4.1 Overview of Optimizing Performance

4.1.1 Strategies for Optimizing Performance

4.2 Maximizing Ports For Fax Transmission

4.3 Maximizing Universal Ports

4.4 System CPU Load Test Examples

4.4.1 AG Board Example

4.4.2 QX 2000 Board Example

4.4.3 CG Board Example

4.1 Overview of Optimizing Performance

A board's DSP resources can support a combination of IVR, fax transmit, and fax receive operations. IVR operations include basic Natural Access telephony functionality such as voice play and record, tone detection, and tone generation. You can control DSP resource allocation for a particular board by editing its configuration in the configuration file.

Hardware-related performance is measured by the following factors:

• Available ports per board.

• Number and type of operations that can be run per port.

For a list of boards that support NaturalFax, refer to Section 1.1, NaturalFax Overview.

4.1.1 Strategies for Optimizing Performance

Use one of the following strategies to make the most efficient use of board resources, according to the design and purpose of the application:

• Maximize the number of ports that can run IVR and fax transmit operations.

• Maximize the number of universal ports on a board. A universal port is a port that can use any combination of IVR or fax functions. If a system is configured for universal ports, the application can use any combination of IVR functions such as voice play, voice record, tone detection, and tone generation with fax transmit or fax receive functions. Alternating between IVR and fax functions within a single phone call is supported in a universal port configuration.
  Universal port does not include speech vocoding functions in the context of this manual.

• Support all ports on a board

All of the configurations presented in the following sections were load tested without echo cancellation, using speech encoding for voice functions. If your application requires a different use of DSP resources (for example, adding echo cancellation or using different speech encoding), you may achieve different performance results.

Refer to the appropriate sample configuration files, as listed in each table, for detailed comments and specific configuration statements to achieve the desired performance.

In some configurations, a board's DSP resources cannot support an application that uses all the ports on the board; some ports must be left idle. If the number of operations performed exceed the board's ability to support them, an NFXEVN_SESSION_DONE event is returned with the reason CTAERR_OUT_OF_RESOURCES.

Test your application carefully using the number of ports and the functionality that it will use during normal operation to ensure you have sufficient DSP resources and host CPU resources.

Refer to the board-specific installation and developer's manual for further information on DSP requirements. Refer to Appendix C of this manual for further information on NaturalFax DSP requirements on AG and CG boards.

4.2 Maximizing Ports For Fax Transmission

The following table shows the maximum number of IVR and fax transmit ports that specific boards can support. The table also lists the example configuration files.

This performance strategy achieves the highest number of ports per board, but all ports may not support all operations. Some of the hardware configurations listed can support only IVR and fax transmit operations, and cannot support fax receive operations on any ports. If you need fax receive capability, configure the system for universal ports (refer to Section 4.3).

Maximum Fax Transmit Ports per Board			Configuration file for...
Board	Ports	Restrictions	mu-law	A-law
AG Dual T, AG Quad T	50	IVR and fax transmit only	agpiqt1fax.cfg	None
AG Dual E AG Quad E	50	IVR and fax transmit only	None	agpiet1fax.cfg
AG 2000/200	8	None	agpi2fax.cfg	agpi2faxa.cfg
AG 2000-BRI	8	None	agpi2bri_fax.cfg	agpi2bri_faxa.cfg
AG 2000/4001	16	None	agpi2fax.cfg	agpi2faxa.cfg
AG 4000/800	30	None	agpi4_8fax.cfg	agpi4_8faxa.cfg
AG 4000/1600	60	None	agpi4fax.cfg	agpi4faxa.cfg
AG 4000/3200	120	None	agpi4dfax.cfg	agpi4dfaxa.cfg
QX 2000	4	None	Not required	Not required
CG 6000	120	None	cgfax.cfg	cgfaxa.cfg
CG 6100C	120	None	None	cg61efax.cfg

1 The AG 2000/400 board provides DSPs for up to 16 ports of IVR and fax, but only eight line interfaces. Eight additional line interfaces must be provided with additional hardware. Note: These calculations are valid for both CompactPCI and non-CompactPCI boards.

4.3 Maximizing Universal Ports

The following table shows the maximum number of universal ports that specific boards can support. The table also lists the example configuration files.

Universal ports can run any IVR or fax operation. In some configurations, using the maximum number of universal ports means none of the remaining ports can be used. Under these conditions, each available port can support all operations, but all ports on the board may not be available. This performance strategy provides the best functionality per port.

For fax-only applications not requiring IVR functions, use the universal port configuration files and number of ports-per-board guidelines in this section.

