Proven Performance for Accenture Duck Creek Policy Administration Commercial Lines



Similar documents
Proven Performance for Accenture Duck Creek Policy Administration Commercial Lines

Benchmark Testing Results: CSC s Integral Life Insurance Software Running on SQL Server

Benchmark Testing Results: Teamcenter Running on SQL Server 2012 and Servers Based on Intel Xeon Processors

SQL Server Consolidation Using Cisco Unified Computing System and Microsoft Hyper-V

Benchmark Testing Results: TEMENOS T24 Running on Microsoft SQL Server 2012

Oracle Database Scalability in VMware ESX VMware ESX 3.5

HP reference configuration for entry-level SAS Grid Manager solutions

Benchmark Testing Results: OpenText Monitoring and Records Management Running on SQL Server 2012

Benchmark Testing Results: Unparalleled Scalability of Itron Enterprise Edition on SQL Server

Analyzing the Virtualization Deployment Advantages of Two- and Four-Socket Server Platforms

Dell Virtualization Solution for Microsoft SQL Server 2012 using PowerEdge R820

Performance characterization report for Microsoft Hyper-V R2 on HP StorageWorks P4500 SAN storage

Benchmark Testing Results: Telvent ArcFM Solution Running on Microsoft SQL Server

Private cloud computing advances

Dassault Systèmes ENOVIA V6 Running on SQL Server 2008 R2 Enterprise

Evaluation Report: HP Blade Server and HP MSA 16GFC Storage Evaluation

Maximum performance, minimal risk for data warehousing

SMB Direct for SQL Server and Private Cloud

Microsoft SQL Server 2008 R2 Enterprise Edition and Microsoft SharePoint Server 2010

OPTIMIZING SERVER VIRTUALIZATION

HP ProLiant BL660c Gen9 and Microsoft SQL Server 2014 technical brief

The Methodology Behind the Dell SQL Server Advisor Tool

Scalability. Microsoft Dynamics GP Benchmark Performance: Advantages of Microsoft SQL Server 2008 with Compression.

Optimizing SQL Server Storage Performance with the PowerEdge R720

Microsoft Windows Server 2003 with Internet Information Services (IIS) 6.0 vs. Linux Competitive Web Server Performance Comparison

EMC Backup and Recovery for Microsoft SQL Server

Dell Microsoft Business Intelligence and Data Warehousing Reference Configuration Performance Results Phase III

How To Write An Article On An Hp Appsystem For Spera Hana

System Requirements Table of contents

7 Real Benefits of a Virtual Infrastructure

Protect SQL Server 2012 AlwaysOn Availability Group with Hitachi Application Protector

Understanding the Benefits of IBM SPSS Statistics Server

Microsoft SQL Server on Stratus ftserver Systems

High Performance SQL Server with Storage Center 6.4 All Flash Array

White Paper. Recording Server Virtualization

System Requirements for Microsoft Dynamics GP 2015

Removing Performance Bottlenecks in Databases with Red Hat Enterprise Linux and Violin Memory Flash Storage Arrays. Red Hat Performance Engineering

EMC Backup and Recovery for Microsoft SQL Server

HP high availability solutions for Microsoft SQL Server Fast Track Data Warehouse using SQL Server 2012 failover clustering

SAS Business Analytics. Base SAS for SAS 9.2

SQL Server What s New? Christopher Speer. Technology Solution Specialist (SQL Server, BizTalk Server, Power BI, Azure) v-cspeer@microsoft.

Accelerate SQL Server 2014 AlwaysOn Availability Groups with Seagate. Nytro Flash Accelerator Cards

PRODUCT BRIEF 3E PERFORMANCE BENCHMARKS LOAD AND SCALABILITY TESTING

Kronos Workforce Central 6.1 with Microsoft SQL Server: Performance and Scalability for the Enterprise

The Advantages of Multi-Port Network Adapters in an SWsoft Virtual Environment

DELL TM PowerEdge TM T Mailbox Resiliency Exchange 2010 Storage Solution

HP SN1000E 16 Gb Fibre Channel HBA Evaluation

Comparing Multi-Core Processors for Server Virtualization

Accelerating Microsoft Exchange Servers with I/O Caching

PSAM, NEC PCIe SSD Appliance for Microsoft SQL Server (Reference Architecture) September 11 th, 2014 NEC Corporation

HP ConvergedSystem 900 for SAP HANA Scale-up solution architecture

HP ProLiant DL580 Gen8 and HP LE PCIe Workload WHITE PAPER Accelerator 90TB Microsoft SQL Server Data Warehouse Fast Track Reference Architecture

Seradex White Paper. Focus on these points for optimizing the performance of a Seradex ERP SQL database:

Windows Server 2008 R2 Hyper-V Live Migration

Brocade and EMC Solution for Microsoft Hyper-V and SharePoint Clusters

Database Server Configuration Best Practices for Aras Innovator 10

EMC Business Continuity for Microsoft SQL Server 2008

Integrated Grid Solutions. and Greenplum

DIABLO TECHNOLOGIES MEMORY CHANNEL STORAGE AND VMWARE VIRTUAL SAN : VDI ACCELERATION

EMC Unified Storage for Microsoft SQL Server 2008

Microsoft Dynamics CRM 2011 Guide to features and requirements

How To Connect Virtual Fibre Channel To A Virtual Box On A Hyperv Virtual Machine

