Ingres Version 6.4 An Overview of the Architecture - PowerPoint PPT Presentation

About This Presentation
Title:

Ingres Version 6.4 An Overview of the Architecture

Description:

Title: Ingres 6.4 DBMS Architecture Subject: Understanding how Ingres works internally Author: Dennis Adams Last modified by: Dennis Adams Created Date – PowerPoint PPT presentation

Number of Views:36
Avg rating:3.0/5.0
Slides: 26
Provided by: DennisA150
Category:

less

Transcript and Presenter's Notes

Title: Ingres Version 6.4 An Overview of the Architecture


1
Ingres Version 6.4An Overview of the Architecture
  • Presented by
  • Quest Software

2
Ingres 6.4 - An Overview
  • Client / Server Architecture
  • Components
  • Locking I/O
  • Server Internals
  • Query Processing

3
Client - Server Architecture
4
Client - Server Architecture
  • User Image
  • Name Server
  • DBMS Server
  • Archiver
  • Recovery Process

5
Name Server
  • isql dbname
  • image activated
  • call Ingres General Communication Facility (GCF)
    routines
  • opens connection to Name Server

6
Name Server
7
Name Server
  • Uses GCA Protocol
  • Receives connection request from User Image
  • Provides mailbox address to User
  • User Image Connects to DBMS
  • What about Ingres / Net ?

8
Name Server - Ingres /Net
9
DBMS Server
  • The Ingres engine
  • Accepts incoming SQL
  • Processes it
  • Executes I/O on behalf of Users
  • Multi - threaded

10
Components
11
Logging and Recovery
  • DBMS Server
  • writes to memory log buffers
  • Recovery Process
  • Reads from memory log buffers
  • Writes to Physical Log File
  • Returns ok to complete to Server
  • DBMS Server
  • writes to the database

12
Logging and Recovery (cont)
13
Logging and Recovery (cont)
  • Circular Log File
  • All transactions which update the database are
    logged
  • Writes Before Image to log file
  • Transaction Complete stamp after complete
  • Write to the log file BEFORE the database

14
Group Commit / Fast Commit
  • Group Commit
  • Sever writes query A to log buffers
  • Sever writes query B to same buffers
  • All commits are written together
  • Fast Commit
  • Server writes forces directly to Log File
  • then writes to the database (Cached)
  • Write-Behind threads write to the database

15
Archiving
  • Log File has periodic Consistancy Points
  • Archiver reclaims space from the Log File
  • Only extracts COMMITED transactions from the Log
    File
  • Writes transactions to the Journals
  • Moves the End Of File marker forward

16
Archiving (cont)
  • If there are Uncommitted Transactions in the Log
    File
  • Archiver Process cannot extract them
  • Cannot move the EOF marker
  • Log File Fills Up
  • Chaos !

17
Components
18
Locking
  • Initiated by Server Process
  • Types
  • Exclusive / Shared
  • Levels
  • Page / Table / Database
  • VMS Cluster
  • II_CSP talks to clusterwide lock manager

19
Locking (cont)
  • Locking Escallation can Occur when
  • MAXLOCKS exceeded
  • Locks Per Transaction exceeded
  • Ingres installation-wide resources low
  • Blocking Locks
  • Cause Deadlock checking to start

20
I/O Handling
  • Unix has Synchronous I/O
  • call iislave processes to offload I/O requests
  • II_NUM_SLAVES
  • how many ?
  • VMS has Asynchronous I/O

21
Server Internals
22
Query Processing - Parsing
  • Query arrives in Server Control Facility (SCF)
  • Store Text in Query Storage Facility (QSF)
  • Parse the Query (PSF)
  • Check the Relational Descriptions (RDF)
  • Check Query Modifications due to Permits, Views
    etc. (QRYMOD)

23
Query Processing - Planning
  • Create a Parsed Query Tree in QSF
  • Optimiser (OPF) uses Query Tree Statistics to
    develop Query Plan
  • Store the Query Plan in QSF
  • Query Execution Facility validates table
    descriptions (QEF)

24
Query Processing - Execution
  • Access the actual files using the Data
    Manipulation Facility (DMF)
  • DMF returns data to QEF
  • QEF returns data to SCF
  • SCF returns data to GCA communications
  • Eventually ends up at the users front end
    program !

25
Conclusions
  • Ingres DBMS is complex
  • Multiple Components
  • Handle with care !!!
Write a Comment
User Comments (0)
About PowerShow.com