Linux/Intel Effectiveness For Running E-Business: A Case Study of Oracle E-Business Outsourcing
An Oracle White Paper January 2003

Linux/Intel Effectiveness For Running E-Business: A Case Study of Oracle E-Business Outsourcing

Executive Overview ............................................................................................. 3 Introduction .......................................................................................................... 3 Why Linux? ........................................................................................................... 4 Price/Performance Measurement of Intel Architecture ................................. 4 UNBREAKABLE LINUX ................................................................................ 4 Architecture for E-Business outsourcing.......................................................... 5 Oracle E-Business Outsourcing Standard Infrastructure .............................. 5 effectiveness of Linux/Intel for outsourcing ................................................... 6 Reliability ............................................................................................................... 7 Reliability Advantages of Oracle E-Business Suite on Linux/Intel and RISC ........................................................................................................................7 Performance ......................................................................................................... 8 Production Performance of Linux/Intel for Oracle Outsourcing .................8 Data Collection Method and Tools ....................................................................8 Workload Characterization ..................................................................................9 Production Workload Study Findings ................................................................9 Oracle Outsourcing Performance Strategy ..................................................... 10 Availability............................................................................................................ 10 Manageability ...................................................................................................... 12 Conclusion ........................................................................................................... 12 Comparative Advantages of Linux/Intel Over RISC ................................... 12 REFERENCES .................................................................................................. 13 APPENDIX-A Performance Data .................................................................. 14 Workload Throughput Efficiency.................................................................... 14 Instance Efficiency ............................................................................................. 14 APPENDIX-B Oracle E-Business Outsourcing Architecture .................. 16 Appendix-c LinuX/Intel and RISC Environments ...................................... 17 Linux/Intel Environment.................................................................................. 17 RISC Environment ............................................................................................. 17 Oracle Applications Standard Benchmark ...................................................... 17

Linux/Intel Effectiveness For Running E-Business: A Case Study of Oracle E-Business Outsourcing

EXECUTIVE OVERVIEW

Linux/Intel provides the cost effective basis for the delivery of a successful EBusiness Outsourcing solution from Oracle. The key to delivering a low-cost solution for Oracle E-Business Suite Outsourcing has been the Linux deployment on inexpensive, high-performance hardware. The Linux deployment adds to the efficiencies of outsourcing initiative with high return on investment, and drastically reduced total cost of ownership. The following are the objectives of this paper:  Why Oracle chose Linux/Intel Architecture as the standard building block for its outsourcing practice.  Describe the current Outsourcing E-Business Suite deployment using the EBusiness Standard Infrastructure architecture  Outline the criteria used to determine the effectiveness of the Oracle E-Business Outsourcing Service using recent studies from production Outsourcing customers
INTRODUCTION

Oracle is leading the industry in recognizing the value of low cost deployment on Linux for Oracle E-Business Suite Outsourcing. Linux allows Oracle to deliver software-as-a-service with a lower total cost of ownership (TCO) than services deployed on other platforms. Linux enables Oracle to continue as a market leader in delivering a reliable solution with solid performance. This paper provides an insight into the Oracle commitment to Linux/Intel (Linux deployment on Intel), and to the growth of Linux into an enterprise-strength deployment platform. In addition, the importance of the E-Business Standard Infrastructure architecture approach will be discussed as the basis of the effectiveness of the Oracle Outsourcing Service, together with other important criteria such as reliability, performance, availability and manageability.
WHY LINUX?

In late 2001 the outsourcing architecture group at Oracle began conducting experiments to assess the effectiveness of an Intel Architecture computer running the

Linux operating system for deployment in outsourcing. At the time the staple of the outsourcing practice had been 4 CPU by 4 Gigabyte memory RISC architecture machines dedicated to a customer, or large multi-CPU RISC machines shared by several customers. The observed reliability and performance of Linux that had previously been deployed at Oracle for ad-hoc applications, combined with the rapid rate of performance improvement of Intel Architecture processors was the motivating factor behind this evaluation.
Price/Performance Measurement of Intel Architecture

The Intel architecture provides the low cost commodity components and competitive flexibility of servers. This architecture allows unparalleled price performance together with the advantages of outsourcing for Oracle E-Business customers. In the study the main metrics that where used were CPU speed, the number of CPUs, the amount of memory available on the computer, and the price of the computer. One attraction of the Intel based architecture was that because of the much lower price one could afford to add degrees of hardware redundancy. Likewise, the replacement costs would be much lower. The following table highlights the most relevant results of the study..

