Compaq 9th Annual Systems Engineering Conference – Living LAB
ECG05/0998 August, 1998
Oracle Fail Safe 2.0 Performance on Compaq
CONTENTS
ABSTRACT ....................1
INTRODUCTION...........3
Living Lab
Demonstrations ..........3
OFS Overview ............ 3
Scope..........................3
TEST APPROACH .........3
TEST ENVIRONMENT .......3
Hardware ................... 3
Software ..................... 3
Benchmark Factory
Overview.................... 6
Client re-connect
implementation...........6
TEST FINDINGS ............6
APPENDIX A ..................7
Installation and
Configuration of Oracle
Fail Safe 2.1 on
Compaq ProLiant
Servers for the 9th ASE
Conference................. 7
INSTALLATION TIPS . 9
APPENDIX B
(PREPARED BY
ORACLE
CORPORATION).........11
ORACLE FAIL SAFE........ 11
ORACLE FAIL SAFE
MANAGER..................... 11
CLIENT APPLICATIONS... 12
FAILOVER TIMING ......... 12
PRODUCT AVAILABILITY 12
FURTHER INFORMATION 12
Oracle Fail Safe(OFS) on Compaq servers has been one of the most
reliable and mature high availability solution on Microsoft Windows
NT MSCS. Compaq’s Database Technology and Solutions
Center(DTSC) has conducted a series of performance tests to
characterize the performance of Oracle Fail Safe running on Compaq
ProLiant servers. The tests also captured the impact of system loads as
additional processors were added to the test systems.
These tests were performed at the Compaq/Oracle International
Competency Center (ICC), located within the Compaq DTSC in San
Bruno, California. The Compaq/Oracle ICC was established to
address Oracle database related issues such as solution development
and validation, advanced technology, customer case studies and proof
of concept projects.
The main focus of the tests were to,
q Evaluate the system performance in the two node fail over
environment
q Capture the fail over performance of the system under various
user loads
q Characterize the performance of the surviving node as well as the
failing node when using one, two, or four processors on both
servers
The test results indicated,
Ø The overall performance in the OFS environment is highly
dependent on the system load of each server.
Ø When the combined system load on the surviving node and the
failing node was under 90%, the performance of the surviving
node is similar to the baseline tests. The baseline tests were
implemented to measure the performance of each server without
the fail over event.
Ø When the combined loads were over 90%, the fail over
performance and the performance of the surviving node degraded
compared to the performance of the baseline tests.
Ø When the surviving node was heavily loaded, adding additional
processors to the node helped to improve the overall system
performance.
ProLiant Servers
ABSTRACT
P
A
P
E
P
P
E
E
R
R
R
P
A
P
A
2
NNOOTTIICCE
The information in this publication is subject to change without notice.
COMPAQ COMPUTER CORPORATION SHALL NOT BE LIABLE FOR TECHNICAL OR
EDITORIAL ERRORS OR OMISSIONS CONTAINED HEREIN, NOR FOR INCIDENTAL
OR CONSEQUENTIAL DAMAGES RESULTING FROM THE FURNISHING,
PERFORMANCE, OR USE OF THIS MATERIAL.
This publication contains information protected by Copyright, except for internal use
distribution; no part of this publication may be photocopied or reproduced in any form without
prior written consent from Compaq Computer Corporation.
This publication does not constitute an endorsement of the product or products that were
tested. The configuration or configurations tested or described may or may not be the only
available solution. This test is not a determination of product quality or correctness, nor does it
ensure compliance with any federal, state or local requirement. Compaq does not warrant
products other than it’s own as stated in product warranties.
Product names mentioned herein may be trademarks and/or registered trademarks of Oracle
Corporation, Client/Server Solutions INC., Compaq Computer Corporation, or other companies
listed or mentioned in this document.
COMPAQ 9th Annual System Engineering Conference – Living LAB
ECG05/0998
August 1998
3
INTRODUCTION
There are three demonstrations designed for the SE
conference living lab session.
Living Lab Demonstrations
Demo #1: OFS overview/simple transaction fail over
This demo illustrates a basic setup of the OFS
environment on Compaq servers, and how a simple query
reacts when the database server fails.
Demo #2: Automatically client re-connect
implementation
Through a sample C program, it shows how the
application program uses Oracle Call Interface(OCI) to
implement the re-connect feature.