Cisco, Citrix, Microsoft, and NetApp Deliver Simplified High-Performance Infrastructure for Virtual Desktops

PERFORMANCE AND SCALABILITY

EMC XtremSF: Delivering Next Generation Storage Performance for SQL Server

PERFORMANCE AND SCALABILITY

Hitachi Virtage Embedded Virtualization Hitachi BladeSymphony 10U

Microsoft SQL Server 2014 Fast Track

Oracle Database Reliability, Performance and scalability on Intel Xeon platforms Mitch Shults, Intel Corporation October 2011

Distribution One Server Requirements

The 8Gb Fibre Channel Adapter of Choice in Oracle Environments

Summary. Key results at a glance:

Veeam Backup & Replication Enterprise Plus Powered by Cisco UCS: Reliable Data Protection Designed for Virtualized Environments

Amazon EC2 XenApp Scalability Analysis

Benchmarking Guide. Performance. BlackBerry Enterprise Server for Microsoft Exchange. Version: 5.0 Service Pack: 4

Dell Microsoft SQL Server 2008 Fast Track Data Warehouse Performance Characterization

The MAX5 Advantage: Clients Benefit running Microsoft SQL Server Data Warehouse (Workloads) on IBM BladeCenter HX5 with IBM MAX5.

Cisco UCS B460 M4 Blade Server

Microsoft Exchange Solutions on VMware

XenDesktop 7 Database Sizing

Best practices for fully automated disaster recovery of Microsoft SQL Server 2008 using HP Continuous Access EVA with Cluster Extension EVA

8Gb Fibre Channel Adapter of Choice in Microsoft Hyper-V Environments

Microsoft Office SharePoint Server 2007 Performance on VMware vsphere 4.1

QLogic 2500 Series FC HBAs Accelerate Application Performance

The Benefits of Virtualizing

ICONICS Choosing the Correct Edition of MS SQL Server

Dell Virtual Remote Desktop Reference Architecture. Technical White Paper Version 1.0

Microsoft Dynamics NAV 2013 R2 Sizing Guidelines for On-Premises Single Tenant Deployments

IT Firm Virtualizes Databases: Trims Servers 85 Percent, Ups Performance 50 Percent

Scalability and Performance Report - Analyzer 2007

Atlantis USX Hyper- Converged Solution for Microsoft SQL 2014

Accelerating Enterprise Applications and Reducing TCO with SanDisk ZetaScale Software

Key Attributes for Analytics in an IBM i environment

Cisco Prime Home 5.0 Minimum System Requirements (Standalone and High Availability)

Lab Validation Report

Transcription:

Proven Performance for Accenture Duck Creek Policy Administration Commercial Lines Benchmark testing confirms scalability and performance of Microsoft SQL Server 2012 and servers based on the Intel Xeon processor E7 family Abstract Accenture, Microsoft, and Intel worked together to test the performance and scalability of Accenture Duck Creek insurance policy-administration software for Commercial Package Policies (CPPs) running on Microsoft SQL Server data management software, the Windows Server operating system, and servers based on the Intel Xeon processor E5 family and the Intel Xeon processor E7 family. The results of the tests were very impressive, confirming that the joint solution can comfortably meet the performance and scalability requirements of very large tier-one insurance carriers. Highlights of the tests include the following: Performance exceeded expectations. At peak production and with a simulated user community of 20,000, the system was able to process approximately $142 million in premium per hour and close to 70,000 transactions per day (including nightly batch renewals). The system also demonstrated tremendous scalability. The tests showed that new business premium tripled from $8.55 million per hour to $25.19 million per hour when the workload and number of processing servers was tripled. Policy page response times were less than 2 seconds for average-size and large-size policies (which make up 99 percent of a typical tier-one insurance carrier s workload). This report provides the full results of the tests, in addition to details about the test environment and the methodology used.

2013 Microsoft Corporation. All rights reserved. This document is provided as-is. Information and views expressed in this document, including URL and other Internet Web site references, may change without notice. You bear the risk of using it. This document does not provide you with any legal rights to any intellectual property in any Microsoft product. You may copy and use this document for your internal, reference purposes. Benchmark Testing Results: Accenture, Microsoft, and Intel ii

Table of Contents Overview... 1 Accenture Duck Creek Policy Administration Commercial Lines... 2 Benefits of SQL Server 2012... 2 Benefits of Servers Based on Intel Xeon Processors... 2 Benchmark Test Environment... 3 Benchmark Test Configuration... 3 Test Hardware and Software... 4 Workloads... 5 Database... 5 Test Methodology... 6 Business Transactions... 6 Test Description and Results... 8 Performance Tests... 8 Scalability Tests... 9 Page Timing Tests... 10 Batch Renewal Tests... 11 Summary... 12 Appendix A: Benefits of SQL Server 2012... 13 Appendix B: Benefits of Servers Based on Intel Processors... 15 Appendix C: Test Data... 16 Appendix D: Performance Statistics... 18 Performance Test Statistics... 18 Scalability Test Statistics... 20 Additional Information... 22 Benchmark Testing Results: Accenture, Microsoft, and Intel iii

