• Two classes: pessimistic (synchronizing before execution) and optimistic (executing and checking consistency)
• Two fundamental primitives: Locking and Timestamping

Distributed Deadlock Management

• Alternatives: prevention, avoidance, and detection/recovery

Reliability of Distributed DBMS

• Ensuring consistency of the database despite failures
• Recovering from failures and bringing databases up-to-date

Replication

• Ensuring consistency of replicas (copies of the same data item)
• Protocols: Eager (forcing updates to all replicas before completion) or Lazy (propagating updates after completion)

Relationship among Problems

• Design of distributed databases affects many areas
• Directory management, query processing, concurrency control, and replication are interrelated