
bff626967090b47c8460f77820ba8b5c.ppt
- Количество слайдов: 16
<Insert Picture Here> Oracle Database 11 g: First Experiences with Grid Computing Plamen Zyumbyulev Head of DBAs
Mobiltel (M-Tel) • M-Tel is the market leader in Bulgaria and one of the most successful mobile operators in Europe • More than 4. 8 M subscribers • Founded in March 1994 • Part of Mobilcom Austria Group • M-Tel became the world fifth to offer HSDPA services
The IT reality • IT becomes more and more complicated • New roles and tasks • Hundreds of systems • Increased system dependencies and integration • Lot’s of new systems • Constant changes / Increased Risk • Everything is important and 24/7 • Increased demand for reliability, availability, performance, etc. • Islands of Computation • Limited scalability, no resource sharing • Must be configured for peak loads • Single point of failure • Slow to adjust to business needs
Challenges • Provide Highest Quality of Service • implement new technologies • decrease the risk • Be more flexible • online changes, virtualization, dynamic provisioning • How to do more with less • automation, consolidation, monitoring
Why Grid? Data Warehouse Technical Benefits • Dynamically configurable Custom Application Service Oriented Applications infrastructure • Effective (automatic) workload ERP Application management CRM Application • No single point of failure • Incremental scaling • Centralised automated monitoring & management Better Quality of Service Fast response to changes Effectiveness Management Mid-Tier Database Tier Storage
Why Grid Computing with Oracle 11 g • HA and DR • Real Application Clusters (RAC) • DR with real time query • Service Oriented Architecture • Resource Management • Automation and simplification • Oracle Enterprise Manager Grid/DB Control • Automatic Diagnostic Workflow. Incident Packaging Service • Online Maintenance • Change assurance • Database Replay • SQL Plan Management • Snapshot standby
Grid Configuration • Hardware • 6+1 Nodes, HP Blade. System c-Class , 2 CPU dual Core 3. 2 GHz, 10 GB RAM • Red. Hat Enterprice Linux 4 • Oracle Stack • Oracle Database 11. 1. 0. 5 • Oracle RAC 11. 1. 0. 5 • Oracle Clusterware 11. 1. 0. 5 • Oracle Automatic Storage Management 11. 1. 0. 5 • Storage • IBM DS 6800 – 28 disks
Our GRID implementation • Big RAC environment • Consolidate many Business Systems into small number of databases • Single database – this is not mandatory • It is possible to have different DB versions • Many services per Business System – OLTP, Batch, Read Only, Test, etc. • Every system/service is separate from the others • Separate tablespaces • Guaranteed resources – Memory, CPU, I/O, etc
Our GRID Implementation … • Many features are configured out of the box • High availability • Backup • Disaster recovery • Force logging can be enabled per tablespace • Reporting instance • Test System • Extremely fast provisioning of new systems, System Cloning, Creation of Test Systems • Services can span databases * • Oracle EM 10 g Grid Control can monitor 11 g GRID
Our GRID Architecture Business BS 1 BS 2 BS 3 BS 4 BS 5 BS 6 systems BS 7 SOA Resource Manager DR with real time query – no idle resources Continuous Redo Shipment and Apply Open in Read Only Single Database – Not mandatory One or more DBs
Fast Test System Creation Business BS 1 BS 2 BS 3 BS 4 BS 5 BS 6 BS 7 SOA systems Replay Driver Resource Manager Process Capture Workload Open in Read/Write Single Database – Not mandatory One or more DBs
Resource Management Rules to automatically assign sessions to Consumer Groups Master Plan Active on instances … directive values, thresholds, actions BS 1 Subplan % BS 2 Subpan % directive values, thresholds, actions BS 1 Group 1 % BS 1 Group 2 % Permitted users … …
Service allocation (DG in RO mode) Business System 1 Service name / Instance N online batch … RO … Business System 2 test online batch … RO test 1 P P A* A** A A D D 2 P P A* D A A D D 3 A A D D P P A* D 4 A A D D P P A* D 5 A A D D 6 A A D P** A A D D DG 1 D D P* D --- P* D A – available; P – preferred; D – disabled; DG – data guard; RO – Read Only; * - implemented with tns; ** - different schema
Service allocation (DG in RW mode) Business System 1 Service name / Instance N online batch … RO … Business System 2 test online batch … RO test 1 P P A* A** A A D D 2 P P P* D A A D D 3 A A D D P P P* D 4 A A D D P P A* D 5 A A D D 6 A A D P** A A D D DG 1 D D D* D --- D* P A – available; P – preferred; D – disabled; DG – data guard; RO – Read Only; * - implemented with tns; ** - different schema
A Paradigm shift We cannot solve our problems with the same thinking we used when we created them. Albert Einstein • Start small • Test systems • Development systems • Temporary/Permanent storage for data from other systems. (ex. 1 GB in Billing System means ~ 15 GB storage) • Non critical production systems • Create “Atmosphere”. Fight Ignorance and Status quo • Build expertise, confidence and habits
The Results • Availability 100% • Lot’s of changes, reconfigurations and experiments • Lot’s of problems, bugs, human errors • No unplanned downtime since the beginning - May 2007 • GRID is in the production since July 2007 • Oracle Database 11 g 32 bit beta 5 for Linux • There is a big list of new important systems waiting for Linux 64 bit production release • Mobiltel gained a lot of experience and confidence with Oracle 11 g GRID computing
bff626967090b47c8460f77820ba8b5c.ppt