Overview Accenture, Microsoft, and Intel worked together on benchmark testing of the Accenture Duck Creek policy administration enterprise software solution for the insurance industry running on Microsoft SQL Server 2012 data management software, the Windows Server 2008 R2 operating system, and servers based on the Intel Xeon processor E5 family and the Intel Xeon processor E7 family. The goal of the tests was to assess the performance and scalability of the commercial lines Commercial Package Policies (CPPs) software solution for the workload volumes of tier-one insurance carriers, defined as carriers with more than $5 billion in annual premium. The benchmark tests were conducted at the Microsoft Platform Adoption Center (PAC) in Redmond, Washington, using a line-of-business CPP with property and general liability. The tests used a starting database that simulates the workloads of a typical tier-one commercial insurance carrier with 20,000 active agents and a $17.5 billion book of business: approximately 3.9 million policies and quotes, with $17.5 billion in written premium on bound policies, and approximately $26.3 billion in additional quoted premium. The performance tests measured the amount of premium the system was able to process for typical business transactions (such as quote, cancel, reinstate, endorse, and new business). The scalability tests investigated how well the system scaled with increasing workload and increasing number of processing servers. The test results were very impressive, confirming that the solution can meet the performance and scalability requirements of very large tier-one insurance carriers. Table 1 shows the amount of premium that the joint solution was able to process, per business transaction type: Table 1. Test results: premium processed on the joint solution Transaction type Daily premium processed Premium processed in a 260-business day year Quotes $182 million $47 billion Cancellations ( $28 million) ( $7 billion) Reinstatements $32 million $8 billion Endorsements $28 million $7 billion New business $154 million $39 billion Additional highlights of the benchmark tests included the following: The system processed more than 54,000 business transactions per day, and close to 70,000 business transactions per day including nightly batch renewals (for an average of 2.8 daily business transactions per user for a 20,000 user community). The system was able to process a batch renewal premium of $153.9 million in 28 minutes. Policy page response times were excellent. For an average-size policy, the page response time was approximately one second. For a large policy (2.5 times larger), the page response time was less than two seconds. Benchmark Testing Results: Accenture, Microsoft, and Intel 1

The system was able to scale almost linearly: new business premium tripled from $8.55 million per hour to $25.19 million per hour when the server count and workload was tripled. This white paper provides the full results of the tests, in addition to details about the test environment and the methodology used. Accenture Duck Creek Policy Administration Commercial Lines The benchmark tests used commercial lines CPP with property and general liability, built on the service-oriented XML-based Accenture Duck Creek EXAMPLE Platform. Accenture offers pre-built out-of-the-box templates with Insurance Services Office (ISO)-based content for CPPs. Carriers can use these templates to launch new insurance products quickly by adapting the base to their own market needs. The templates can inherit more bureau-based content as updates are released, ensuring that they are compliant with the latest bureau standards. For more information about Accenture policy software, see www.accenture.com/microsites/software/pc-insurance/policy-software/pages/templates.aspx. Benefits of SQL Server 2012 The underlying core of an insurance solution is the data collected; this data is maintained in an enterprise database such as Microsoft SQL Server 2012. Running on the Windows Server operating system, SQL Server 2012 provides a reliable, high performing, and scalable foundation for the Accenture policy administration solutions. For more detail about the advantages SQL Server brings to the solution, see Appendix A: Benefits of SQL Server 2012. Benefits of Servers Based on Intel Xeon Processors Servers based on the Intel Xeon processor E5 family and the Intel Xeon processor E7 family are well suited to the demands of large insurance carriers. Servers based on the Intel Xeon processor E5 family are often deployed in distributed architectures that are highly resilient and easily scaled through the addition of more servers. Servers based on the Intel Xeon processor E7 family provide the high-end scalability and availability needed to support large, transaction-intensive databases. The benchmark test environment takes advantage of both the distributed architecture of the Intel Xeon processor E5 family to support the needs of the application and the high-end scalability and availability of the Intel Xeon processor E7 family needed to support the large SQL Server database. For more detail about the role of servers based on Intel Xeon processors, see Appendix B: Benefits of Servers Based on Intel Xeon Processors. Benchmark Testing Results: Accenture, Microsoft, and Intel 2

Benchmark Test Environment To evaluate the performance and scalability of the Accenture Duck Creek policy administration software running on SQL Server, Windows Server, and servers based on Intel Xeon processors, a team from Accenture and Microsoft set up a test environment to emulate the workloads of a large insurance carrier. The following sections describe the environment used in the benchmark testing, in addition to the general methodology used by the test team. Benchmark Test Configuration The benchmark tests ran on Accenture Duck Creek EXAMPLE Platform 4.3.1, Microsoft SQL Server 2012, Windows Server 2008 R2, Microsoft Hyper-V Server 2012, and servers based on the Intel Xeon processor E5 family and the Intel Xeon processor E7 family. The tests used a node-based configuration, in which all components of the Accenture Duck Creek EXAMPLE Platform, including the application and presentation (web) layers, were loaded onto virtual machines (or processing nodes) on a single virtual machine host server. In this configuration, each virtual machine acts as a complete, self-contained processing engine. This configuration was selected because node-based configurations have been demonstrated to perform slightly better than tier-based configurations, in which the application and presentation layers are placed on separate servers. The tests used the following components: 12 virtual machines on a virtual machine host 1 database server 2 load controllers, 3 load agents, and 1 load balancer 1 storage array Figure 1 shows the configuration the team used for the benchmark tests. Figure 1. Benchmark test configuration Benchmark Testing Results: Accenture, Microsoft, and Intel 3

