Browse Prior Art Database

Assessing Processing Requirements for Interactive Voice Response Applications

IP.com Disclosure Number: IPCOM000118708D
Original Publication Date: 1997-May-01
Included in the Prior Art Database: 2005-Apr-01
Document File: 4 page(s) / 96K

Publishing Venue

IBM

Related People

Martin, D: AUTHOR

Abstract

A key factor which prospective purchasers of Interactive Voice Response (IVR) systems consider is the total number of simultaneous telephone calls that the system can handle.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Assessing Processing Requirements for Interactive Voice Response
Applications

      A key factor which prospective purchasers of Interactive Voice
Response (IVR) systems consider is the total number of simultaneous
telephone calls that the system can handle.

      Unlike a telephone exchange, where the limitation is generally
based on the "wiring", the limiting factor in IVR systems is normally
the processor Central Processing Unit (CPU) and Input/Output (I/O)
(disks) subsystem.

      This is because a typical IVR application requires access to a
large number of pre-recorded and digitized voice phrases (segments)
and to other resources; e.g., a remote database server - in order to
carry out a transaction with a caller.  The processing power required
during one telephone call to access these resources can be very
significant.

      The variety of IVR applications is huge.  They range from the
simple playback of one message (e.g., a weather forecast) to complex
'office' systems handling voicemail messages, paging systems, call
forwarding, directory services, user options, etc., all on the same
system.

      Most IVR system suppliers offer a range of processors, the most
powerful of which may typically be ten times as powerful as the least
powerful.  Purchase prices will, if anything, show a greater spread.

      It is desirable to assess the processing requirements of an
IVR system realistically, since excess telephone lines or excessively
powerful processors will be over-priced and uncompetitive.

The described approach uses the following data:
  1.  Published comparable processor and I/O subsystem performance
       benchmarks for each system, e.g., the "Spec" and/or "TPC"
       benchmarks.
  2.  Comparable performance measurements, on a "base" IVR system,
       of each type of IVR activity, e.g., for each of: voice
       recording, voice playback, speech recognition, external host
       access, compression algorithms, etc.  Durations must also be
       carefully controlled and recorded.
  3.  Detailed information on calls:
      o  Average call length (or list of lengths if a number
          of very different IVR applications are to be used)
      o  Per-call count and timings for such as recordings,
          database accesses, listening to menus, entering data
          on phone keypad, speech recognition events, etc.
  4.  Detailed costs of all software and hardware components.
       Typically, an IVR system has a base price plus hardware
       price per telephone line plus software licence per
       telephone line.

The following steps are then performed using this data:
  o  Construct database of machines versus performance benchmarks
      (enumer...