Demo #3: OFS Performance characterization
This demo illustrates the OFS performance under various
number of users and system load.
OFS Overview
Oracle Fail Safe is the easy-to-use high availability
option for Oracle databases.
• Oracle® Fail Safe databases are highly available with
fast automatic failover capabilities. Oracle Fail Safe
server optimizes instance recovery time for Oracle
Fail Safe databases during planned and unplanned
outages.
• Oracle Fail Safe configuration and management is
straightforward with Oracle Fail Safe Manager, the
easy-to-use graphical user interface (GUI). This is
integrated with Oracle® Enterprise Manager for
comprehensive database administration.
• Oracle Fail Safe databases are configured within
virtual servers that allow client applications to
access Oracle Fail Safe databases at a given network
name at all times, regardless of the cluster node
hosting the database.
3. Partitioned Workload
The ‘partitioned workload’ is a variation of
‘active/active’ solution. The only difference is that one
server is setup as an application server where the other is
a database server.
Appendix B addresses answers to some of the frequently
asked questions of OFS, and the document was downloaded from the Oracle web site, http://www.oracle.com/.
Scope
The major intent of this tech note is to share the findings
of the performance characterization tests using the
‘active/active’ solution on Compaq servers. All tests were
performed in a controlled laboratory environment. The
goal was to establish a consistent test bed where all the
test results can be compared fairly. Therefore, the test
environment was not tuned for optimal performance. The
demo environment used at the 9th ASE conference is a
smaller scale of environment in the lab.
There are three major ways to deploy the OFS solutions.
1. Standby (Active/Passive)
The ‘standby’ solution is expensive since it requires one
additional server standing by to take over the existing
server in the event of failure. The ‘standby’ solution will
be able to provide the best response time since the backup
server is idle all the time.
2. Active/Active
The ‘active/active’ solution may be the most commonly
used solution since both servers can be utilized at all
times. Meanwhile they are both configured to back each
other up in the event of failure.
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
Software
Database Server 1
- Windows NT 4.0
MSCS
- Oracle 8.0.4; OFS
2.0
- Complex OLTP
workload;
- DB Size: ~8GB
Database Server 2
- Windows NT 4.0
MSCS
- Oracle 8.0.4; OFS
2.0
- AS3AP database
- DB Size: ~16GB
Database
Clients
- Windows
NT 4.0
August 1998
The overall test environment is illustrated as Figure 1 in
the next page. Appendix A documented 14 steps for
building a two-node OFS environment.
4
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
5
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
6
Test Methodology
Benchmark Factory 97 was the tool used to build the
workloads on both database servers. The tool also
provided the capability of simulating users from the
client system in the test environment, displaying real
time statistics, and collecting the test statistics for the
tests.
Benchmark Factory Overview
Benchmark Factory (BF) has two main components. The
first component is the Visual Control Center (VCC);
used to prepare, execute and evaluate the tests performed
by Benchmark Factory. The VCC module also holds the
data repository where all the test results are kept for
future analysis. The other component is the Agent. This
module can either be installed on the same machine as
the VCC or on one or more other machines. The agent
allows many users to login to the Server Under Test
(SUT) from any given client machine. In a typical test,
you would have many agents, each running on their own
client machine, and one VCC running on its own
machine.
Visual Control Center
The Visual Control center (VCC) is the main component
of Benchmark Factory 97. From here, you are able to set
up all of your tests as well as populate your test
databases. You can also control the other component of
the product, the Agent, from the VCC and watch the
statistics of the actual tests being performed in real time.
The Agent
The agent is the workhorse of the Benchmark Factory
product. It is the application that resides on the machines
that are to be used as the client machines for the tests.
This agent is installed from a special directory created
during the installation of the VCC. . Installation from
this directory makes the setup of the agent on a large
number of machines a very quick and simple task. Once
the agent is installed on the client machine, it
automatically attempts to connect up with the VCC and
is instantly ready to perform a test. The agent gets all its
instructions from the VCC, making it very simple to
control hundreds of machines running the agent module
from a single location.
Detail product information can be found on
http://www.benchmarkfactory.com/.
Client re-connect implementation
To fail over a database transaction requires special
implementation. The application program needs to take
most of the responsibility to handle the return code from
the database server and to make the appropriate response.