Test Hardware and Software Table 2 describes the hardware and software used in the test environment. Table 2. Details of hardware used in the test environment Servers Details Virtual machine host (1) Virtual machines (12) Database server (1) Load-generation agents (3) and load-generation controllers (2) Load balancer (1) Storage array (1) HP ProLiant DL980 G7 o 1 TB random access memory (RAM) (128 x 8 GB dual in-line memory modules [DIMMs]) o 8 x 10 core Intel Xeon processor E7-4870 2.4 GHz CPU o 1 x 10 Gbps Intel network interface card (NIC) active Windows Server 2012 Datacenter / HyperV Accenture Duck Creek EXAMPLE Platform Server and Accenture Duck Creek EXAMPLE Platform Express Windows Server 2008 R2 Enterprise (64-bit) 8 cores allocated, 16 GB RAM Accenture Duck Creek EXAMPLE Platform 4.3.1.758 HP ProLiant DL980 G7 o 512 GB RAM (64 x 8 GB DIMMs) o 8S* x 10 core Intel Xeon processor E7-4870 2.4 GHz CPU o 1 x 10 Gbps Intel NIC active Windows Server 2008 R2 Enterprise (64-bit) Microsoft SQL Server 2012 (64-bit) HP ProLiant BL460c (blade servers) o 48 GB RAM o 2 x 4 core Intel Xeon processor E5540 2.53 GHz o Dual embedded Broadcom NetXtreme II 5708 Gigabit Ethernet (GbE) NIC Windows Server 2008 R2 Enterprise (64-bit) Microsoft Visual Studio Ultimate 4.0 F5 load balancer o 4 GB RAM o 1 x 1 Gb/s NIC active o RAM Cache enabled o 100 Mbps compression enabled 1 x X-IO Hyper Intelligent Storage Element (ISE) 730 (21 TB) for databases and TempDB o 30 x 900 GB serial attached Small Computer Interface (SAS) 10k drives o 10 x 200 GB solid state drives (SSDs) 1 x X-IO ISE 2 (12 TB) for logs and backup o 20 x 600 GB SAS 10k drives Storage array connected via Fibre Channel host bus adapter (HBA) o QLogic QLE2672 o Dual channel, 16 Gbps per channel *In the benchmark testing, only 40 cores were used. Benchmark Testing Results: Accenture, Microsoft, and Intel 4

Workloads The team used the Accenture Duck Creek ISO Commercial Lines templates in the benchmarking tests. These templates are highly complex insurance products. The team used three policy sizes in the benchmark tests, each with an increasing number of risks and locations. All tests used a CPP policy that includes commercial property and general ledger (GL). Note that average and large policies make up more than 99 percent of the workload of a typical tier-one insurance carrier s workload. Average An average policy includes two locations and two buildings per location. Each building has two associated risks. GL is assigned with two coverages per location and two exposures per coverage. Large A large policy is configured like an average policy, except with five locations. Ultra-large An ultra-large policy is configured like an average policy, except with 80 locations. The property risks on the large policy were limited to 220. Table 3 shows the workloads used in the tests, in premium per policy type: Table 3. Premium for policy transaction types Policy size New business, cancel, and reinstate Endorse Average $6,895 $10,327 Large $21,442 $27,674 Ultra-large $365,377 $368,358 Database The database used for the benchmark tests represents a very large commercial book of business, containing 3,912,500 total quotes/policies. The database was preloaded, or primed, with 94 percent average-size policies, 5 percent large-size policies, and 1 percent ultra-large size policies and quotes. Table 4 shows details of the database: Benchmark Testing Results: Accenture, Microsoft, and Intel 5

Table 4. Database details Test Number of quotes Number of policies Representative premium for quotes Representative premium for policies Average 2,206,605 1,471,100 $15,214,851,750 $10,143,234,500 Large 117,375 78,250 $2,516,754,750 $1,677,836,500 Ultra-large 23,475 15,650 $8,577,225,075 $5,718,150,050 Test Methodology The team used Microsoft Visual Studio Ultimate 2012 to develop the test cases. The workload was increased at a rate of one simulated user per second until the desired, final user load was reached. The test was then run at that load for 120 minutes. The tests measured the number of business transactions of different types (quote, cancel, reinstate, endorse, new business, and enquire) per hour. These results were then translated into premium per hour, based on the policy size and mix of business transactions. Two SQL Server queries were used to determine the number of business transactions completed within the window of time: one query was run at the zero-minute mark and a second query was run at the 120-minute mark. Business Transactions The tests measured the number of business transactions, which are complex, multi-step activities representing the typical activities performed by insurance carriers. The team ran six types of business transactions: New business Inquiries Cancellations Reinstatements Endorsements Quotes Because of the length of the commercial property and GL entry process, complete automated quote entry was built for average-size policies only. Table 5 summarizes the activities involved in creating each business transaction. The instruction count shown in parentheses is the actual number of user interface level system instructions that were coded to perform each activity; these are summarized in the numbered lists for brevity. Benchmark Testing Results: Accenture, Microsoft, and Intel 6