Platform

Number of CPU per box 2 2

CPU Speed 2.4 GHz

Physical Memory 6 GB

Average Response time (sec)1 1.3 1.7

90% Response time (sec)2 2.2 3.8

List price

Linux RISC

$8860 $53330

750 MHz 8 GB

NOTE: 90% percentile refers to transactions run faster than all transactions.
UNBREAKABLE LINUX

Oracle has made Linux unbreakable by initiating and continuing to support the following initiatives with cooperation from Red Hat:  Move Linux further along the maturity curve with the addition of Enterprise class features  Provide World Wide Support such as front line support for the technology stack and relief patches  Bring Oracle Real Applications Clusters (RAC) to Linux  Deploy Linux within Oracle IT
Measured based on Oracle Application Standard Benchmark (OASB) transactions for 251 concurrent users.
1 2

Based on the studies performed today Linux/Intel Architecture is the standard building block for various other applications at Oracle. Examples are:  Infrastructure for Oracle’s main website www.oracle.com  Oracle Application Demonstration Services (ADS), the infrastructure used by the Oracle Applications sales force for demonstrating Oracle e-business suite capabilities  Oracle Education  Oracle Files Online, the centralized file sharing and collaboration facility at Oracle.  Mid-tier deployment of Oracle’s internal global, single instance of E-Business Suite
ARCHITECTURE FOR E-BUSINESS OUTSOURCING

The architecture for Oracle E-Business Suite Outsourcing deployment is based on the Oracle E-Business Standard Infrastructure. It is composed of a set of tested elements known to deliver high reliability, optimal performance and availability. The elements of the technology stack include the following:  Hardware  Operating system  Oracle database  Oracle E-Business Suite In addition, Oracle E-Business Suite has pre-integrated and pre-configured elements, with tools to support standard service operations.
Oracle E-Business Outsourcing Standard Infrastructure

The elements of the Oracle E-Business Outsourcing Standard Infrastructure are described in the following table. Technology Stack Element Hardware Operating System Oracle Stack Pre-integrated Configuration Details Dell hardware standard configuration Red Hat Advanced Server 2.1 Linux Oracle Server 8.1.x, and Oracle E-Business Suite 11.5.x Oracle Discoverer Oracle Portal Oracle Express Server Oracle Enterprise Data Warehouse

Technology Stack Element Pre-configured

Configuration Details Production, Test, and Dev instances Standard disk layout, and naming conventions 230 GB (90 GB Production), 60 GB (Test), 60 GB (Dev), and 20 GB overhead

Storage

@Oracle model Network Attached Storage (NAS, NFS protocol) @Customer model Local attached storage

Network Virtual Private Network (VPN) technology System Initialization Deployment Architectures Cloning tools used to standardize deployment Single-Tier (General Business) 15 in Production 45 (various stages of pre-production) Split-Tier (for mid-tier scalability) 3 customers @Oracle 1 customer @Customer User Concurrency Single-Tier Up to 300 users (approx. 3,000 named users) Split-Tier for scalability 300-1,000 users (approx. 10,000 named users)

EFFECTIVENESS OF LINUX/INTEL FOR OUTSOURCING

The delivery of the Oracle Outsourcing Service has been derived as a result of the EBusiness Outsourcing Standard Infrastructure architecture approach. The criteria used to determine the effectiveness of Oracle Outsourcing Service delivery includes the following:  Reliability  Performance  Availability

 Manageability These criteria will be discussed with reference to recent studies of Oracle E-Business Suite Outsourcing production customers.
Reliability

Reliability of the Oracle E-Business Outsourcing Standard Infrastructure (hardware, infrastructure, database and application) is the result of the collaborative effort of Oracle, Dell and Red Hat. Reliability of these components is the foundation for the effectiveness of Oracle EBusiness Outsourcing on Linux/Intel. This reliability enables the remaining criteria of performance, availability and manageability to be viable.
Reliability Advantages of Oracle E-Business Suite on Linux/Intel and RISC

Unbreakable Linux/Intel Collaborative partnership with Oracle/Dell/Red Hat to ascertain stack reliability Oracle-Linux development team dedicated to OS platform Oracle provides front-line, holistic and rapid support for Linux/Intel and Oracle issues

RISC  Oracle and RISC vendor relationship to ascertain stack reliability  Oracle develops on the RISC platform (database and applications)  Oracle provides support for Oracle issues.

