Jump to: navigation, search

Traffic and Capacity Testing

Overview

Use this section to determine the required capacity of your GVP servers, based on anticipated traffic characteristics or by running tests on an existing system.

When measuring peak capacity on a single GVP machine, CPU usage is usually the determining factor—memory has not been an issue in most test cases. Therefore, the sample test results in this section concentrate on CPU usage and other criteria.

In addition, the Media Resource Control Protocol (MRCP) server that supports Automatic Speech Recognition (ASR) applications, must not share a host with a GVP server. You can use multiple MRCP servers for a particular test, however, it is important that the MRCP resources do not cause a bottleneck during testing.

This section contains test summary tables to assist in the difficult task of sizing in the face of so much raw data contained by the tables in the following sections. Each table is prefaced with a description of its intent, with suggestions for interpreting and applying the data.

The complexity of VoiceXML and CCXML applications impacts capacity testing, therefore, the Genesys QA performance testing results in this section are derived from test cases using four different VoiceXML applications and two different CCXML applications.

VoiceXML Application Profiles

VoiceXML performance testing was conducted on four major application profiles. Their characteristics are outlined in the tables below. The call flow duration for each application profile is for a single call or CD1 (see Call Duration (CD) and Peak Capacity (PC)).

Profile: VoiceXML_App1 Profile: VoiceXML_App2
A simple DTMF-only application designed to refill calling cards.
  • Total number of digits (DTMF input only) = 52, including:
    • Toll free number from the back of the card
    • Refill card PIN number
    • Refill dollar amount
    • Credit card number
    • Credit card expiration date
    • Zip Code of caller
  • Number of VoiceXML pages = 18
  • VoiceXML complexity = low
  • Number of audio prompts = 9
  • Number of audio files used in prompts (no TTS) = 107
  • ECMA script complexity = moderate
  • Number of VoiceXML pages = 6
  • Number of Java script help functions in each VoiceXML page = 13
  • Call flow duration:
    • 74 seconds (pre GVP 8.1.2)
    • 76 seconds (GVP 8.1.2 and higher)
A complex application designed for insurance coverage inquiries.
  • Speech input, including:
    • Type of request
    • ID card number
    • Confirmation
    • Relationship with insurance plan holder
    • Date of birth confirmation
  • Number of VoiceXML pages = 10
  • VoiceXML complexity (~ 1 MB of content) = High
  • Number of audio prompts = 7
  • Number of audio files used in prompts (3 with TTS) = 29
  • ECMA script complexity = high
  • Call flow duration:
    • 70 seconds (ASR engine)
    • 55 seconds (ASR engine simulator)
Profile: VoiceXML_App3 Profile: VoiceXML_App4
QA ASR/TTS load application.
  • Speech input, including:
    • Words
    • Digits
    • Hotkey (NGI)
    • Yes or no confirmation
  • Number of VoiceXML pages = 1
  • VoiceXML complexity = low
  • Number of audio prompts = 7 prompts involve 7 audio files and 7 TTS
  • ECMA script complexity = low
  • Call flow duration = 62 seconds
Composer-generated application designed for IVR-assisted banking.
  • Input a total of 20 digits (DTMF only):
    • Input current customer number
    • Confirm contact ID
    • Input debit menu option
    • Input debit banking menu
    • Input personal option
    • Input 6 digit secure code
  • Number of VoiceXML pages = 20
  • VoiceXML complexity = medium (~ 400 KB of content)
  • Number of audio prompts = 6 (no TTS, 12 audio files)
  • ECMA script complexity = moderate (4 general JavaScript function files)
  • Call duration = 85 seconds
Profile: VoiceXML_App5 Profile: VoiceXML_App6
VoiceXML_App1 with IVR recording function.


In addition to running the VoiceXML_App1 application, IVR recording was also started when the VoiceXML_App1 began and the call was recorded until the end.

Recording details

  • No of channels = 2
  • Recording type = mp3
  • Bit rate = 16 kbps
  • Recording destination = http
  • Recording metadata = enabled
Simple IVR recording application with continuous speech input from the caller.
  • Number of VoiceXML pages = 1
  • VoiceXML complexity = low
  • Number of audio prompts = 2 (2 audio files)
  • Call flow duration = 75 seconds (NGI)

Recording details

  • No of channels = 2
  • Recording type = mp3
  • Bit rate = 16 kbps
  • Recording destination = http
  • Recording metadata = enabled


CCXML Application Profiles

CallControlXML (CCXML) performance testing was conducted on two major application profiles. Their characteristics are outlined below. The call flow duration for each application profile is for a single call or CD1 (see Call Duration (CD) and Peak Capacity (PC)).

Profile: CCXML_App1 Profile: CCXML_App2
An outbound application that joins multiple call legs, dialogs, and conferences.
  • Includes the following steps:
    • Call customer and connect to a dialog
    • Call agent and connect to dialog
    • Exit agent dialog
    • Exit customer dialog
    • Create conference
    • Join customer and agent to conference
    • Disconnect agent
    • Disconnect customer
    • Destroy conference
  • Number of CCXML (JSP) pages = 2
  • CCXML complexity = medium
  • Customer call duration = 8.7 seconds
  • Agent call duration = 8.6 seconds
  • Conference call duration = 6 seconds
Simple conference recording call.
  • Includes the following steps:
    • Create a call to agent
    • Agent receives an invite and a dialog is created for agent to ring back
    • Agent answers the call and a conference is created to join caller and agent
    • Conference is established and dialog is created for recording
    • Call is disconnected from caller after 15 seconds of recording
  • Number of CCXML pages = 1
  • Number of VoiceXML pages = 2
  • CCXML complexity = medium
  • Call duration = 21 seconds

top | toc

Feedback

Comment on this article:

blog comments powered by Disqus
This page was last modified on July 11, 2018, at 00:47.