Table 5. Activities in business transactions Business transaction (actual Activities number of instructions) New business (12) 1. Log on 2. Select the search tab 3. Search for a specific policy number 4. Open the policy 5. Click Next twice to get to the pricing page 6. Click Issue 7. After completion, click Logout Inquiry (29)* 1. Log on 2. Select the search tab 3. Search for a specific policy number 4. Open the policy 5. Click through all tabs in the policy 6. After completion, click Logout Cancel (13) 1. Log on 2. Select the search tab 3. Search for a specific policy number 4. Open the policy 5. Select Cancel 6. Set a cancel date 7. Click OK to complete the process 8. After completion, click Logout Reinstate (13) 1. Log on 2. Select the search tab 3. Search for a specific policy number 4. Open the policy 5. Select Reinstate 6. Set a reinstate date 7. Click OK to complete the process 8. After completion, click Logout Endorse (65)* 1. Log on 2. Open an existing policy and add an endorsement 3. Add a location to property 4. Add two buildings 5. Add one risk to each building and set its limits 6. Price the policy and move to GL 7. Add two coverages, each with two exposures to the new location 8. Price the policy and issue the endorsement 9. After completion, click Logout Quote (133)** Included CPP with property (number of locations, buildings per location, and risks per building depended on the quote size) and GL (multiple coverages, each with multiple exposures, depending on the quote size). * Because of the length of this test, only a summary of the total work is shown **Because of the length of this test, only a description of the resultant policy is shown Benchmark Testing Results: Accenture, Microsoft, and Intel 7

Test Description and Results To evaluate the performance and scalability of the Accenture Duck Creek policy administration software running on SQL Server, Windows Server, and servers based on Intel Xeon processors, the team from Accenture and Microsoft conducted four types of tests: Performance Tests Scalability Tests Page Timing Tests Batch Renewal Tests The test environment, including the hardware and software used, is described in the section Benchmark Test Environment. The details of these tests and a summary of the test results are described in the sections that follow. Note: The tests measured the number of business transactions (BTs), which are complex, multistep user-driven interactions. For more information, see the section Business Transactions. Performance Tests The performance tests were designed to identify the maximum throughput of the system under a peak load and to examine how the system performs when the total load exceeds the capacity. The team loaded the test environment with an increasing workload and measured the number of total business transactions completed per policy size and transaction type. The team then translated the results to premium processed per hour, based on the premium per policy size and per business transaction type. Following are specifics of the performance tests: Number of virtual machines: 12 (see the section Benchmark Test Configuration for more information) System capacity: 70 percent of peak capacity, peak capacity, and 277 percent of peak capacity (overloaded) Run time per test: 120 minutes Test composition: 94 percent average-size policies, 5 percent large-size policies, and 1 percent ultra-large size policies, each with the following mix of business transactions: new business (50 percent), endorse (20 percent), inquire (10 percent), cancel (10 percent), and reinstate (10 percent) Table 6 shows the results of the performance tests, in premium processed per hour, per transaction type and policy size. Benchmark Testing Results: Accenture, Microsoft, and Intel 8

Table 6. Performance test results System capacity Size Hourly premium processed Cancel Reinstate Endorse New business Sum per policy size Total 70 percent Peak 277 percent Average $6,333,058 $6,050,363 $18,201,338 $29,693,318 $60,278,077 Large $900,564 $943,448 $2,075,550 $4,888,776 $8,808,338 Ultralarge $1,826,885 $2,192,262 $3,315,222 $12,057,441 $19,391,810 Average $8,274,000 $8,667,015 $25,254,679 $43,428,158 $85,623,852 Large $1,479,498 $1,672,476 $3,569,946 $7,493,979 $14,215,899 Ultralarge $4,201,836 $5,846,032 $6,077,907 $25,941,767 $42,067,542 Average $6,960,503 $7,105,298 $20,814,069 $35,664,388 $70,544,258 Large $1,222,194 $1,275,799 $3,085,651 $5,875,108 $11,458,752 Ultralarge $3,653,770 $4,019,147 $1,289,253 $19,913,047 $28,875,217 ~$88M ~$142M ~$111M The performance tests show that at peak production, the system was able to process approximately $142 million in premium per hour. At 277 percent capacity, an overloaded condition, the system was still able to process substantial premium. Note: For the number of business transactions measured, see Appendix C: Test Data. For statistical details of key sub-systems during the performance tests, see Appendix D: Performance Test Statistics. Scalability Tests The scalability tests were designed to confirm that adding processing nodes makes it possible to maintain response times and transactional throughput per user as the load increases. The team increased the number of Accenture Duck Creek EXAMPLE Platform processing nodes and the workload in a fixed ratio and kept the system operating slightly below peak utilization (at approximately 70 percent of peak). The team measured the total number of business transactions per policy size and transaction type. Following are specifics of the scalability tests: Number of virtual machines: 4, 8, and 12 Workload: 1x (full workload), 2x (double workload), and 3x (triple workload) Run time per test: 120 minutes Test composition: 94 percent average-size policies, 5 percent large-size policies, and 1 percent ultra-large size policies Benchmark Testing Results: Accenture, Microsoft, and Intel 9