Oracle 8 Server on NT supplies interfaces, Oracle Call
Interface (OCI) , for the database application to handle
the fail over condition. It provides calls for applications
to be able to connect back to the database instance, after
the instance fails from one server to another. It also
provides the ability for resuming the position and
continuing the fetch operation after the fail over
completes. In the event of INSERT, UPDATE, or
DELETE situations, it will be the responsibility of the
application program to verify if the transaction had been
committed before fail over occurred using the return code
from the OCI calls. The application then determines if
the transaction needs to be re-issued after reconnecting to
the database instance.
TEST FINDINGS
-Overall performance was reasonably well
The overall performance in the OFS environment was
good when both the failing node and the surviving node
had a light or medium system load at the time the fail
over occurred.
-System load impacts performance
We experienced performance degradation when the
system load on the surviving node became very high after
taking over all the work from the failing node.
-Additional processors relieve system load
Adding additional processors to the surviving node
relieved the system load especially when the majority of
the work was CPU bound.
-Automatically client reconnect
The read-only transactions were handled by Oracle Call
Interface (OCI). When the fail over occurred, all users on
the failing node can be re-connected to the surviving
node automatically if the application program is
implemented properly. The read-only transaction could
also be resumed after it was re-connected to the surviving
node.
-Reasonable time to re-connect clients
When one node failed, the time to reconnect the clients
from the failing node to the surviving node is dependent
to the system load and the number of users.
When both system loads were light or medium, each user
on the failing node took 40 to 60 seconds to reconnect to
the surviving node. If the combined system load was over
90%, it took from 1 to 2 minutes.
The number of users required to reconnect to the
surviving node also impacted the reconnect time.
Comparing the reconnection of 20-user vs. 40-user cases,
the average time to reconnect each user in the 40-user
case was 15% longer than the 20-user case.
-Longer execution time for transactions experiencing
fail-over
The database transactions running on the surviving node
experienced very little impact when the system was not
heavily loaded after the fail over occurred.
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
7
The database transactions running on the failing node
took at least 1 minute longer to complete when the fail
over occurred. The additional time was used to reconnect
the user to the surviving node. It took longer when the
surviving node was heavily loaded. After the user was
Appendix A
Installation and Configuration of Oracle Fail
Safe 2.1 on Compaq ProLiant Servers for the
9th ASE Conference
This document outlines the fourteen steps required to setup
a two-node OFS system. These steps built the environment
below with the two servers being able to back each other up
in the event of system failure.
- Two ProLiant Servers
- Two Client System
- One Shared Fiber Storage System
The document assumes that readers are knowledgeable
about Compaq, Oracle Server on NT, Microsoft Windows
NT, and Microsoft Cluster Server Products.
Before step 1, ensure all physical connections between the
servers, clients, the fiber storage, and network links are all
properly connected.
Compaq High Availability web site will have the following
document available for a detail description of installation
and configuration of OFS on Compaq Servers. The
document is titled, ‘Implementing Oracle Failsafe 2.1.2 and
Microsoft Cluster Server on Compaq ProLiant Clusters’.
STEP 1: Update all firmware for both nodes to their
current versions.
q ProLiant Firmware. (We used version 04/29/98)
q Fiber Channel Storage Array Subsystem Firmware.
q Fibre Adapter Cards and Connectors.
q Smart2 Controller
q Hard drives
STEP 2: Run the Compaq SmartStart CD to both nodes.
(version 4.0)
q Run the system eraser utility.
q Run the system configuration utility.
q Set all cards to their appropriate IRQ setting. See the
Hardware Configuration layout in the Appendix.
q Use the ACU utility to partition the shared storage to its
appropriate RAID level and Logical Drives.
q Create a Compaq partition on the local drive.
STEP 3: Install Windows NT on both nodes. (We used
NT 4.0 Enterprise Edition)
reconnected to the surviving, it resumed processing the
transaction.
The details of the test results is documented in the DTSC
Lab Report, which will be available on Compaq web site
in the near future.
Note: Microsoft Cluster Server requires that all computers
in the cluster belong to the same domain. Login for each
node requires that both nodes have the same username,
password, and administrator privileges. The following steps
were performed for our demo setup configuration. If you
are installing a cluster into an existing domain
environment, it is not required to setup a PDC or BDC.
q Install server1 (node1) as the Primary Domain
Controller (PDC).
q Install server2 (node2) as the Backup Domain
Controller (BDC).
q Set virtual memory = 500MB (This step is optional).
q Set the foreground and background application as
equal.
q From the Control
PanelèNetworkèServicesèServerèProperty and
select “Maximize throughput for network applications”.
