Archive

Posts Tagged ‘Ignify eCommerce Performance Benchmark’

Ignify eCommerce Performance Benchmark for Online Store Traffic Volume

September 27th, 2013 Brijen Bhatt No comments

Ignify conducted a benchmark to simulate web store performance for demonstrating how catalog browsing and order placement processes scale under a heavy load with a large size database on commodity hardware. A low-level, 4-core server with 16GB RAM and a single SATA hard drive in RAID1 array was able to serve 84,000 page views and 2,000 plus orders in an hour.

Note that 3 web servers with the same database server will handle up to 250,000 page views and 6,500 orders per hour.

High Level Results Summary

The benchmark demonstrated that the web store architecture scales significantly without any performance bottlenecks.

  • 84,000 pages viewed in an hour.
  • 2,185 orders of 3 lines each received in an hour.
  • 2,400 search requests served in an hour, each request resulted in an average of 2,500 search results.

We simulated the following activities in the environment. The diagram below shows the activity flow that we simulated in the environment.

Ignify eCommerce Transaction Flow

Fig.1.0 Ignify eCommerce Transaction Flow

Throughput Measurements

Scenarios

Concurrent users load

Peak hours

Volume per minute

Page View Count

700

1

1400

Order Placement

700

1

36

Keyword Search

700

1

40

Responsetime Measurements

Request

Avg. Response (ms)

Home Page

553

Keyword Search

325

Category Product Listing

553

Product Details

282

Add Item to Cart

87

View Cart

6

Login

338

Address Information

132

Shipping Quotation

137

Payment Information

302

Order Completion

36

The average checkout response time was 1.74 seconds.

Utilization Summary

This section shows the performance characteristics of application server and database server instances that were running on workloads.

Application Server Statistics

Measure

Benchmark

Average CPU Utilization

56%

Peak Memory Utilization

43%

Average Response Time

80ms

Database Server Statistics

Measure

Benchmark

Average CPU Utilization

16%

Peak Memory Utilization

77%

Average Response Time

7ms

Hardware Configurations

Application Server Hardware Specification

  • Single Application Server hosting IIS 7 and Ignify eCommerce Web Store
  • Microsoft Virtual Machine
  • X64-based PC
  • Intel(R) Xeon(R) CPU X7550 @ 2.00GHz 1994 Mhz, 4 Core(s), 4 Logical Processor(s)
  • Windows Server 2008 R2 Enterprise 64 bit
  • 16GB RAM

Database Server Hardware Specification

  • Single Database Server hosting SQL Server 2008 R2
  • Dell PowerEdge R410
  • X64-based PC
  • Intel(R) Xeon(R) CPU E5506 @ 2.13GHz 2128 Mhz, 4 Core(s), 4 Logical Processor(s)
  • Microsoft Windows Server 2008 Enterprise SP1
  • 16GB RAM

Database Disk Configuration

  • SATA RAID1 HDD 750GB
  • 7,200 RPM (Revolutions Per Minute)

Benchmark Data Composition

The benchmark ran on a 16GB database (excludes logs) running with uncompressed data. The system was configured as follows:

  • 500,000 items in the catalog across 800 categories
  • 500,000 customers with 2,500,000 addresses
  • 175,000 price trade agreements
  • 10,000 promotional offers
  • 800,000 orders with total 2,000,000 lines

Benchmark Methodology

The Load Storm was used to generate load, which simulates real user behaviors like logging in, purchasing, clicking links, and browsing the site.

The New Relic was used to monitor the environment that records necessary measurements during the load.

Performance Benchmark Runtime Architecture

Fig.2.0 Performance Benchmark Runtime Architecture

Measurements were recorded on single application and database servers. The benchmarks were taken when the concurrency reached to the steady state. The steady state was maintained for an hour with linear load in repeated runs. Scenarios were considered successful when the deviation among consecutive 3 runs for response time and throughput was 3% or less.

Recommendations

Given the low utilization of the database server, additional web servers can be used to scale up the capacity of the web store. For example:

  • 250,000 page views can be served and 6,500 orders can be placed in an hour using 3 web servers and the same database server.
  • 24,000 search requests can be served in an hour using 10 web servers and the same database server.

Disclaimer

These benchmark results were performed in a controlled lab environment in which Ignify eCommerce was the only running application (that is, no other applications were running on these servers). Performance tests and ratings reflect the approximate performance of Ignify eCommerce. Any difference in system hardware, software design or configuration, customizations, transaction mix, data composition, or indexes may affect actual performance. The transaction mix and data composition affects sizing and hardware requirements. Inherently, Ignify believes that the Ignify eCommerce stack and application are built to scale, and changes to the code base need to go through localized performance testing. A customer or partner should consult other sources of information to evaluate the performance of systems or components they consider purchasing.

If you have questions, or if you would like a copy of the Ignify eCommerce benchmark test, please email us at ecommerce@ignify.com.

Brijen Bhatt is a Sr. Manager of eCommerce at Ignify. Ignify is a technology provider of eCommerceCRM, and ERP software solutions to businesses and public sector organizations. Ignify has won the worldwide Microsoft Partner of the Year Award in 2013, 2012 and 2011. Ignify has been included as the fastest growing business in North America for 7 years in a row by Deloitte, Inc. Magazine and Entrepreneur Magazine from 2007 to 2013.