Call (800) 766-1884 for Oracle support & training
Free Oracle Tips


Oracle Consulting Support
Oracle Upgrades
Use New Oracle Features
Oracle Replication Support
Oracle Training
Remote Oracle DBA
System Documentation
Oracle Tips
Oracle Performance
 

Free Oracle Tips


 

HTML Text

BC Oracle tuning

Oracle training

Oracle support

Remote Oracle

 

Donald K. Burleson

Oracle RAC Tips

Components of a Cluster

Cluster configuration is usually comprised of two or more servers or nodes and an external SCSI or Fiber Channel array. The cluster nodes communicate through high-speed interconnects and are controlled by cluster software or a cluster manager application. The cluster functions as a single computing resource, but it is composed of a logical stack of integrated components. Let us examine the characteristic features of the important cluster components.

Cluster Nodes

A node in a cluster can be as simple as a computer with a single processor, or it can have a symmetric multiprocessor (SMP) or NUMA architecture. An independent copy of the Operating Environment usually characterizes each cluster node. However, some may share a single boot image from the central, shared disk storage unit. This could be, for example, an independent physical server such as the Sun Enterprise™ 6800, or a single domain within a Sun Enterprise 15000, or an inexpensive Dell workstation running a Linux operating system. There are two main architectural approaches for multiple processor systems:

  • Symmetric Multi-Processors (SMP)
  • NUMA (or DSM - Distributed Share Model)

Let us look at these architectures.

SMP Systems

In the SMP architecture, the computer utilizes multiple internal processors (CPU/s) and they all share the same system memory and IO resources. [Fig 3.4] Sharing is achieved through the use of a high-speed system bus. One multi-CPU-aware copy of the O/S runs on the computer and controls all of the processors. Thus, the SMP architecture is also called the 'Shared Memory Architecture'. In an SMP environment each processor executes tasks independently. A database application like Oracle, which has the ability to spawn multiple processes, is quite suitable for the SMP architecture and takes advantage of the SMP architecture to perform many parallel operations simultaneously.

Fig 3.4 SMP based node

With current technology advances, present SMP machines are able to house up to 100 CPUs in a single SMP computer system. Table (3.1) shows some of the popular SMP-based servers offered by the leading server vendors.

Table 3.1 - SMP Servers

The advantages of utilizing SMP-based servers include:

  • They provide an incremental path to improve performance by adding extra processors as needed.
  • Many applications designed for a single processor work seamlessly with SMP and take advantage of the existence of additional processors.
  • SMP technology is mature and used widely.
  • Administrative overhead from the operating system point of view is minimal.

However, scalability has often been a problem with SMP. Although many SMP-based systems from Sun and IBM support 64 to 128 processors, it is not enough to add a few sockets to a plug-in processor; we need to upgrade the entire machine to achieve higher computing power. A high-speed bus and large memory are needed.

Another issue with SMP is the lack of high-availability. If one component of the SMP machine should fail, the entire machine can become unusable, regardless of the number of processors present. In addition, SMP servers with large numbers of processors are prohibitively expensive. For example, the simple addition of two processors with companion memory boards to an existing system may cost over a hundred thousand dollars.

However, when several individual SMP nodes are aggregated to form a cluster, they provide a relatively inexpensive way to achieve scalability. Smaller SMP nodes (such as those provided by DELL and HP-Compaq) are relatively inexpensive, and the aggregation of such SMP systems would result in a powerful computing cluster (at the time of this writing, a dual node, dual CPU cluster with a shared 5-36 gig disk array can be purchased for around $24,000.00).

NUMA Architecture

In NUMA (non-uniform memory access) architecture, multiple processors within a computer system are grouped. This configuration is usually called a 'Quad' (or Node Card in a SGI server), and it has its own memory and I/O controller. Quads are connected by high-speed interconnects. Unlike a cluster, the quads are part of a single node. Thus, a NUMA system can be thought of as a large SMP system.

However, the memory is non-uniformly distributed to the processors. Each quad has its own localized memory, but the memory is accessible to other quads. To the processors, all of the memory in a NUMA machine appears the same; the only difference is in access time. NUMA is also called distributed shared memory (DSM) architecture. Good examples of NUMA systems include the Sequent (now IBM) servers and Silicon Graphics (SGI) 2000 / 3000 series.


For more information, see the book Oracle 11g Grid and Real Application Clusters 30% off if you buy it directly from Rampant TechPress . 

Written by top Oracle experts, this RAC book has a complete online code depot with ready to use RAC scripts.  



  ”call


Download your Oracle scripts now:

www.oracle-script.com

The definitive Oracle Script collection for every Oracle professional DBA

  
 

Oracle performance tuning software 
 
 
 
 

Oracle performance tuning book

 

 
 
 
Oracle performance Tuning 10g reference poster
 
 
 
Oracle training in Linux commands
 
Oracle training Excel
 
Oracle training & performance tuning books
 
   

Copyright © 1996 -  2014 by Burleson. All rights reserved.

Oracle® is the registered trademark of Oracle Corporation.