(This step is optional.)
q If the Installer brings up the nhloader, exit this screen.
This will be used later to install MSCS.
STEP 4: Install Service Pack 3 (SP3) to both nodes.
q Select “Yes” to overwrite the “cpqflxe.sys” file.
q Reboot both the machines.
Note: After installing SP3 to both nodes, use the Disk
Administrator to verify all default partitions and all drive
letters from server1 are the same as server2. We used drive
C for system and Oracle files and drive D for the CD ROM.
STEP 5: Run Compaq NTSSD from the SmartStart CD
to all required nodes.
q Go to \cpqsupportsw\ntssd and run setup.exe.
q Select “Express setup” unless there are overhead issues,
in which case only load the support drivers for Fibre
Channel, Network Card, SCSI, and Smart2 card.
q Invoke NT Disk Administrator to create all required
partitions on the shared storage. All drive letters from
server1 must be the same on server2. We used drive
C=system files and Oracle files, drive D=CD ROM,
drives E, F, and G as the shared disks.
q Reboot the machines.
Note: If you have NT Service Pack 4 (SP4), the Compaq
Fibre Channel Storage Driver is located in
\Drivlibs\storage\cpqcalm\i386 directory.
STEP 6: Create Windows NT hosts and lmhosts files for
the entire cluster.
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
8
Note: Lmhosts file is not needed if the cluster is within the
same subnet as all the clients. When modifying the lmhosts
file from the DOS prompt, refresh the cache with the
commands Nbstat-R and Nbstat-c.
q Locate the hosts and lmhosts files in
\winnt\system32\drivers\etc on server1.
q Add the computer’s name and the public
communication’s IP to the hosts and lmhosts files for
all nodes and clients which participate in the cluster.
(Failsafe requires that the hosts file includes the
computer’s name and IP.)
STEP 7: Check and Test the IP bindings for all required
nodes.
q At the dos prompt on each node, ping the computer’s
name to verify that the return IP address is correct for
“public communication”. If the IP address is incorrect,
correct the IP address for the network card.
STEP 8: Install Microsoft Cluster Server 1.0 (MSCS) on
both nodes.
q Click on “Start” è “Run” and at the “Open” box type
in “nhloader” and <enter>.
q Select “Continue” to continue the installation.
q Select “Microsoft Cluster Server” and “Start
Installation”.
q When prompted, insert disk 2 of Windows NT 4.0
Enterprise Edition in the CD-ROM drive and click
“OK” to continue.
q At the “Microsoft Cluster Server Setup” screen select
“Next” to continue, “I Agree”, then “Next”.
q Select one of the following three options:
a) Form a new cluster – select if this is the first
cluster installation in the configuration. Use this
option for the first server.
b) Join an existing cluster – select if a server is
joining a cluster that already exists. Use this
option for the second server. (You will see less
steps in the installation for this option.)
c) Install Cluster Administrator only – select if the
Cluster Administrator console, used to monitor the
cluster, is to be installed on a computer system.
q Enter the name for the cluster being formed or joined,
then click “Next”.
q Click “Next” for “Setup will place the cluster files in
this folder”.
q Add all available unshared disks to the right side as
shared cluster disks, then click “Next”.
q “Select the shared disk on which to store permanent
cluster files” screen allows you to select the Quorum
drive. The Quorum drive contains the cluster files,
keeping MSCS alive. Select a “shared” disk or use the
default, then click “Next” twice.
q Enter the “Network Name” that will be used for
network communication, and enable the network to be
used by the cluster as:
a) Use for all communications – select if this network
is to be used for all private and public
communications.
b) Use only for internal cluster communications –
select if the network is only used for the
“heartbeat” or private communication between the
two cluster nodes.
c) Use only for client access – select if the network is
only used for client or public communication.
q Click “Next”, enter the IP address and subnet mask for
the cluster, then click “Next” again.
q Click “Finish” to copy the files, then reboot the nodes.
STEP 9: Install Oracle on both nodes. (We used v.8.0.4.
Apply any necessary patches after the installation.)
Note: For additional help, refer to Oracle installation
documentation.
q Place the Oracle CD in the CD-ROM drive of the
primary server, server1.
q When the installation screen appears, click on “Begin
Installation”.
q At the “Oracle Installation Settings” screen, the
following should appear for the Oracle Home:
q Name=DEFAULT_HOME
q Location=C:\ORANT
q Language=ENGLISH
q Click “OK” if the above is correct.
q Select “Oracle8 Enterprise Edition” as the installation
option, then click “OK”.
q Select Cartridges and Options if purchased, else click
“OK”. (This document does not support options.)
q At “Select a Starter Database Configuration” screen,
select “None” and “OK”. (Select “Typical
Configuration” if you want a sample database to
automatically be created during the installation.)
q Select “Yes” if you want to install Legato Backup
Manager at the “Installation Legato Storage Manager”
screen, else click “NO”.
q Select whether to install Oracle Online Documentation,
which requires 75MB of hard disk space.
q Click “OK” at the “Reboot Needed” screen, close all
open windows and reboot the system.
q After reboot, confirm that the following services are
started by using the “Services” utility in the “Control
Panel”:
q OracleAgent80---------------started
q OracleTNSListener80------started
q OracleWebAssistant80-----started
q If any of the above services are not started, highlight
the service and click on the “Start” button located on
the right side of the screen.
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
9
STEP 10: Update both nodes with the latest patch for
Oracle. (We used v.8.0.4.1.3)
q Stop OracleTNSListener80 and OracleWebAssistant80
services before installing the patch.
q Run the setup.exe and install the patch.
STEP 11: Install Oracle Fail Safe for Oracle8 and Fail
Safe Manager to both nodes.
Note: Ensure that all required patches are installed. Verify
that the Microsoft Cluster server and both nodes are up and
running prior to installing Fail Safe.
q Place the Oracle Fail Safe CD in the CD-ROM drive
and click “Begin Installation”.
q Verify that the “Oracle Installation Settings” are the
same as in Step 9, then click “OK”.
q Click “Install” at the screen where Oracle Fail Safe
Server and Oracle Fail Safe Manager are highlighted.
q Select “Install the Fail Safe Documentation and Quick
Tour”.
q Select “Yes” to “Install Oracle Intelligent Agent
8.0.4.1.2”.
q Reboot the machines.
STEP 12: Create the Fail Safe Group (Virtual Server).
q Login to the Fail Safe Manager on server1.
q Select “Yes” when asked if you want to verify the
cluster.
q Click “OK” if you see a message stating that it could
not see the standalone databases. (This requires that the
Oracle Enterprise Manager be up and running, this step
will be performed later.)
q From the Fail Safe Manager, click on “Fail Safe
Groups” and select “Add”.
q Enter the Group name. This will be an alias in the
Microsoft Cluster Administrator for the resource group
that is automatically created.
q Select “Networks accessible by clients” and enter
“client connection” for the network to use, the network
name, IP address, and the subnet mask.
q Specify the “Failover Period” and “Failover
Threshold”, then click “Next”.
q Select a “Preferred Node” for the failback, else click
“next”.
q Click “OK” to confirm, and repeat the above steps for
server2.
q Verify the Fail Safe group and cluster by selecting the
“verify group” and “verify cluster” options from the
Fail Safe Manager menu. If either fails, delete and
rebuild them.
STEP 13: Create a Standalone Database.
A standalone database is a database that does not belong to
the Fail Safe Group. When a database already exists or is
created, it is a standalone database until it is moved to a Fail
Safe Group, then it becomes a Fail Safe Database. A
sample database can be created using the Fail Safe
Manager. When creating a custom database, place the
database on the shared disk and verify the database using
the “verify standalone database” option from the Fail Safe
Manager menu.
STEP 14: Add the Standalone Database to the Fail Safe
Group.
If a standalone database is on the shared disk, but does not
display as a GUI, use the Fail Safe Manager menu option
“verify standalone database”. After the verification is
successful, add the standalone database to the Fail Safe
Group by selecting “Add database to Fail Safe group” from
the Fail Safe Manager menu.
INSTALLATION TIPS
q Change Oracle Default Password
Some of the Oracle “create database” scripts use
“internal/oracle” or “internal/manager” as the password.
To change the default password to suit your needs, use the
orapwd80.exe utility located in the Orant\bin directory.
q Tnsname.ora
In the tnsname.ora file, modify the host name to the name
of the virtual computer used by the clients. Use the Fail
Safe “aware” tnsname.ora file, which can be found on the
Oracle Fail Safe CD, instead of the default tnsname.ora file.
q Windows NT Server Network Binding Order
When two network cards are in a system, Windows NT
should bind the IP address to the cards according to the
order in which they are specified in Control
PanelèNetworkèTCP/IPèProperty. To ensure that the
order is correct, before installing Microsoft Cluster Server,
check and test the binding order of the cards. To do this, at
the DOS prompt ping the computer names for the nodes
being used for public communication or client access. You
should see the network IP binding for the card used for
public communication. If you see the IP binding for the
heartbeat or private communication, switch the IP addresses
on the network cards.
q Remote Procedure Call (RPC) Locator Error for
MSCS
Sometimes when opening Microsoft Cluster Administrator,
an error message indicating that there is a problem with
RPC may appear. Go to Control PanelèServices and stop
the RPC locator and restart it again. A fix for this problem
is available from the Microsoft Web site, see cluster setup.
q Fiber Channel Storage Host Adapter Driver
If you have the Microsoft Windows NT Service Pack 4
(SP4) CD, the Compaq Fiber Channel Storage Host Adapter
Driver is located in the \Drvlibs\storage\cpqcalm\i386
directory.
q Recovery From a System Crash
After a crash, move the database to the surviving node, then
reload Microsoft Windows NT 4.0, Service Pack 3, MCSC,
Oracle 8.0.4, and Fail Safe 2.1.x onto the crashed system.
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
10
Move the crashed system’s database from the “Fail Safe
group” to a “standalone database”. Now, add this
standalone database back to the Fail Safe group. This
process will automatically reload the instances on both
nodes.
q Create or Add a Tablespace in the Fail Safe
Environment
When creating or adding a tablespace, remember to put the
data files on the desired shared disk. Verify the group by
using the Fail Safe Manager.
q Create a Custom Database
Use Oracle Database Assistant to create a typical or custom
database, locating the parameter, log, data, init, trace,
control, and redo log files onto the shared storage array
drives.
q Performance and Capacity Considerations
Create or use a simulator application tool to simulate the
number of users, workload, and type of applications being
used, before actually deploying your configuration into a
production environment.
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
Appendix B (Prepared by Oracle Corporation)
Oracle® Fail Safe 2.1
Frequently Asked Questions
®
Oracle Fail Safe
11
What is Oracle Fail Safe?
Oracle Fail Safe is the easy-to-use high availability
option for Oracle databases.
• Oracle® Fail Safe databases are highly available
with fast automatic failover capabilities. Oracle
Fail Safe server optimizes instance recovery
time for Oracle Fail Safe databases during
planned and unplanned outages.
• Oracle Fail Safe configuration and management
is straightforward with Oracle Fail Safe
Manager, the easy-to-use graphical user
interface (GUI). This is integrated with Oracle
Enterprise Manager for comprehensive database
administration.
• Oracle Fail Safe databases are configured within
virtual servers that allow client applications to
access Oracle Fail Safe databases at a given
network name at all times, regardless of the
cluster node hosting the database.
How do Oracle Fail Safe and Oracle Parallel
Server compare?
Oracle Fail SafeOracle Parallel Server
Targets departmental and
workgroup customers
seeking highly available
Windows NT cluster
solutions.
Provides highly available
single-instance
databases.
Targets enterprise-level and
corporate customers seeking
highly available and scalable
cluster solutions.
Provides scalable, highly
available multi-instance
parallel databases.
Oracle Fail SafeOracle Parallel Server
single-instance Oracle
database.
Maximum of 2 nodes in
a cluster; current
limitation is required by
underlying Microsoft
Cluster Server.
Oracle database and cluster
configuration knowledge.
Supports more than 2 nodes
in a cluster; does not use
Microsoft Cluster Server
software.
How do the Oracle Fail Safe releases differ?
• Oracle Fail Safe 2.1 supports both Oracle7
®
and Oracle8 databases.
• Oracle Fail Safe 2.0 supports only Oracle7
databases.
• Oracle Fail Safe 2.0 and 2.1 provide a software-
only solution for Windows NT clusters layered
over Microsoft Cluster Server, and feature
Oracle Fail Safe Manager to configure groups,
set conditions, and optimize failover and
failback of Oracle databases.
• Oracle Fail Safe 1.0 and Oracle Fail Safe 1.1 are
earlier custom hardware-based solutions that are
specific to Compaq ProLiant servers and require
the Compaq Online Recovery Server.
What network protocols and disk formats does
Oracle Fail Safe Release 2.1 support on Windows
NT?
• TCP/IP network protocol
• NTFS formatted disks on a shared storage
interconnect for the Oracle Fail Safe database
files
Windows NT only
solution.
Tuning and other
management operations
are the same as for a
Available on all major
platforms.
Tuning and other
management operations
require multi-instance
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
Oracle Fail Safe Manager
What is Oracle Fail Safe Manager?
August 1998
12
Oracle Fail Safe Manager is a configuration and
management GUI that includes wizards, drag-anddrop capabilities, online help, and tutorials to:
• Configure standalone Oracle databases into
Oracle Fail Safe databases
• Manage Oracle Fail Safe databases in clusters
• Monitor, verify, and help to load balance Oracle
Fail Safe databases
When do I use Oracle Fail Safe Manager and
Oracle Enterprise Manager tools?
• Use Oracle Fail Safe Manager for cluster-related
configuration and management operations of
Oracle Fail Safe databases.
• Use Oracle Enterprise Manager for routine
database administration tasks (such as database
backup and restore operations, or SGA analysis)
of Oracle Fail Safe databases.
How does Oracle Fail Safe Manager differ from
Microsoft Cluster Administrator?
• Oracle Fail Safe Manager and Oracle Fail Safe
server have built-in knowledge of Oracle
databases and their operation on Windows NT
clusters.
• Microsoft Cluster Administrator is not adequate
for configuration and management of Oracle
databases.
Client Applications
What changes do I need to make on the client to
access an Oracle Fail Safe database?
• No changes are required to the application code.
• On the client system, the virtual server address
for the Oracle Fail Safe database has to be
registered.
Do my applications connect using the cluster
alias?
• No. The cluster alias is used only for cluster
administration purposes by Microsoft Cluster
Administrator and Oracle Fail Safe Manager, as
shown in the figure below.
Client Database
Applications
Virtual Server Address for
Oracle Fail Safe Database
Oracle Fail Safe
Manager
Cluster Alias
• Applications connect to the Oracle Fail Safe
database using the virtual server address.
How does a failover affect database applications?
• The failover appears as a brief network outage.
• Client applications simply reconnect using the
virtual server address and continue processing
transactions against the Oracle Fail Safe
database.
Failover Timing
What are typical failover times?Oracle Fail Safe
server optimizes the time it takes to fail over an
Oracle Fail Safe database. Failover times vary and
depend on the following three parameters:
• Workload on the Fail Safe database server
• Failover policy information
• Cluster hardware configuration
For well-designed applications, failover for both
planned and unplanned outages occurs in seconds.
Product Availability
How is Oracle Fail Safe 2.1 packaged and
distributed?
The Oracle Fail Safe 2.1 production CD supports
Oracle8 release 8.0.4 and Oracle7 releases 7.3.3 and
7.3.4.
What additional software and hardware are
required for Oracle Fail Safe 2.1?
• Oracle Enterprise Manager Version 1.5
• Windows NT 4.0 Enterprise Edition with
Service
Pack 3, and Microsoft Cluster Server (MSCS)
• A Microsoft validated cluster configuration
Further Information
Where can I find more information?
• Oracle Fail Safe Product Data Sheet
• Oracle Fail Safe white papers
• Oracle Fail Safe Concepts and Administration
Guide
• Oracle Fail Safe Manager online help and
tutorial
•Oracle Fail Safe Quick Tour
Windows NT Cluster
COMPAQ 9th Annual Systems Engineering Conference
ECG05/0998
August 1998
13
Also, please visit the “Oracle for Windows NT
Clusters” site at http://www.oracle.com/ and click
on Clustering Solutions.
Oracle Corporation World Headquarters
500 Oracle Parkway
Redwood Shores, CA 94065
Oracle is a registered trademark of Oracle Corporation. Oracle7 and Oracle8 are
trademarks of Oracle Corporation. Windows NT is a trademark of Microsoft
Corporation. Compaq is a registered trademark of Compaq Computer
Corporation. All other company and product names mentioned are used for
identification purposes only, and may be trademarks of their respective owners.