Distributed Databases Basic Concepts

Concepts. Advantages and disadvantages of distributed databases. Functions and architecture for a DDBMS. Distributed database design. Levels of transparency. Comparison criteria for Distributed DBMSs.

Distributed Databases

Connolly & Begg. Chapter 22. Third edition

COMP 302 Valentina Tamma COMP 302 Valentina Tamma

Why distributed databases?

Some initial motivations:
The development of computer networks promotes decentralization. In a company, the database organization might reflect the organizational structure, which is distributed into units. Each unit maintains its own database. Sharing of data can be achieved by developing a distributed database system which: makes data accessible by all units stores data close to where it is most frequently used.

Concepts
Distributed Database
A logically interrelated collection of shared data (and a description of this data), physically distributed over a computer network.

Distributed DBMS (DDBMS)

Software system that permits the management of the distributed database and makes the distribution transparent to users.

COMP 302

Valentina Tamma

COMP 302

3 Valentina Tamma

An example of DDBMS

DDBMS - characteristics
Collection of logically-related shared data. Data split into fragments. Fragments may be replicated. Fragments/replicas allocated to sites. Sites linked by a communications network. Data at each site is under control of a DBMS. DBMSs handle local applications autonomously. Each DBMS participates in at least one global application.
4 Valentina Tamma

COMP 302

Valentina Tamma

COMP 302

These are not DDBMSs

Distributed Processing A centralized database that can be accessed over a computer network.

These are not DDBMSs

Parallel DBMS A DBMS running across multiple processors and disks designed to execute operations in parallel, whenever possible, to improve performance.
Based on premise that single processor systems can no longer meet requirements for cost-effective scalability, reliability, and performance. Parallel DBMSs link multiple, smaller machines to achieve same throughput as single, larger machine, with greater scalability and reliability.

COMP 302

6 Valentina Tamma

COMP 302

7 Valentina Tamma

Parallel DBMS
Main architectures for parallel DBMSs are:
Shared memory, Shared disk, Shared nothing.

Parallel DBMS

(a) shared memory (b) shared disk (c) shared nothing

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Advantages of DDBMSs
Reflects Organizational Structure Improved Sharing and Local Autonomy Improved Availability A failure does not make the entire system inoperable Improved Reliability Data may be replicated Improved Performance Data are local to the site of greatest demand Economics Many small computers cost less than a big one! Modular Growth easy to add new modules

Disadvantages of DDBMSs
Complexity Cost Especially in system management Security network must be made secure Integrity Control More Difficult Lack of Standards Lack of Experience Database Design More Complex due to fragmentation, allocation of fragments to a specific site, ..
COMP 302
11 Valentina Tamma

COMP 302

10 Valentina Tamma

Types of DDBMS
Homogeneous DDBMS All sites use same DBMS product (eg.Oracle) Fairly easy to design and manage. Heterogeneous DDBMS Sites may run different DBMS products (eg. Oracle and Ingress) Possibly different underlying data models (eg. relational DB and OO database) Occurs when sites have implemented their own databases and integration is considered later. We wont consider heterogeneous DDBMSs here.
COMP 302

Multidatabase System (MDBS)

DDBMS in which each site maintains complete autonomy. DBMS that resides transparently on top of existing database and file systems and presents a single database to its users. Allows users to access and share data without requiring physical database integration. Unfederated MDBS (no local users) and federated MDBS.

12 Valentina Tamma

COMP 302

Valentina Tamma

Overview of Networking
Network - Interconnected collection of autonomous computers, capable of exchanging information.
Local Area Network (LAN) intended for connecting computers at same site. Wide Area Network (WAN) used when computers or LANs need to be connected over long distances. WAN relatively slow and less reliable than LANs. DDBMS using LAN provides much faster response time than one using WAN.

Overview of Networking

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Functions of a DDBMS
Expect DDBMS to have at least the functionality of a DBMS (see Connolly & Begg. Chapter 2. Third edition) Also to have following functionality:

Reference Architecture for DDBMS

Due to diversity, no accepted architecture equivalent to ANSI/SPARC 3-level architecture for DBMSs. A possible reference architecture consists of:

Extended communication services. Extended Data Dictionary. Distributed query processing. Extended concurrency control. Extended recovery services. Extended security control.

Set of global external schemas. Global conceptual schema (GCS). Fragmentation schema and allocation schema. Set of schemas for each local DBMS conforming to 3-level ANSI/SPARC .

Some levels may be missing, depending on levels of transparency supported.

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Reference Architecture for DDBMS

Reference Architecture for DDBMS

Global Conceptual Schema is the logical description of the DB as if it were not distributed. It contains definitions of entities, relationships, constraints, security, and integrity information. Fragmentation and Allocation Schemas describe how data are logically partitioned, and where they are located, taking replication into account. Local Schemas are the logical descriptions of the local DBs.

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Components of a DDBMS

Distributed Databases
Issues in Distributed Database Design

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Issues in Distributed Database Design

Three key issues we have to consider:
Data Allocation: where are data placed? Data should be stored at site with "optimal" distribution. Fragmentation: relation may be divided into a number of sub-relations (called fragments) , which are stored in different sites. Replication: copy of fragment may be maintained at several sites.

Issues in Distributed Database Design

Definition and allocation of fragments carried out strategically to achieve:

Locality of Reference Improved Reliability and Availability Improved Performance Balanced Storage Capacities and Costs Minimal Communication Costs.

Involves analysing most important transactions, based on quantitative/qualitative information.

COMP 302

27 Valentina Tamma

COMP 302

Valentina Tamma

Fragmentation
Quantitative information may include:

Data Allocation
Four strategies regarding placement of data: Centralized Partitioned (or Fragmented) Complete Replication Selective Replication

frequency with which a transaction is run; site from which a transaction is run; performance criteria for transactions.

Qualitative information may include transactions that are executed such as:
type of access (read or write); predicates of read operations.

COMP 302

Valentina Tamma

COMP 302

30 Valentina Tamma

Data Allocation
Centralized: Consists of single database stored at one site with users distributed across the network. (This is not a DDB but distributed processing!!) Partitioned: Database partitioned into disjoint fragments, each fragment assigned to one site. Complete Replication: Consists of maintaining complete copy of database at each site. Selective Replication:Combination of partitioning, replication, and centralization.

Fragmentation
A relation R is divided into fragments r1, r2, rn, which contain enough information to allow reconstruction of R
Example: We have a relation Sells(pub, address,price,type) Type is bitter or lager. We can split Sells into twp dfferent fragments: SellsBitter= type = bitter(Sells) SellsLager= type = lager(Sells)
COMP 302
28 Valentina Tamma

COMP 302

31 Valentina Tamma

Comparison of Strategies for Data Distribution

Usage

Why Fragment?
Applications work with views rather than entire relations. Data is stored close to where it is most frequently used. Data that is not needed by local applications is not stored With fragments as unit of distribution, transaction can be divided into several sub-queries that operate on fragments. Data not required by local applications is not stored and so not available to unauthorized users.

Efficiency

Parallelism

Security

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Why Fragment?

Types of Fragmentation
Four types of fragmentation:
Horizontal, Vertical, Mixed, Derived.

Two main disadvantages:

Performance Integrity.

Other possibility is no fragmentation:

If relation is small and not updated frequently, may be better not to fragment relation.

COMP 302

Valentina Tamma

COMP 302

Pearson Education Limited 1995, 2005

Valentina Tamma

Horizontal and Vertical Fragmentation

Two types of fragmentation: Horizontal

Mixed Fragmentation

Vertical

COMP 302

Valentina Tamma

COMP 302

Pearson Education Limited 1995, 2005

Valentina Tamma

Horizontal Fragmentation
Each fragment consists of a subset of the tuples of a relation R.
Defined using Selection operation of relational algebra: p(R) Example:
Relation: Sells(pub, address,price,type) Fragments:

Horizontal Fragmentation
This strategy is determined by looking at predicates used by transactions. Involves finding set of minimal (complete and relevant) predicates. Set of predicates is complete, if and only if, any two tuples in same fragment are referenced with same probability by any application. Predicate is relevant if there is at least one application that accesses fragments differently.

SellsBitter= type = bitter(Sells) SellsLager= type = lager(Sells)

COMP 302

43 Valentina Tamma

COMP 302

Valentina Tamma

Vertical Fragmentation
Each fragment consists of a subset of attributes of a relation R. Defined using projection operation of relational algebra: a1,an(R) Determined by establishing affinity of one attribute to another. Example:
Relation: Bars(name,address,licence,employees,owner) Fragments:

Mixed Fragmentation
We can also mix horizontal and vertical fragmentation. We obtain a fragment that consist of an horizontal fragment that is vertically fragmented, or a vertical fragment that is horizontally fragmented. Defined using Selection and Projection operations of relational algebra. p( a1,an(R)) a1,an(p(R))
45 Valentina Tamma

name,address,licence (Bars) name,address,employees,owner(Bars)

COMP 302

COMP 302

46 Valentina Tamma

Example - Mixed Fragmentation

S1 = staffNo, position, sex, DOB, salary(Staff) S2 = staffNo, fName, lName, branchNo(Staff) S21 = S22 = S23 =
branchNo=B003(S2 ) branchNo=B005(S2 ) branchNo=B007(S2 )

Derived Horizontal Fragmentation

A horizontal fragment that is based on horizontal fragmentation of a parent relation. Ensures that fragments that are frequently joined together are at same site. Defined using Semijoin operation of relational algebra:
Ri = R >F Si, 1iw

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Example - Derived Horizontal Fragmentation

S3 = S4 = S5 =
branchNo=B003(Staff) branchNo=B005(Staff) branchNo=B007(Staff)

Derived Horizontal Fragmentation

If relation contains more than one foreign key, need to select one as parent. Choice can be based on fragmentation used most frequently or fragmentation with better join characteristics.