o o Average-size policies consists of quote (50 percent) and new business (50 percent) Large-size and ultra-large size policies consists of new business (50 percent) and endorse (50 percent) Table 7 shows the results for the scalability tests. Table 7. Scalability test results Workload, (number of virtual machines) New business Quote Endorse Premium Increase Premium Increase Premium Increase Total premium Increase 1x (4 virtual machines) 2x (8 virtual machines) 3x (12 virtual machines) $17,107,168 -- $10,852,730 -- $9,065,756 -- $37,025,654 -- $38,554,940 2.25x $22,326,010 2.1x $17,375,718 1.92x $78,256,668 2.1x $50,372,491 2.95x $30,434,530 2.8x $19,359,650 2.14x $100,166,671 2.7x The scalability tests show that as the workload increases in proportion to the number of virtual machines (or processing nodes), the amount of premium processed increases almost linearly. Figure 2 shows the results graphically. Figure 2. Scaling as number of nodes and number of users increase Note: For the number of business transactions measured, see Appendix C: Test Data. For statistical details of key sub-systems during the scalability tests, see Appendix D: Performance Test Statistics. Page Timing Tests The page timing tests were designed to measure the average amount of time it takes to process a policy page. Benchmark Testing Results: Accenture, Microsoft, and Intel 10

The team ran two page timing tests: one for average-size policies and one for large-size policies. The team measured the average time to process a page with the system under peak productive load and with no load. The load was generated using a secondary test controller and agents. The tests used a single simulated user, while the load was generated using the performance test (see the Performance Tests section) at peak production. Each test used comparable sets of pages and incorporated page rating. Following are specifics of the page timing tests: Number of virtual machines: 12 Run time per test: 30 minutes Test 1 composition: 100 percent average-size policy, quote Test 2 composition: 100 percent large-size policy, endorse Table 8 shows the results of the policy page timing tests. Table 8. Policy page timing test results Test Average page times without load Average page times with peak load Average quote 0.58 seconds 0.95 seconds Large endorse 1.12 seconds 1.64 seconds The page timing tests show that with increasing load on a static system, policy page response times increase; however, with a maximum (or peak) productive load, they still maintain an average of less than two seconds. These response times comfortably meet the requirements of large insurance carriers. Batch Renewal Tests The batch renewal tests were designed to measure the number of renewals through a web service call in a typical customer implementation. The team generated a batch script to process renewals through a web service call; with no other activity in the environment, the script was simultaneously run against each of the 12 virtual machines running Accenture Duck Creek EXAMPLE Platform Server and the number of renewals was measured. A quiet system with no other activity occurring was tested because renewals are typically run after business hours. Following are specifics of the renewal tests: Number of virtual machines: 12 Run time per test: 60 minutes Test composition: 94 percent average-size policies, 5 percent large-size policies, and 1 percent ultra-large size policies Benchmark Testing Results: Accenture, Microsoft, and Intel 11

Table 9 shows the results of the renewal tests. Table 9. Renewal test results Number of virtual machines Policy size Renewal premium Total renewal premium Average $196,431,655 12 Large $37,802,246 Ultra-large $100,113,298 $334,347,199 The tests showed that the joint solution could meet the renewal requirements of a typical tier-one insurance carrier. During a 60-minute timeframe, the system processed more than $334 million renewals in a quiet system. This means that the joint solution could renew a $40 billion book of business by running for 20 minutes per night for 365 days. Summary The results of the benchmark tests conducted by Accenture and Microsoft with the Accenture Duck Creek insurance policy administration software for CPP running on SQL Server 2012, Windows Server 2008 R2, and servers based on Intel Xeon processors were impressive. Highlights of the tests include the following results: At peak production and with a simulated user community of 20,000, the system was able to process approximately $142 million in premium per hour and close to 70,000 transactions per day (including nightly batch renewals). The number of business transactions processed per day exceeded 54,000 (for an average of 2.8 business transactions per user per day for a 20,000 user community). The system was able to process a batch renewal premium of $153.9 million in 28 minutes. The tests showed that new business premium tripled from $8.55 million per hour to $25.19 million per hour when the workload and number of processing servers was tripled. Policy page response times were less than 2 seconds for average and large policy sizes. These results confirm that the joint solution can meet the performance and scalability requirements of large tier-one insurance carriers. Benchmark Testing Results: Accenture, Microsoft, and Intel 12