To quantify the reliability of Linux/Intel and the RISC platforms, the following metrics were used: Hardware/Operating System Metrics:  Hardware replacements required  Hardware-related Technical Assistance Requests (TARs)  Operating system-related TARs  Patching history for the platform (collectively) Software Metrics:  Software- and hardware-related TARs  Database- and Oracle E-Business Suite-related TARs  Patching history for the applications  QA/system specific reports from hardware vendors

The reliability of the Linux/Intel platform and Oracle’s E-Business Suite is the basis for the high availability enjoyed by the E-Business Suite Outsourcing customers. The results indicate the reliability between Linux/Intel and RISC are comparable. Both platforms display reliability and stability necessary for critical business applications. The advantages of Linux/Intel over RISC are the cost benefits and performance.
Performance

Effective server performance measurement for Oracle E-Business Outsourcing customer systems is determined by the overall throughput of the characterized workload such as server, database and Oracle E-Business Suite. From a server perspective, the metrics include server resource usage such as CPU utilization, memory and I/O. From a database and E-Business Suite perspective, the metrics include application workload throughput, response time, and instance efficiency. User workload activity then determines the overall ability of the system to perform at peak time.
Production Performance of Linux/Intel for Oracle Outsourcing

Based on a recent study of selected Oracle Outsourcing E-Business production customers, Linux/Intel was found to deliver the most cost effective performance solution. For more information concerning performance data, refer to Appendix A of this paper. The criteria used to determine the cost effectiveness of the Linux/Intel solution were as follows:  Workload throughput efficiency (database and server)  Instance (database efficiency)  Workload characterization  Response time
Data Collection Method and Tools

Daily collection of data during business hours of the customer time zone  Database-centric data collection: Oracle Statspack  Enterprise – wide service delivery data: Oracle IT system management portals  Operating system-related data: operating system tools and utilities.
Workload Characterization

The workload profiles for the Outsourcing Production customers studied were as follows:  E-Business Suite interactive users: Forms and Reports

 E-Business Suite batch processing: GL journaling
Production Workload Study Findings

Results from this study, indicate the overall workload throughput and instance efficiency of the Oracle E-Business Outsourcing Linux/Intel and RISC servers to be high performers. The systems are well tuned, and no stress of system resource utilization such as disk, memory or CPU, could be evidenced at medium to high load samplings. Workload efficiency is the measure of the E-Business Suite Application performance. Physical reads and writes per second, executions per second, and transactions per second are impressive when you consider the comparative cost differences between Linux/Intel and RISC. Highly efficient CPU utilization ranging from 60% to 85%, was found across both types of servers. Instance efficiency is the measure of database behavior. Efficient database behavior was evidenced by high buffer hit rates and high in-memory sorts, regardless of the activity (interactive or batch). No waits on redo indicate efficient processing throughput and efficient writes to disk. Overall, this is indicative of well-tuned disk I/O. The response times of the Linux and RISC servers were measured using a ping facility to measure Oracle E-Business Suite activity timed for 10 minute intervals. FND-Web.ping was used to report core information about the Applications database server roundtrips. The period of response time gathering was simultaneous with the workload throughput efficiency and instance efficiency result measurements. System Linux/Intel Oracle EBusiness Suite RISC Oracle E-Business Suite Mean Average Response Peak Average Response Time (sec) Time (sec) 0.11 0.23 0.18 0.23

The preceding results were collected during the business peak trading hours. Response times were collected from a centralized monitoring location, not necessarily at the customer’s site. On closer examination of the results per customer, for each there are minimal peaks in load for Linux and RISC, which dissipate rapidly, to resume fast round trip results. Overall, the response times are acceptable when considering the overall activity of the system, the efficiency of the workload throughput and that the systems are production status, with standard logging and recovery procedures in place.

