Jerry Held

1
(No Transcript)
2
Merrill HoltDirector Parallel Server Product
Management Oracle Corporation
3
Oracle Parallel Server Cache Fusion Delivers
Scalability Oracle9i
4
Oracle9i Parallel Server

• Oracle9i Parallel Server is designed for todays
  most demanding deployments
• Server consolidation means very large user
  populations
• Critical e-business requires full time service
• Rapid growth shortens capacity planning

5
Oracle9i Parallel Server

• Major changes in Oracle9i
• Scalability with full Cache Fusion architecture
• Improved availability
• Improved manageability

6
Oracle Parallel Server Architecture
Users
Network
Centralized Management Console
No Single Point Of Failure
Low Latency Interconnect VIA or Proprietary
High Speed Switch or Interconnect
ClusteredDatabase Servers
Hub or Switch Fabric
Storage Area Network
Drive and Exploit Industry Advances in Clustering
Mirrored Disk Subsystem
7
Database Challenges Scalability

• Scale to Millions of Users
• Transparently
• All types of applications
• Scale workloads without limits
• Grow storage easily

Oracle Parallel Server
8
Scalability Overview

• Transparent application scalability
• Enhanced IPC (Inter-Process Communication)
• Simplification of lock configuration
• Tuning flexibility
• Tuning automation

9
Out-of-the-Box Transparent Application Scalability

• In the past clustered databases scaled well for
  specific types of applications
• Data Warehouse
• Parallel-enabled OLTP
• Oracle Parallel Server with Cache Fusion is a
  breakthrough in parallel database technology
  delivering transparent scalability to all types
  of applications

10
Cache Fusion Architecture
Users

• Full Cache Fusion
• Cache-to-cache data shipping
• Shared cache eliminates slow I/O
• Enhanced IPC
• Allows Flexibleand Transparent Deployment

11
Full Cache Fusion

• Oracle9i Cache Fusion increases performance and
  scalability by reducing communication latency
• Data is shipped directly over high speed
  interconnect
• Minimize disk I/O

12
Cache Fusion ManagesInter Instance Requests

• Readers and writers accessing instance A gain
  access to blocks in instance Bs buffer cache
• Four types of block contention
• Coordination by Distributed Lock Manager

Lock StatusBlock inCache B
Request for BlockCache A
Read
Read

Write
Read
Read
Write
Write
Write
13
Transparent Scaling

• Scalable Instance Coordination
• Multiple LMDs (Lock Manager Daemon)
• Intelligent message batching
• Enhanced flow control and buffer handling
• Faster, concurrent cross instance calls using IPC
  instead of multiple DLM operations
• Enhanced asynchronous IPC

14
Distributed Lock Manager
LMD
SGA
Inter-instance lock processing
Shared Pool
LMON

LockDatabase
LCK
DLM
15
Processes in a Parallel Server Instance
Fulfills lock manager service for PCM locks
Handles CacheFusion Processing
LMD
Other Processes - PMON - SM ON - DBWR - LGWR -
CKPT
Provides Cluster Monitoring
LMON
HandlesNon-PCMLocks
Foreground Processes (Oracle Shadow Processes)
LCK
16
Enhanced IPC

• Global IPC service layer
• Each requesting server process handles asynch IPC
• Efficient I/O implementation
• Remote memory operations for direct DMA
• Exploit modern low latency interfaces
• New V tables for IPC statistics

17
Lock Simplification and Automation

• Automatic DLM configuration
• Automatic derivation of DLM configuration
• No INIT.ORA lock parameters required
• Improved lock efficiency and memory management
• Dynamic lock affinity
• DLM locks dynamically remastered reducing locking
  overhead
• Cache layer determines policy for remastering

18
Tuning Flexibility

• Improved insert concurrency
• Flexibility in free-list group mangement
• Add free-list groups to existing database objects
• Altering free-list groups of existing database
  objects

19
Database ChallengesAvailability

• Be available all the time
• design for fault tolerance
• provide fast andreliable fault recovery
• eliminate maintenance downtime
• Support mission-critical business operations

Oracle Parallel Server
20
No Single Point of Failure

• Parallel Server
• Exploits clusters for very high availability
• Overcomes the limitations of traditional failover
  solutions by providing
• Concurrent processing
• Load Balancing
• Fast time to recovery
• No single point of failure
• Network, Storage
• CPU, OS

21
NEW in Oracle9i
Availability Features

• Reduced time to recovery
• Concurrent lock reconfiguration and instance
  (cache) recovery
• Lock replay only for dead masters/locks
• Deferred/lazy lock remastering
• Batching of reconfiguration messages
• Parallel replay processing by multiple LMDs
• Optimal lock invalidation
• Optimized special reconfiguration case (e.g.
  2-gt1)
• Fast/reliable detection of node/network failure

22
Database ChallengesManageability

• Create one virtual system to configure and manage
• Single system image for the database integrated
  with the cluster

Oracle Parallel Server
23
Oracle Parallel Server Manageability

• Single system Image
• Single Oracle Database
• One virtual system to configure and manage
• Single management console
• Cluster-wide Monitoring Diagnostics

24
NEW in Oracle9i Parallel Server Manageability
Features

• Improved Single System Image
• Exploit cluster file systems
• Improved tools infrastructure for management
• Oracle Universal Installer (OUI)
• Enterprise Manager (EM)
• Database Configuration Assistant (DBCA)
• Net8 Assistant (Net8CA)
• Recovery Manager (RMAN)
• First pass diagnostics

25
Improved Configuration

• Centralized, persistent configuration storage
• Eliminates consistency problems with the per node
  text file based Parallel Server configuration in
  prior versions.
• Utilities to migrate previous Parallel Server
  configurations
• Enhanced DBCA Net8CA functionality
• Robustness
• Capability to add and delete instances

26
Improved Administration

• Dynamic node addition to cluster
• Add a node in both the system and Parallel Server
• Better integration with Oracle Enterprise Manager
• View and update server side initialization
  parameter file (SPFILE)
• Assign private rollback segments to Parallel
  Server instances

27
Improved Administration

• Better integration with OEM (Contd.)
• Enhanced monitoring capability and events
  available with OEM and performance packs
• Cache Fusion statistics
• Statistics for interconnect block traffic
• Statistics for the entire database or per
  instance
• New EM events associated with new statistics

28
Improved Diagnosibility

• Server side capability
• First pass analysis on problems
• Reduces the turnaround time to resolve bugs
• Enables analysis of intermittent, hard to
  reproduce problems
• Reduces need for diagnostic patches

29
Microsoft Windows Platform Integration

• Integration with Microsoft Cluster Server
• Cluster Pre-install Tool
• Disk diagnostics
• Extensive consistency checks across nodes
• Multi-Instance Capability
• Enable multiple instances on a single node
• Multi-Database Capability
• Support multiple databases on the same cluster

30
Oracle9i Delivers Performance and Scalability

• In todays rapidly changing environment
• Internet users and transaction volumes grow
  extremely rapidly
• Data warehouse systems must support very large
  data volumes
• Server consolidation drives the need for large
  scale systems
• Oracle9i provides the ability to deploy as you
  grow

31
A
32
(No Transcript)