Appendix A: Benefits of SQL Server 2012 The benchmark tests were run on SQL Server 2012, a comprehensive, integrated, and reliable datamanagement software solution that provides an enterprise-ready database framework for Accenture Duck Creek policy administration software. Using SQL Server 2012 provides customers with many benefits, including: Scalability and performance features. SQL Server includes many features that can help Accenture customers scale and maintain performance as the number of policies they handle grows. Required nines uptime availability and data protection. Accenture customers can protect their mission-critical databases from downtime and data loss with the new integrated high availability and disaster recovery solution, SQL Server 1, 2, 3 AlwaysOn. Faster deployment. On average, SQL Server database administrators (DBAs) can install and configure new database servers in 1.5 hours, while the largest competitor s DBAs can take 6 hours or more. 4 Lower hardware costs. SQL Server can run on standard commodity server hardware, which can dramatically lower the total cost of ownership (TCO). Lower software costs. The list price of SQL Server is a third of the largest competitors; 5 in addition, SQL Server includes major database-related features, such as high availability; remote disaster recovery; partitioning; data compression; transparent data encryption; spatial; master data management; complex event processing; extract, transform, and load (ETL); online analytical processing (OLAP); data mining; reporting services; and self-service business intelligence (BI) tools. Competitors licensing models add costs for options and add-ins. 6 Simpler systems management and lower staffing costs. SQL Server DBAs can typically manage four times as many physical databases as a competitor s DBAs, leading to an estimated annual savings of $5,779 in administrative costs 1 http://www.microsoft.com/sqlserver/en/us/product-info/why-sql-server.aspx 2 http://www.microsoft.com/casestudies/microsoft-sql-server-2008-r2-enterprise/stratus-technologies/protect-your-mission-criticaldatabases-from-downtime-and-data-loss-with-six-nines-uptime-availability/4000007136 3 http://download.microsoft.com/download/d/2/0/d20e1c5f-72ea-4505-9f26- FEF9550EFD44/Microsoft%20SQL%20Server%20AlwaysOn%20Solutions%20Guide%20for%20High%20Availability%20and%20Disaster%2 0Recovery.docx 4 http://www.alinean.com/pdfs/microsoft_sql_server_and_oracle-alinean_tca_study_2010.pdf 5 http://www.microsoft.com/sqlserver/en/us/tools/cost-savings-calclator.aspx 6 http://www.microsoft.com/sqlserver/en/us/product-info/competitor-compare.aspx Benchmark Testing Results: Accenture, Microsoft, and Intel 13

per database, a 460 percent difference in annual cost of administration per database. 7 Fewer security vulnerabilities. Since 2002, SQL Server has recorded the fewest reported vulnerabilities of any of the major database platforms as compiled by the National Institute of Standards and Technology (NIST). 8 By using SQL Server 2012, Accenture customers can save with reduced hardware, administration, and support fees, which translate into substantially lower costs over the life of the system. For more information about SQL Server, see http://www.microsoft.com/enus/sqlserver/default.aspx. For the top 12 reasons to choose SQL Server, see http://www.microsoft.com/enus/sqlserver/product-info/top-twelve.aspx. 7 http://www.microsoft.com/sqlserver/en/us/product-info/why-sql-server.aspx 8 http://itic-corp.com/blog/2010/09/sql-server-most-secure-database-oracle-least-secure-database-since-2002 Benchmark Testing Results: Accenture, Microsoft, and Intel 14

Appendix B: Benefits of Servers Based on Intel Processors Servers based on Intel Xeon processors are well suited to the demands of policy administration software and can provide dramatic cost savings in comparison with proprietary reduced instruction set computers (RISCs) and mainframe systems. Servers based on the Intel Xeon processor E5 family are often deployed in distributed architectures that are highly resilient and easily scaled through the addition of servers. Servers based on the Intel Xeon processor E7 family provide the high-end scalability and availability needed to support large, transaction-intensive databases. These large and robust servers include integrated support for: Scalable performance. A single eight-socket server based on the Intel Xeon processor E7-8800 product family provides up to 80 processing cores, 160 threads, and 4 TB of memory for high-volume, transaction-intensive workloads. Large cache configurations (up to 30 MB per processor) keep frequently used data close to the processor cores to reduce latency and increase throughput. Data protection. Advanced circuitry detects and corrects errors across all major subsystems and communication pathways and isolates uncorrectable errors so they do not poison other data. High availability. Built-in failover and automatic reconfiguration capabilities help to maintain uninterrupted operation. Machine Check Architecture Recovery (MCA Recovery) works in tandem with Windows Server to provide automated system recovery in complex error scenarios. Serviceability. Built-in support for predictive failure analysis and partitioning lets administrators identify and replace problematic components before they impact performance and without bringing down the system for maintenance. SQL Server 2012 can take advantage of the scalability features of Intel Xeon processor E7 family: 8 sockets or more, 10 cores, and 20 threads per socket and up to 2 TB of RAM in an eight-socket configuration. These servers deliver advanced reliability 9 to increase the mission-critical resilience. The hardware and software work together to deliver advanced support for data error detection and recovery, multiple levels of redundancy, and system recovery. For more information about Intel Xeon processors, see http://www.intel.com/p/en_us/embedded/hwsw/hardware/xeon-previous. 9 http://www.intel.com/assets/pdf/whitepaper/323479.pdf Benchmark Testing Results: Accenture, Microsoft, and Intel 15