Could use derived fragmentation for Property: Pi = PropertyForRent >branchNo Si, 3i5

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

How can we define fragments correctly?

In defining fragments we have to be very careful. Three correctness rules: Completeness Reconstruction Disjointness.

Correctness of Fragmentation
Completeness: If relation R is decomposed into fragments r1, r2, rn, each data item that can be found in R must appear in at least one fragment. This ensures no loss of data during fragmentation.

COMP 302

37 Valentina Tamma

COMP 302

38 Valentina Tamma

Correctness of Fragmentation
Recostruction: we must be able to reconstruct the entire R from fragments. For horizontal fragmentation is union operation. R = r1 r2 rn, For vertical fragmentation is natural join operation. R = r1 >< r2 >< >< rn, To ensure reconstruction we have to include primary key attributes in all fragments.

Correctness of Fragmentation
Disjointness: if data item x appears in fragment ri, then it should not appear in any other fragment. Exception: vertical fragmentation, where primary key attributes must be repeated to allow reconstruction. For horizontal fragmentation, data item is a tuple For vertical fragmentation, data item is an attribute.

COMP 302

Valentina Tamma

COMP 302

39 Valentina Tamma

Correctness of Horizontal Fragmentation

Relation: Sells(pub, address,price,type) type={Bitter, Lager} Fragments: SellsBitter= type = bitter(Sells) SellsLager= type = lager(Sells) Correctness rules Completeness: Each tuple in the relation appears either in SellsBitter, or in SellsLager Reconstruction: The Sells relation can be reconstructed from the fragments Sells = SellsBitter SellsLager Disjointness: The two fragments are disjoint, there can be no beer that is both Lager and Bitter
COMP 302
43 Valentina Tamma

Correctness of Vertical Fragmentation

Relation: Bars(name,address,licence,employees,owner) Fragments: r1 =name,address,licence (Bars) r2 = name,address,employees,owner(Bars)
Correctness rules Completeness: Each attribute in the Bars relation appears either in r1 or in r2 Reconstruction: The Bars relation can be reconstructed from the fragments Bars = r1 >< r 2 Disjointness: The two fragments are disjoint, except for the primary key, name, which is necessary for reconstruction

COMP 302

43 Valentina Tamma

Transparencies in a DDBMS

Distributed Databases
Transparency in Distributed databases

Distribution Transparency Transaction Transparency Performance Transparency DBMS Transparency

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Distribution Transparency
The user has to perceive the DDB as a single, logical entity
Fragmentation Transparency: the user does not need to know that data is fragmented Location Transparency: the user does not need to know the location of data items Replication Transparency: the user is unaware of relication of data. Naming transparency: items in a database must have a unique name, but users dont need to worry about it.

Naming Transparency
Each item in a DDB must have a unique name. DDBMS must ensure that no two sites create a database object with same name. Solution 1: create central name server. Disadvantages:
loss of some local autonomy; central site may become a bottleneck; low availability; if the central site fails, remaining sites cannot create any new objects.

COMP 302

51 Valentina Tamma

COMP 302

54 Valentina Tamma

Naming Transparency
Solution 2: prefix object with identifier of site that created it. Example: Beer created at site S1 might be named S1.Beer. Disadvantage: loss of distribution transparency.

Naming Transparency
Solution 3: use aliases for each database object. Example: S1.Beer might be known as local_Beer by user at site S1. The DDBMS has task of mapping an alias to appropriate database object.

COMP 302

55 Valentina Tamma

COMP 302

56 Valentina Tamma

Transaction Transparency
Ensures that all distributed transactions maintain distributed databases integrity and consistency. Distributed transaction accesses data stored at more than one location. Each transaction is divided into number of subtransactions, one for each site that has to be accessed. DDBMS must ensure the indivisibility of both the global transaction and each sub-transactions. Must ensure both concurrency transparency, and failure transparency

Example - Distributed Transaction

Relation: Sells(pub, beer,price,type) Fragments: SellsBitter= type = bitter(Sells) SellsLager= type = lager(Sells) The two fragments are at two different sites. Transaction T prints out the names of all pubs in the relation sells. This transaction is split into two sub-transactions, one for each fragment.

