当前位置:首页 >> IT/计算机 >>

PRIMECLUSTER介绍


Presentation Title

High Availability motivated by High cost of Downtime
$/hr
Airline Reservations Home Catalog Sales Pay per View Banking Telecommunications Credit Card Sales Brokerage
0 500

$ 89,500 $ 90,000 $ 150,000 $ 360,000 $ 370,000 $ 2,600,000 $ 6,450,000
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500

Example of Availability Levels in Banking: 99,5 % ? 99,9 % = 35 h = $ 12,600,000 cost-saved per year 99,9 % ? 99,99% = 8h = $ 2,880,000 cost-saved per year
Source: IDC, Gartner Group and Contingency Planning Research

The Five Steps to High Availability

The Result of High Availability

PRIMECLUSTER Overview
? Combination of Reliant Cluster and Synfinity Cluster ? Reliant Cluster ? ? ? ? Originally developed by Pyramid in 1989 First UNIX clustering product First UNIX cluster product to support Oracle Parallel Server (1993) Core sold to Veritas to create Veritas Cluster Server

? Offers integrated high-availability, parallel, and scalable network services ? Synfinity Cluster ? Developed by Fujitsu for mainframe market ? Contributes SAN-based cluster disk and file system ? Provides multi-link network trunking ? First to deliver cluster file/volume systems for Solaris ? Combined, these products represent more than two decades of clustering experience

PRIMECLUSTER

Overview

Fail-over provides High Availability Common scenarios – Fail-over

Stop Job Job Problem! Node A

Failover

Start Job

Node B Reliable Storage

Common scenarios - Mutual Standby

Stop Job A Job A Problem! Node A Reliable Storage

Job B

Start Job A
Node B

Mutual Standby – High Availability with maximised resource usage

A variation…

Stop Job A Job A Problem! Node A Reliable Storage

Stop Job B Job B

Start Job A
Node B

One way standby - High Availability with maximised resource usage in priority Situations

A Closer Look at Failover

IP A Job A

Kill!
I’m Ok. Are u?u? I’m Ok. are
Private LAN RCI

IP A Start Job A Recover FileSystem Application Database Node B

Node down
Node A

Reliable Storage

Parallel Database – Scalable Solutions

IP A Oracle RAC Problem!
Node A

RAC RAC

IP B Oracle RAC

Node B

Reliable Storage
Database Remains Up Raw Disk

Technology Sustaining High Availability

Total System Availability

Data / Network accessibility Parallel Database Data backup Redundant Redundant data volumes file systems Hardware reliability Redundant memory Redundant controllers Redundant network

Redundant power

Redundant CPU

RAID

High availability stages

Automatic restart of the applications

Automatic detection of system/ component failures

PRIMECLUSTER - Building Availability
? PRIMEPOWER + IHV RAID ensures hardware reliability ? The best cluster foundation is hardware that does not fail! Total Application Availability ? PRIMECLUSTER Foundation plus ISV applications provide data and network reliability Automatic Restart Automatic Detection/Correction ? ISV provides data management and backup/restore of Applications of System Component Failures functions PRIMECLUSTER Services ? PCL provides data integrity Data / Network Reliability ? PRIMECLUSTER services (HA, Parallel, and Scalable) provide Total Availability ? Cluster configuration insures system resources are available ISV PRIMECLUSTER Foundation ? Applications services are assured of resources/restart
Parallel Database Data Backup Redundant File Systems Redundant Data Volumes Redundant Networks

Hardware Reliability
Redundant Memories

Redundant Power Supplies

Redundant CPU

Redundant controllers

RAID

PRIMEPOWER

IHV

Key Features
? Application setup is done using “wizards” ? Simple prep work enables application resource agents and rules to be built automatically ? Specialized wizards for Oracle9i RAC, SAP R/3, and EMC SRDF are available ? Maintains existing customer environment ? Applications require no changes ? Applications are not “ported” to the cluster ? No changes to maintenance, diagnostic procedures ? Works with existing file systems (i.e, VxFS, UFS, etc) ? Data can be abstracted into global resources: ? Single-image file system(s) ? Global Disk and File System (GDS/GFS) ? Introduces new volume and file managers ? Uses SAN for data transfer

Key Features (continued)
? Supports Solaris and Linux in the same cluster
? Powerful