Oracle Outsourcing Performance Strategy

Initial workload characterization is performed to indicate the workload requirements. As Oracle E-Business Suite systems are used, the systems are re-evaluated for the following:  Established usage patterns  Increased customer requirements  Data volume increases  Resource requirements changes Workload characterization is the key to the proactive performance management strategy that is currently available to E-Business Suite Outsourcing customers. The proactive strategies currently in place are depicted in the following table.

Management Component Performance Strategy

Details Outsourcing Management System including data collection (repository), reporting and publishing facility Oracle Statspack, Enterprise Manager, operating system utilities and tools AIM (Application Implementation Method) sizing and capacity planning tool

Performance Monitoring Capacity Planning

Availability

Service availability is the result of effective reliability and end-to-end system performance. The metrics used to determine availability of service for Oracle’s EBusiness Outsourcing customers are:  Uptime  Scheduled outages  Unscheduled outages These metrics are computed using the various management tools utilized by Oracle IT, and fed into highly informative portals, refreshed in real-time to provide 24x7 service to Oracle IT personnel and Outsourcing customers.

Platform Availability for Oracle Outsourcing Oct-Dec 2002
100

Percentage

90

Linux RISC

80 Week 2 Week 3 Time Week 4 Week 12

The preceding graph shows the availability of a sample of Linux/Intel and RISC based customers for the last calendar quarter. Criteria for customer system sampling included: · · · · Medium to high load Different industries Varied production histories (growth) Different time zones

Despite high workloads and scheduled outages, platform robustness is the main contributing factor to the high availability of the Linux/Intel platform. The data was collected from an equal number of systems of each type and percentages derived across the platform type. In the majority of the cases for the Linux/Intel platform servers, the data was predominantly in the 99 to 100 % range, with few exceptions. The RISC server data did have more variation in collected data, however, the same collection criteria was used for comparison. The availability percentage measurement, was created using a summation of the individual critical components of the Oracle E-Business Suite solution. These components were the following:  Concurrent Manager  Database  Listener  Web server The availability measurements include scheduled and unscheduled availability. Both the Linux/Intel and RISC based customers enjoy a high level of availability, with rapid resolution for unscheduled outages.

Linux/Intel customers enjoy availability at a substantially lower cost than RISC-based customers. Also, when unscheduled outages do occur, Linux/Intel- based customers have reduced turnaround time to re-established service, due to the support to the combined pre-configured servers used within Oracle’s E-Business Outsourcing Service.
Manageability

Effective systems management of an enterprise IT organization, with the inherent costs and overhead can be prohibitive. The growing burden of operational cost, training, hardware and software upgrades and maintenance for IT, is now not an option for many organizations, and alternatives are being explored. Indeed, the cost of an organization operating it’s own IT from an applications (ERP) perspective, can be as much as 2 to 3 times the cost of software alone. Many organizations are turning to outsourcing as a means to drastically reduce cost. Management of Oracle E-Business Outsourcing service for Linux/Intel and RISC is undertaken using the same management strategies, processes, procedures, and expertise. Both platforms are managed by Oracle without additional costs for the customer.
CONCLUSION

“We’ll be running our whole business on Linux.” -Oracle Chairman and CEO Larry Ellison

As outlined in the following table, Linux/Intel is the cost effective solution to the success of the Oracle E-Business Outsourcing Service. The Oracle commitment to the Linux/Intel direction enterprise-wide, has been delineated by example of extensive deployment internally, and continues to do so, moving forward.
Comparative Advantages of Linux/Intel Over RISC

Linux/Intel Low cost commodity servers Fewer, faster CPUs (horizontal scaling)

RISC High cost proprietary servers Slower and greater number of CPUs (vertical scaling)

Linux/Intel

RISC

Fewer GB of physical memory required Greater number GB of physical to achieve high throughput when running memory required to achieve high Oracle E-Business Suite workloads throughput when running comparable Oracle E-Business Suite workloads Low maintenance costs: warranty includes replacement of parts with maintenance, at the time of server purchase Server delivered, pre-configured Higher maintenance costs: Maintenance is not included in the purchase price Server separately purchased, full setup required

High reliability and availability at lower High reliability and availability at server maintenance cost ($20K per year) Low software licensing costs (no per processor charges) High licensing costs (per server and per processor)

Software patch turnaround very rapid due Software patching turnaround latency to Open Source community input possible Rapid support and replacement of units Slower support and replacement of units High performance at lower cost High performance at high cost

High scalability at lower cost (addition of High scalability at higher cost (if servers to mid tier) addition of RISC to mid-tier) High scalability and reliability at lower cost for RAC with Linux/Intel High scalability and reliability at higher cost for RAC with RISC

Oracle will continue to deliver the highest quality of service for its Oracle E-Business Outsourcing Service customers. Oracle continues to streamline its internal processes to deliver and meet rigorous industry standards.
REFERENCES

 Oracle E-Business Suite Outsourcing, Oracle White Paper, March 2002.  Linux/Intel Effectiveness as a Scalable Database Server: A Case Study of Lithonia Lighting E-Business Suite Implementation, January 2003  Linux/Intel Effectiveness as a Mid –Tier Application Server: A Case Study of Oracle’s single Global Instance E-Business Suite Implementation, January 2003

 Total Cost of Ownership for Linux in the Enterprise, Robert Frances Group, July 2002

Linux Market Growth
12 10 8

Million

6 4 2 0 1999 2001 2003 2005

APPENDIX-A PERFORMANCE DATA
Workload Throughput Efficiency

Data collection per database, averaged per platform: Workload Profile Redo size Log reads Phys reads Phys writes User calls Executes Trans Linux/Intel (/sec) 81660.9 22578.5 1259.8 54.2 45.8 188.2 1.4 RISC (/sec) 66093.3 17350.9 1157.3 36.1 91.4 233.1 1.5 Linux/Intel(/t RISC rans) (/trans) 52051.1 31999.2 866.5 28.7 36.8 142.6 41589.4 9153.6 603.8 19.4 59.1 148.5

Data collection per server, averaged per platform: Workload Profile Linux/Intel Mean Average/ Peak Average OS-CPU Utilization OS Response Time
Instance Efficiency

RISC Mean Average/ Peak Average 60/80 (%) 0.23 sec/.23 sec

60/85 (%) 0.11 sec/0.18 sec

Data collection per database, averaged per platform:

Instance Efficiency (%) Buffer Nowait Buffer Hit In-Memory Sort Latch Hit Redo Nowait

Linux/Intel 99.91 100 99.99 99.89 99.50

RISC 99.97 96.87 99.97 99.88 99.43

APPENDIX-B ORACLE E-BUSINESS OUTSOURCING ARCHITECTURE

The Oracle E-Business Outsourcing Standard Infrastructure delineates the building blocks from the principal vendors.
Linux/Intel e-Business Outsourcing Architecture Common Architecture (One customer) Customer Intranet Zone A Customer Control Zone

Internet

Zone B Oracle Networks Zone

Oracle VPN/Firewall

Hosting DMZ Router

BigIP Primary BigIP Secondary Zone C Oracle Hosting Operations Control Zone

Dell Database/Application Tier

Tape Server

Tape Silo

NAS Storage Clusters

APPENDIX-C LINUX/INTEL AND RISC ENVIRONMENTS

Linux/Intel Environment

    

Dell PowerEdge 2650 ($8860) 2 CPUs @2.4GHz w/512K L2 cache 6GB memory@400MHz 2x36GB disks Red Hat 2.1 Advanced Server

RISC Environment

    

Sun SunBlade 750 ($53,330) 2 CPUs @750MHz w/8MB L2 cache 8GB memory 6x36GB disks Solaris 8

Oracle Applications Standard Benchmark

Standard Oracle Applications Standard Benchmark (OASB) was conducted for Oracle internal usage. The following are exceptions to the conduct of the benchmark tests:  2 out of 22 transactions not working.  OE Insert Order (working in base).  PO View Purchase Order (bug 1767346).  Rebuilt workload with 20 working transactions.

Effectiveness of Linux/Intel in Oracle E-Business Outsourcing Service January 2003 Author: Josie De Marco Contributing Authors: Brom Mahbod, Kurt Graustein, Bill Weils, David Wong Oracle Corporation World Headquarters 500 Oracle Parkway Redwood Shores, CA 94065 U.S.A. Worldwide Inquiries: Phone: +1.650.506.7000 Fax: +1.650.506.7200 www.oracle.com Oracle is a registered trademark of Oracle Corporation. Various product and service names referenced herein may be trademarks of Oracle Corporation. All other product and service names mentioned may be trademarks of their respective owners. Copyright © 2003 Oracle Corporation All rights reserved.