COMP 302

57 Valentina Tamma

COMP 302

58 Valentina Tamma

Example - Distributed Transaction

T prints out names of all staff, using schema defined above as S1, S2, S21, S22, and S23. Define three subtransactions TS3, TS5, and TS7 to represent agents at sites 3, 5, and 7.

Concurrency Transparency
All transactions must execute independently and be logically consistent with results obtained if transactions executed one at a time, in some arbitrary serial order. Same fundamental principles as for centralised DBMS. DDBMS must ensure both global and local transactions do not interfere with each other. Similarly, DDBMS must ensure consistency of all subtransactions of global transaction. Techniques for concurrency control. Usually different from the ones for DBMS.

COMP 302

Valentina Tamma

COMP 302

59 Valentina Tamma

Concurrency Transparency
Replication makes concurrency more complex. If a copy of a replicated data item is updated, update must be propagated to all copies. Could propagate changes as part of original transaction, making it an atomic operation. However, if one site holding copy is not reachable, then transaction is delayed until site is reachable.

Concurrency Transparency
Could limit update propagation to only those sites currently available. Remaining sites updated when they become available again. Could allow updates to copies to happen asynchronously, sometime after the original update. Delay in regaining consistency may range from a few seconds to several hours.

COMP 302

Valentina Tamma

COMP 302

Valentina Tamma

Failure Transparency
DDBMS must ensure atomicity and durability of global transaction. Means ensuring that sub-transactions of global transaction either all commit or all abort. Thus, DDBMS must synchronize global transaction to ensure that all sub-transactions have completed successfully before recording a final COMMIT for global transaction. Must do this in presence of site and network failures.
COMP 302 Valentina Tamma

Performance Transparency
DDBMS must perform as if it were a centralized DBMS:

DDBMS should not suffer any performance degradation due to distributed architecture. DDBMS should determine most cost-effective strategy to execute a request.

COMP 302

63 Valentina Tamma

Performance Transparency
Distributed Query Processor (DQP) maps data request into ordered sequence of operations on local databases. It must consider fragmentation, replication, and allocation schemas. DQP has to decide:

Performance Transparency
DQP produces execution strategy optimised with respect to some cost function. Typically, costs associated with a distributed request include:

which fragment to access; which copy of a fragment to use; which location to use.

I/O cost; CPU cost; communication cost.

COMP 302

64 Valentina Tamma

COMP 302

65 Valentina Tamma

Performance Transparency - Example

Property(Pno, City) Renter(Rno,Max_Price) Viewing(Pno, Rno) 10000 records in London 100000 records in Glasgow 1000000 records in London

Performance Transparency - Example

Assume:
Each tuple in each relation is 100 characters long. 10 renters with maximum price greater than 200,000. 100 000 viewings for properties in Aberdeen. Computation time negligible compared to communication time.

SELECT p.pno FROM property p INNER JOIN (renter r INNER JOIN viewing v ON r.rno = v.rno) ON p.pno = v.pno WHERE p.city=Aberdeen AND r.max_price > 200000;

COMP 302

66 Valentina Tamma

COMP 302

67 Valentina Tamma

Performance Transparency - Example

Dates 12 Rules for a DDBMS

0. Fundamental Principle

To the user, a distributed system should look exactly like a non-distributed system.
1. 2. 3. 4. 5. 6. Local Autonomy No Reliance on a Central Site Continuous Operation Location Independence Fragmentation Independence Replication Independence

COMP 302

68 Valentina Tamma

COMP 302

69 Valentina Tamma

Dates 12 Rules for a DDBMS

0. Fundamental Principle

Distributed Transaction Management

DDBMS must ensure:
synchronization of sub-transactions with other local transactions executing concurrently at a site; synchronization of sub-transactions with global transactions running simultaneously at same or different sites.

To the user, a distributed system should look exactly like a nondistributed system. 7. 8. 9. 10. 11. 12. Distributed Query Processing Distributed Transaction Processing Hardware Independence Operating System Independence Network Independence Database Independence

Global transaction manager (transaction coordinator) at each site, to coordinate global and local transactions initiated at that site.

Note: last four rules are ideal!

COMP 302

69 Valentina Tamma

COMP 302

4 Valentina Tamma

Distributed Concurrency Control

Techniques for Distributed Concurrency Control must ensure distributed serializability. Locking protocols (extensions of 2PL protocol) Distributed Deadlock management. Timestamping methods (extend the definition of timestamp so that it includes a site identifier)

COMP 302

Valentina Tamma