Solaris back-end/database server ? Inexpensive Linux front-end/application servers
? Sixty-four nodes in a cluster ? Object-oriented design is flexible, allows high level of granularity for monitoring and control ? Fast node failure detection (PRIMEPOWER) ? Positive node-kill insures application data integrity

Operational Components
? Cluster Foundation (CF) ? Infrastructure for all core cluster processes ? Cluster services ? Reliant Monitor Services (HA) ? High availability monitoring and response ? Scalable Internet Services (SIS) ? Load balancing of network services ? Parallel Application Services (PAS) ? Parallel database services for Oracle9i RAC ? Custom Services ? Special request fulfillment ? Global Disk and File Systems (GDS/GFS) ? SAN-based cluster volume management and file system ? Global Link Services (GLS) ? Multi-node Ethernet trunking

Operational Components
Enterprise Management

Applications

PRIMECLUSTER
Cluster Management
High Availability Parallel Applications Scalable Internet Custom Services

Cluster Foundation
Solaris
PRIMEPOWER SERVER

Solaris
PRIMEPOWER SERVER

Solaris
Sun SERVER

Linux
Primergy Server

Linux
Primergy Server

Linux
Primergy Server

System Area Network/Global Disk and File System

GDS: Global Disk Service
? Cluster volume manager allowing simultaneous access from more than 2 nodes ? Provides cluster-wide, consistent device naming ? applications use the same name whatever node they are running on

/dev/sdfsk/class / rdsk/oracle_vol1

/dev/sdfsk/class / rdsk/oracle_vol1

/dev/sdfsk/class / rdsk/oracle_vol1

SAN

Local Disc Mirroring Volume management, Hot spare & Striping

Mirror volumes (Including inter-RAID mirroring)

GDS: Global Disk Service
? Core Functionality ? Software RAID0, 1, 0+1 ? Software partitioning (up to 256 partitions/slices per volume) ? Root disk mirroring ? Software volume management on both local and shared disks ? Quick recovery after panic through logging based recovery ? Hot swap/spare disk management ? Easy Management ? Java based easy/intuitive administration ? Automatic detection of shared disk ? Intuitive GUI interface ? Extended data management ? Instant backup/restore by detaching/re-attaching a portion of the mirror disk (Snapshot)

GFS: Global File Service
Multiple servers simultaneously access the same file systems using their own Fibre Channel array access paths
PRIMEPOWER
user application user application user application user application user application

PRIMECLUSTER GFS
SAN

GFS: Global File Service
? High availability filesystem ? Log-based quick recovery (fsck) ? Redundant meta-data ? Superior data integrity versus NFS (cache coherence issues) ? High performance filesystem ? Direct data access from each node (not via interconnect) ? Each node caches file data locally ? Extent-base block allocation ? Provides "local" version of GFS for local file-system ? Very large scale file system for HPC applications ? 1 TB standard, 32TB has been implemented

GLS: Global Link Service
? Provides a variety of methods for implementing redundant public network paths ? Caters for failure in NICs and network infrastructure ? Transparent recovery operation ? Scalable pathing (Trunking) is also available ? Application fail-over if all paths fail
Application
GLS I/F NIC NIC

Applications
GLS I/F NIC NIC

Allows application to be failed-over to another node if all NIC fail

PRIMECLUSTER Products and Packages
You can purchase package products with a single product Id or

?mix and match? for specific requirements CF is a prerequisite for EVERYTHING (except GLS) GDS is a prerequisite for GFS Current version 4.1A20 Individual Products
CF Enterprise Edition HA Server Parallel Server Scalability Server ? ? ? ? RM S ? ? WT ? ? ? ? PAS ? SIS ? GD S ? ? ? GFS GLS ? ? ? ? ? ?

Packages


相关文章:
如何选择与实施双机热备及高可用性方案
高效、冗余的集群心跳协议――PRIMECLUSTER 可拥有多达 8 条心跳线路;采用自有的...LifeKeeper 集群软件:支持多点集群及双机 LifeKeeper 软件介绍 美国 SteelEye ...
ETERNUS DX8400 DX8700 磁盘存储系统
1.1.10.提高存储系统可用性的单元间镜像——PRIMECLUSTER 为实现更高的业务连续性,PRIMECLUSTER(*注 1)为 ETERNUS 磁盘存储系统提供了 单元间镜像功能。该功能可...
更多相关标签: