Database Technologies 2 (CSY3024)

DBMS architecture & features

Dr James Xue
School of Science & Technology University of Northampton

Database architectures
There is a wide variety of different DBMS architectures Teleprocessing File-Server Architecture Two-tier Client-Server Architecture Three-tier Client-Server Architecture N-tier Client-Server Architecture Peer-to-Peer Architecture Distributed DBMS Cloud Architecture Mobile databases
2

Teleprocessing
Traditional multi-user system architecture Single mainframe and multiple terminals Heavy load on the central mainframe Run application programs and DBMS Format data for presentation on terminals Tendency to replace mainframes with network of personal

computers (downsizing)

File server architectures Client-server architectures .

File-Server Architecture

A file-server is a computer that is connected to a network and

mainly servers as a shared storage

E.g., for shared access to databases

In a file-server architecture the processing is distributed over the

network

Workstations (application and DBMS) request data (files)

4

File-Server Architecture (cont.)

SQL request example

Since the file-server is not SQL-aware, the DBMS must

request files corresponding to the Branch and Staff relations from the file server, rather than just the staff names that satisfy the request Disadvantages
Heavy network traffic Each work stations runs a full instance of the DBMS Complex integrity, concurrency and recovery control Multiple DBMSs may concurrently access the shared file

Two-tier Client-Server Architecture

Application consists of a client (1st tier) and a server (2nd tier) that

might run on different machines

Clear separation of concerns between client and server Thin client vs thick client Less or more application logic on the client side

Support decentralised business environments

6

Two-tier Client-Server Architecture

Client (1st tier) tasks Presentation of data (user interfaces) Business and data application logic Send the database request to the server and process the results Server (2nd tier) tasks Manage (concurrent) data access (data services)

Business logic (e.g., validation of data)

Authorisation, integrity checks, query/update processing, recovery control,

Different possible client-server topologies Single client and single server Multiple clients and single server Multiple clients and multiple servers
7

Three-tier Client-Server Architecture

In the 1990s, the three-tier client-server architecture was

introduced to address the enterprise scalability issue

e.g., driven by emerging web applications

Application consists of a presentation tier (client) , a logic tier

(application server) and a data tier (database server) that might run on different platforms
8

Three-tier Client-Server Architecture

Presentation (1st tier ) tasks Presentation of data (user interfaces) Basic input validation (thin client) Send request to the server and visualise results Logic tier (2nd tier) tasks Business logic Data processing logic Data tier (3rd tier) tasks Basic data invalidation Manage (concurrent) database access (data services)
Authorisation, integrity checks, query/update processing, recovery

control

Three-tier Client-Server Architecture

10

N-tier architecture

The three-tier architecture can be extended with

additional intermediary tiers for increased flexibility

Increases the flexibility for load balancing by introducing multiple web

servers and/or edege servers.

11

Peer-to-Peer Architecture

Systems changing information and services in a peer-to-peer (P2P)

manner without a central authority

No global schema and need for schema integration (matching)

Data and service sharing No dedicated clients and servers Sites may dynamically form new client/server relationships
12

Middleware
Software that connects (mediates) between software

components or applications
provide a uniform interface

Hide complexity of heterogeneous and distributed components and

There exists different types of middleware Remote procedure call (RPC)

Java RMI CORBA XML RPC

Asynchronous publish/subscribe Subscribe for different types of messages

13

Distributed DBMS (DDBMS)

Distributed database Legally related connection of shared data and metadata that is distributed over a network Distributed DBMS Software system to manage the distributed databases in a transparent way
14

Distributed DBMS (DDBMS)

Distinction between local and global transaction Local transaction
Accesses only data from the site from which the transaction was initiated

Global transaction Accesses data from several different sites

Reasons for building a distributed DBMS Data sharing

Possibility to access data that resides at other sites

Autonomy Each site retains a certain degree of control over the local data Availability If one site fails, the other sites may still be able to continue operating Data might be replicated at several sites to increase the availability
15

Distributed DBMS (DDBMS)

Costs and scalability Use cluster of PCs in stead of large mainframe systems Integration of existing DBMSs Coexistence of legacy systems with new applications Dynamic organisational structure Mergers and acquisitions

Implementation issues Transactions have to be executed atomically across different sites (two-phase commit protocol)
Commit decision is left to a single coordinator

Distributed concurrency control Deadlock detection has to be carried out across multiple sites
16

Parallel Database Architectures

Parallel machines (multiple processors) can be used to

speed up the transaction processing Different models for parallel database architectures
Shared-memory, shared-disk, shared-nothing and hierarchical

17

Parallel Database Architectures

CPU CPU CPU CPU Interconnection network

Memory

(Shared-memory)

Shared memory Processors and disks have access to shared memory via a bus Very efficient communication between processors Not scalable since bus becomes a bottleneck

18

Parallel Database Architectures

Memory Memory Memory Memory CPU CPU CPU CPU

Interconnection network

(Shared-disk)

Shared disk All processors can access all disks via a communication network Each processor has its own memory Certain degree of fault tolerance if processor/memory fails Also disks maybe have fault tolerance (e.g., RAID architecture) Interconnection to the disk systems becomes bottleneck
19

Parallel Database Architectures

Memory Memory CPU CPU Interconnection network

(Shared-nothing)
CPU Memory

CPU

Memory

Shared nothing Each node consists of a processor, memory and one or more disks High-speed interconnection network between processors More scalable than shared memory or shared disk model Increased communication costs for non-local disk access

20

Parallel Database Architectures

Hierarchical Combines the different models (composition) Top-level is shared nothing between nodes
Each node can be a shared memory or shared disk subsystem

21

Object Databases
Motivation to overcome weakness of relational

approach.
Richer data models. Closer integration with programming languages.

Better for situations that there are many objects, but less

data for each object. E.g., CAD ODMG created OQL standard, but very few vendors implemented it

22

Object Databases
Kinds of object databases Object relational (Oracle, DB2, PostgreSQL) Semantic data model (Jasmine) Programming language centred (Objectivity, FastObjects, ObjectStore). Commercial Object database (ObjectDB)

23

Mobile databases
Users want to access information on the move via mobile

devices

Tourist information systems Salesperson who is visiting their customers Emergency services

New requirements for mobile DBMSs Small footprint databases that can run on mobile devices with limited resources Location dependent queries Context-aware queries
24

Mobile databases (cont.)

Communicate with centralised database server via wireless network

or fixed Internet connection Replicate data on a centralised server and on a mobile device
Synchronisation challenges

Caching of data and transactions to cope with potential network

connection failures Opportunistic (peer-to-peer based) information exchange with other mobile DBMSs
e.g., dynamic P2P Bluetooth connections with other devices in range

(proximity-based information exchange)

Security Which portion of a database can/should be replicated on a mobile device?

25

Cloud computing

Internet-based computing with on-demand and pay-per-use access

to shared resources, data and software Main characteristics

Elasticity Pay only for the services that are actually used (pay-as-you-go) Web-based access (e.g., Web Service API or browser)
26

Cloud computing Layers

27

Cloud data service example

Amazon Simple Storage Service (Amazon S3) Online storage service with unlimited storage space Store objects (up to 5GB in size) in buckets Web Service API Amazon SimpleDB Distributed database written in Erlang Offers a Web Service API Makes use of S3 and EC2 On demand scaling Non-relational data store
Schemaless Hashtables with set of key-value pairs

28

29

Exercises
List the benefits of multi-tiered architecture Compare with the MVC (Model-View-Controller)

architecture

30