SANRAD Application Note:

Storage Connectivity Comparison with SANRAD and

Oracle 10g Release 1
APP-011-01

Copyright SANRAD 2005

All rights reserved. The copyright and all intellectual property rights in this article belong to SANRAD. It is strictly forbidden to copy,
duplicate or otherwise use this article or any part thereof in any way shape or form without the prior written consent of SANRAD.
 Important Notice
The configurations described or tested in this application note are not the only available solution. This
application note is not intended nor may it be construed as an endorsement of any product(s) tested.
This application note provides no warranty of any kind. SANRAD limited warranties for SANRAD
products are stated in separate documentation accompanying each product. No damages or remedy
of any kind shall be recoverable by any party for any claim or loss in any way arising or alleged to
arise from or as a result of this application note, whether considered alone or in addition to any other
claim.

APP-011-01 Copyright SANRAD 2005

Table of Contents
SANRAD Storage System Solutions .............................................................1
General Recommendations:.............................................................................................1
The Storage Side (Disk Array):..................................................................................1
The Server(s) Side: ...................................................................................................2
The Network Side: .....................................................................................................2
The V-Switch Side: ....................................................................................................2
ORACLE 10g and SANRAD V-Switch Performance Study ...........................2
Test Method: ....................................................................................................................2
Test Environment: ............................................................................................................5
Test Results: .................................................................................................7
Summary.......................................................................................................8

i
APP-011-01 ii Copyright SANRAD 2005
SANRAD Storage System Solutions
This application note is intended for IT professionals considering iSCSI and SANRAD V-Switch as a
means for centralizing database storage for one to many Oracle 10g database servers. There are several
benefits to moving to a shared storage model, but with these benefits there are certain considerations
regarding what type of solution best fits the needs of the organization. Performance, guaranteed data
availability and future scalability costs are concerns that need to be addressed when placing Oracle
Database on a centralized network storage solution.
This application note assumes knowledge and experience with Oracle Database administration. It is also
recommended to familiarize yourself with V-Switch basic configuration and operation.
When planning the storage strategy for Oracle 10g, companies should design the Oracle storage based
on a combination of capacity, availability, RAID performance and cost to best suite their needs and
budget.

General Recommendations
In an iSCSI environment each component can contribute to improved capacity, availability and
performance.

The Storage Side (Disk Array)

The user should consider:
• Type of disk array: SCSI controller or FC controller as well as if disk array will have
one controller or two.
• Type of disks: SCSI or SATA, SCSI perform better but cost more, one SATA disk
can go up to 400 GB with 7200 RPM at the same price of one SCSI disk with 120 GB
and 15000 RPM
• Number of spindles: The number of disks in a LUN. The more the better for
performance even more important then the disk size.
• Disk array cache size: 256 MB, 512 MB or 1024 MB. SANARD recommends at
least 512 MB

For example, an FC disk array with SATA disks and at least 512 MB cache controller
is a good combination of performance and cost for up to 500 users.

APP-011-01 1 Copyright SANRAD 2005

The Server(s) Side
The user needs to decide between an iSCSI software initiator and iSCSI HBA hardware.
The advantage of the iSCSI HBA is that the TCP/IP offload engine within the HBA for
TCP checksum calculations takes the load off the server CPU. If CPU utilization is above
40% without the use of iSCSI, it is recommended to use an HBA to avoid server CPU
saturation when iSCSI is used.

Note: While the IO subsystems will impact the Oracle database performance it is
equally important to emphasize fine-tuning the database which can
improve performance dramatically. There are numerous fine-tuning guides
and articles available on the web for Oracle databases.

The Network Side

It is highly recommended to use a Gigabit network and, if possible, dedicated Ethernet subnet.

The V-Switch Side

For better performance it is recommended to dedicate port ETH2 or ETH3 (V-Switch 3000 only) on the V-
Switch for the database and transaction log volumes. Refer to the V-Switch User Manual for more
details.

ORACLE 10g and SANRAD V-Switch

Performance Study

Test Method
To test the performance of Oracle using SANRAD V-Switch 3000, we created a test environment (see
Figure 1) that included a Linux RedHat Enterprise Linux 3.0 server running Oracle 10g (release 10.1.0.3)
with a SANRAD V-Switch 3000 (firmware ver. 2.2.7.5) connected to a JMR FC RAID array. The disk
array was configured as JBOD with 4 available LUNs – one for each drive. On the V-Switch we allocated
a single-disk volume that was attached through iSCSI to the Oracle server. We also created a small FC-
attached environment by connecting the server using FC HBAs directly to the disk array so we could
compare the performance between the SANRAD iSCSI V-Switch and FC-attached storage. Oracle 10g
database was installed according to the guidelines provided in Oracle 10g Release 1 Quick Installation
Guide for Linux x86. (http://download.oracle.com/docs/html/B15521_01/toc.htm). After initial
installation and post-installation steps were carried out, a DBCA wizard application was used to create a
new instance of Oracle database on the target drive. Default settings were used whenever possible
throughout the installation, configuration and DB instance creation. The process was repeated for each
instance that was created for different storage architecture used in the test.

APP-011-01 2 Copyright SANRAD 2005

Performance was tested using the Quest Software Benchmark Factory software
(http://www.quest.com/benchmark_factory/index.asp). This utility creates objects necessary for
benchmarking within a designated tablespace and manages standard TPC (Transaction Processing
Performance Council) benchmarking and result reporting. The suite was installed on a separate IBM
eServer 305 machine which was connected to the Oracle database under test through GigE switch. Four
different benchmarks were run as a batch jobs – TPC-C, TPC-B, AS3AP and Scalable Hardware. Each
benchmark was run on each instance of the database corresponding to the storage connectivity method –
local disk, FC-attached, IP-SAN, IP-SAN w/iSCSI HBA. Results were logged and analyzed. An additional
set of benchmarks was re-run against the same database instance on a different machine to provide for a
control group in this test.

Note: The goal of the test was to achieve a relative performance comparison between
storage connectivity scenarios, rather than an absolute TPS (Transactions Per
Second) number possible. Therefore, the configuration that was used during the
test does not reflect an optimum configuration and setup designed to achieve the
highest TPS result.

Here’s a description of each benchmark from the TPC website

(http://www.tpc.org/information/benchmarks.asp):

TPC-C
The TPC-C benchmark measures on-line transaction processing (OLTP) workloads. It combines read-
only and update intensive transactions simulating the activities found in complex OLTP enterprise
environments. The TPC-C benchmark minimizes the diversity of operations found in a production
application and emphasizes the application's essential performance characteristics.
The TPC-C benchmark represents any company that sells, manages, or distributes products or services.
The TPC-C benchmark reports performance metrics in orders fully processed per minute. This is
expressed as tpmC.
The scale factor determines the amount of information initially loaded into the benchmark tables. For the
TPC-C benchmark, each scale factor represents one warehouse as per TPC-C specification. The TPC-C
benchmark involves a mix of five concurrent transactions of different types and complexity. The database
is comprised of nine tables with a wide range of records.
A maximum of 10 users should be run against each warehouse. For example, user loads of 1, 5, and 10,
set the scale to 1. If using other user load values, change the scale factor accordingly.

TPC-B

The TPC-B benchmark measures transaction throughput in terms of how many transaction per second
(TPS) a system can handle. The TPC-B benchmark can be compared to an electronic data processing
batch application that runs overnight when no customer users are logged on. The TPC-B benchmark is
characterized by:
• Significant disk input/output
• Moderate system and application execution time
• Transaction integrity

APP-011-01 3 Copyright SANRAD 2005

AS3AP
The AS3AP benchmark is an American National Standards Institute (ANSI) Structured Query Language
(SQL) relational database benchmark. The AS3AP benchmark provides the following features:
• Tests database processing power
• Built-in scalability and portability that tests a broad range of database systems
• Minimizes effort in implementing and running benchmark tests
• Provides a uniform metric and straight-forward interpretation of benchmark results
Systems tested with the AS3AP benchmark must support common data types and provide a complete
relational interface with basic integrity, consistency, and recovery mechanisms. The AS3AP tests systems
ranging from a single-user microcomputer Database Management System (DBMS) to a high-performance
parallel or distributed database.

Scalable Hardware
The Scalable Hardware benchmark measures relational database systems. This benchmark is a subset
of the AS3AP benchmark and tests the following:
• CPU
• Disk
• Network
• Any combination of the above three entities
The scale factor determines the amount of information initially loaded into the benchmark tables. For the
Scalable Hardware benchmark, each scale factor represents one user accessing the system. Two tables
are created in the database, and they are loaded with a varying number of rows.
For each virtual user, a separate set of data must be created. Therefore the scale factor used when
loading the database should be the size of the maximum user load. For example, with user loads of 1, 5,
and 10, a scale factor of 10 should be used.
The Scalable Hardware benchmark has a scaling factor of one.

APP-011-01 4 Copyright SANRAD 2005

Test Environment
Table 1. Oracle 10g Server -
Model DELL PowerEdge 2650
Processor 2 X 3.06 GHz Intel Xeon
Memory 2 GB
Network Card 2xBroadcom NetXtreme Gigabit Ethernet
FC HBA Qlogic 2312 2Gb FC HBA
Latest drivers and firmware from
www.qlogic.com
Internal SCSI Controller Dell PERC 3/Di – simple volume container
Operation System RedHat EL 3.0
Kernel 2.4.21-20ELsmp
iSCSI Initiator linux-iscsi 3.4.4
Application Oracle 10g release 10.1.0.3
iSCSI HBA Qlogic QLA4010C

Table 2. V-Switch Setup

Model V-Switch 3000
Firmware Version 2.2.7.5
StoragePro Version 2.2.7.5
Storage Connection Type Fiber Channel (FC)

Table 3. Storage Subsystem Setup –

Make JMR
Model STORBLADE RAID (FC)
Cache 512MB
Disks 4x34 GB FC
Raid Capability 0, 0+1, 1, 5, 50, JBOD
Interfaces 4x2GB FC

Table 4. Benchmark System Setup

Model IBM eServer 305
Processor 1 X 1.0 GHz Intel Pentium 4
Memory 1 GB
Network Card 2xBroadcom NetXtreme Gigabit Ethernet

Operating System Microsoft Windows 2003 Server Ent. Edition

Service Pack 1 + Hotfixes

Application Quest Benchmark Suite

Oracle 10g SQL Client

APP-011-01 5 Copyright SANRAD 2005

 iSCSI-Software-attached
Database Instance
iSCSI-HBA-attached
Database Instance

FC-attached
Database Instance

Benchmark Suite Host

iSCSI/Ethernet

Local Disk
Database instance

Oracle10g-RHEL3.0x86

Figure 1. Test Setup

APP-011-01 6 Copyright SANRAD 2005

 Test Results
The following tables and graphs describe the performance test results (higher numbers are good).

Table 5. Test Results (Blue numbers=lowest value, Red=highest)

TPC Benchmarks

STORAGE TPC-B TPC-C AS3AP

CONNECTIVITY

Local Disk 103.1 1.04 0.01

Fibre Channel 314.8 1.04 0.02

iSCSI-HBA 328.9 1.04 0.02

iSCSI-Software 223.6 1.03 0.02

Table 6. Test results continued (Blue numbers=lowest value, Red=highest)

Scalable Hardware
Benchmark

STORAGE CPU- INSERT- READ- UPDATE- TRANSACTION

CONNECTIVITY INTENSIVE INTENSIVE INTENSIVE INTENSIVE MIX

Local Disk
4553 210.6 3692 277.5 313.6
Fibre Channel
4510 199.9 3712 277.1 318.4
iSCSI-HBA
4509 154.8 4144 248.5 247
iSCSI-Software
3424 118.5 3683 247.5 290.1

APP-011-01 7 Copyright SANRAD 2005

 TPC-B
(measured in TPS)

iSCSI-Software

iSCSI-HBA

Fibre Channel

Local Disk

0 50 100 150 200 250 300 350

Chart 1 – TPC-B results

5000
4500
4000
3500
3000
Local Disk
2500 Fibre Channel
2000 iSCSI-HBA
1500 iSCSI-Software
1000
500
0
CPU- INSERT- READ- UPDATE- MIX
INTENSIVE INTENSIVE INTENSIVE INTENSIVE

Chart 2 – Scalable Hardware results

Summary
The test results show that Sanrad V-Switch iSCSI solution offers a cost-effective alternative to
FC-attached storage. The difference in the numbers between the iSCSI solution and the FC solution is
negligible, on average. The results achieved with connecting the database instance through iSCSI-HBA
are similar or better to the numbers of FC-attached instance. Therefore, the SANRAD V-Switch can be
used for as a viable solution for I/O intensive OLTP applications such as Oracle 10g Database.

APP-011-01 8 Copyright SANRAD 2005