Oracle TimesTen in-memory database integration

Post on 17-Aug-2015

53 views 0 download

Tags:

transcript

Oracle TimesTen

in-memory database

integration

Jakub Žitný

Supervisor: Miroslav Potocký

Oracle TimesTen

› In-memory SQL database

› One of the ways to

speed up Oracle DB

as caching layer for Oracle DB

read-only and writethrough

faster than fully-cached Oracle DB

› Non-trivial setup

04/09/2014 Jakub Žitný 2/13

Previous work (1)

› TimesTen benchmarks

Oracle local disk

Oracle NAS

TimesTen

the speedup depends on the cache

configuration

04/09/2014 Jakub Žitný 3/13

Previous work (2)

› Caching modes

Read-only

Async writetrough

Sync writetrough

User managed

04/09/2014 Jakub Žitný 4/13

Integration

1. More testing for specific use-cases

2. Small experiments

3. Further integration

Standalone TimesTen IMDB

Caching layer for Oracle DB

4. E-groups and other “big“ use-cases

04/09/2014 Jakub Žitný 5/13

My contribution

› Setup automation

Bash script

RPM package

Puppet/LDAP env integration proposal

› Caching layer config preparation

› Foundation for further work

04/09/2014 Jakub Žitný 6/13

TimesTen setup automation

› System prerequisities

kernel, users, security, tools, env, ..., ...

› TimesTen installation

› TimesTen and Oracle configuration

› Cache users and cache groups setup

more than 30 tasks for basic installation

04/09/2014 Jakub Žitný 7/13

Bash script ‘ttdeploy‘

› 1000+ lines

› checks everything

› logs everything

› verbose mode

› debug mode

› interactive mode

› man pages

04/09/2014 Jakub Žitný 8/13

RPM package

› Download one of the RPMs

› Install it on Red Hat distro

rpm –i ttpackage.rpm

› Enjoy

04/09/2014 Jakub Žitný 9/13

Puppet integration proposal

› OS-level preparations in Puppet

› Clean service installation with RPM

› Service-level configurations from LDAP

› Not in production yet

04/09/2014 Jakub Žitný 10/13

Documentation

› Man pages

› Report

› Tutorial blogpost for TimesTen tryout

04/09/2014 Jakub Žitný 11/13

Future

› Automated analysis of candidate data

› More testing

› More benchmarks

04/09/2014 Jakub Žitný 12/13

Thank you.

04/09/2014 Jakub Žitný 13/13