Appendix C: Test Data This section provides data collected during the benchmark testing. This data was used to calculate the amount of premium that the system was able to process. Table 10 shows the performance test results for a peak load (in business transactions *BTs+), and the calculated averages, which are also based on typical daily work schedules and workloads for large insurance carriers. Table 10. Peak performance test results System capacity Policy size Cancel Reinstate BT/2-hour test period Quote /endorse New business Inquire Total BT/2- hour test period Total BT /hour Total BT/ hour Peak load Average 2,400 2,514 4,891 12,597 2,704 25,106 12,553 Large 138 156 258 699 148 1,399 699.5 Ultralarge 23 32 33 142 29 259 129.5 13,382 Table 11 shows the scalability test results for 12 virtual machines (in BTs) and the calculated averages, which are also based on typical daily work schedules and workloads for large insurance carriers. Table 11. Scalability test results (in business transactions per hour) Number of virtual machines Policy size BT/2- hour test period New business Quote /endorse Total BT/2- hour test period Total BT /hour Total BT/ hour 12 Average 4,315 4,414 8,729 4,364.5 Large 246 247 493 246.5 Ultra-large 42 34 76 38 4,649 To calculate the premium processed per BT per day, a typical 8-hour work schedule was used: 6 hours of quote/endorse BTs (from both the scalability and performance tests) and a 2-hour mix of the BTs (from the performance tests). Table 12 shows the resulting BTs per hour used to calculate the premium processed per business transaction per typical day. Benchmark Testing Results: Accenture, Microsoft, and Intel 16

Table 12. Business transactions per hour for a typical day Business transactions (BTs) per day Size Quote (48.2%) Cancel (4.7%) Reinstate (4.9%) Endorse (12.5%) New business (24.5%) Inquire (5.2%) Sum BTs/day Total BTs/day Average 26,484* 2,400 2,514 4,891 12,597 2,704 51,590 Large -- 138 156 1,740** 699 148 2,881 Ultra-large -- 23 32 237*** 142 29 463 54,934 ( 2.8 BT /user /day) *The quote value used is from the scalability test for 6 hours: (4,414 BTs X 6 hours); note that large and ultra-large scalability tests for quotations were not run **The endorse value used combined the results from the scalability test for 6 hours and the 2-hour performance test results: (258 BT for 2 hours +[247 BTs X 6 hours]) ***The endorse value used combined the results from the scalability test for 6 hours and the 2-hour performance test results: (33 BT for 2 hours +[34 BTs X 6 hours]) To calculate the average premium processed per transaction type, the test team used the premium values per transaction, as shown in Table 13. Table 13. Premium for policy transaction types Policy size New business, cancel, and reinstate Endorse Average $6,895 $10,327 Large $21,442 $27,674 Ultra-large $365,377 $368,358 Table 14 shows the amount of premium that the joint solution was able to process, per business transaction type, taking typical daily work schedules and workloads into account Table 14. Test results: premium processed on the joint solution Transaction type Daily premium processed Quotes $182,607,180* $47,477,866,800 Cancellations ( $27,910,667) ( $7,256,773,420) Reinstatements $32,371,046 $8,416,471,960 Endorsements $28,336,089** $7,367,383,140 New business $153,727,807 $39,969,229,820 Premium processed in a 260 businessday year *For quotes, only average-size policies were tested: (26,484 BT x $6,895) **To show the difference between the original polcy and the updated policy, the following calculation was used: ((4,891 BT x [$10,327-$6,895])+(1,740 BT x [$27,674-$21,442])+( 237 BT x [$368,358-$365,377])) Benchmark Testing Results: Accenture, Microsoft, and Intel 17

Appendix D: Performance Statistics This section provides detailed information about the state of key sub-systems of the test servers. This section includes utilization counters for: CPU CPU utilization is averaged across the processing nodes and aggregated for SQL Server. Total network throughput The network throughput is averaged per processing node and aggregated for SQL Server. Microsoft.NET common language runtime (CLR) global memory The memory utilization is averaged across the processing nodes. Note: These statistics only apply to the performance and scalability tests. Performance Test Statistics The figures that follow show the performance test results for CPU utilization, network throughput, and memory utilization. Figure 3 shows the performance test CPU utilization. Figure 3. Performance test CPU utilization Figure 4 shows the performance test network utilization. Benchmark Testing Results: Accenture, Microsoft, and Intel 18

Figure 4. Performance test network utilization Figure 5 shows the performance test memory utilization. Figure 5. Performance test memory utilization Benchmark Testing Results: Accenture, Microsoft, and Intel 19

Scalability Test Statistics The figures that follow show the scalability test results for CPU utilization, network throughput, and memory utilization. Figure 6 shows the scalability test CPU utilization. Figure 6. Scalability test CPU utilization Figure 7 shows the scalability test network utilization. Benchmark Testing Results: Accenture, Microsoft, and Intel 20

Figure 7. Scalability test network utilization Figure 8 shows the scalability test memory utilization. Figure 8. Scalability test memory utilization Benchmark Testing Results: Accenture, Microsoft, and Intel 21

Additional Information The following references provide more information about Accenture, Microsoft, Intel, X-IO, HP, Qlogic, and F5: For more information about Accenture, visit: www.accenture.com For more information about Microsoft, visit: www.microsoft.com For more information about Intel, visit: www.intel.com For more information about X-IO, visit: www.xiostorage.com For more information about HP, visit: www.hp.com For more information about Qlogic, visit: www.qlogic.com For more information about F5, visit: www.f5.com Benchmark Testing Results: Accenture, Microsoft, and Intel 22