Maximum Universal (Fax and IVR) Ports per Board			Configuration file for...
Board	Ports	Restrictions	mu-law	A-law
AG Dual T AG Quad T	45	Remaining ports must be unused.	agpiqt1fax.cfg	None
AG Dual E AG Quad E	45	Remaining ports must be unused.	None	agpiqe1fax.cfg
AG 2000/200	8	None	agpi2fax.cfg	agpi2faxa.cfg
AG 2000-BRI	8	None	agpi2bri_fax.cfg	agpi2bri_faxa.cfg
AG 2000/4001	16	None	agpi2fax.cfg	agpi2faxa.cfg
AG 4000/800	30	None	agpi4_8fax.cfg	agpi4_8fax.cfg
AG 4000/1600	60	None	agpi4fax.cfg	agpi4faxa.cfg
AG 4000/3200	120	None	agpi4dfax.cfg	agpi4dfaxa.cfg
QX 2000	4	None	Not required	Not required
CG 6000	120	None	cgfax.cfg	cgfaxa.cfg
CG 6100C	120	None	None	cg61efax.cfg

1 The AG 2000/400 board provides DSPs for up to 16 ports of IVR and fax, but only eight line interfaces. Eight additional line interfaces must be provided with additional hardware.

4.4 System CPU Load Test Examples

The tables in the following sections provide examples of host CPU loading. Use this information to estimate how a fax application will affect your host system resources. The percentage of CPU resources used will vary from the statistics presented here depending on the following factors:

• The image size, format, and resolution of the documents used in your fax application.

• The host system configuration.

• Other processing that occurs during NaturalFax processing.

4.4.1 AG Board Example

The following statistics were taken from tests performed on a system with an AG 4000/3200 board and a Pentium II 300 MHz PC with 64 MB of RAM. Conversion Enabled indicates that all ports performed on-the-fly conversions. ECM Enabled indicates that all ports used error correction mode.

Total Number of Ports	Number of Transmit Ports	Number of Receive Ports	Conversion Enabled	ECM Enabled	Avg. System CPU Usage
60	30	30	No	No	12%
60	30	30	Yes	No	20%
60	30	30	No	Yes	13%
60	30	30	Yes	Yes	21%
90	45	45	No	No	17%
90	45	45	Yes	No	28%
90	45	45	No	Yes	18%
90	45	45	Yes	Yes	30%
120	60	60	No	No	20%
120	60	60	Yes	No	34%
120	60	60	No	Yes	24%
120	60	60	Yes	Yes	38%

4.4.2 QX 2000 Board Example

The following table presents the host load test results from QX 2000-based systems with a Pentium 200 MHz processor with 64 MB of RAM. Conversion Enabled indicates that all ports performed on-the-fly conversions. ECM Enabled indicates that all ports used error correction mode.

Total Number of Ports	Number of Transmit Ports	Number of Receive Ports	Modem	Rate	Conversion Enabled	ECM Enabled	Avg. System CPU Usage
4	2	2	V.29	9.6	No	No	1.1%
4	2	2	V.29	9.6	Yes **	Yes	2.5%
4	2	2	V.17	14.4	No	No	2%
4	2	2	V.17	14.4	Yes *	Yes	3.7%
4	2	2	V.17	14.4	Yes **	Yes	4%
32	16	16	V.17	14.4	No	No	28.5%
32	16	16	V.29	9.6	No	Yes	21.5%
32	16	16	V.29	9.6	No	Yes	21.5%
32	16	16	V.29	9.6	Yes *	No	37.8%
32	16	16	V.17	14,4	Yes *	No	53%
32	16	16	V.29	9.6	Yes *	Yes	36%
32	16	16	V.29	9.6	Yes **	Yes	56%

^* Conversions from A3 to A4 page width format, LOW to SUPER_HIGH resolution, 1D to 2D encoding format.

^** Conversions from A3 to A4 page width format, LOW to SUPER_HIGH resolution, 1D to MMR encoding format.

4.4.3 CG Board Example

The following table presents statistics obtained from tests performed on a CompactPCI system with a CG 6000C board and a Pentium II 266 MHz board with 64 MB of RAM running Windows NT 4.0. Conversion Enabled indicates that all ports performed on-the-fly conversions. ECM Enabled indicates that all ports used error correction mode.

Total Number of Ports	Number of Transmit Ports	Number of Receive Ports	Conversion Enabled	ECM Enabled	Avg. System CPU Usage
60	30	30	No	No	15%
60	30	30	Yes	No	24%
60	30	30	No	Yes	16%
60	30	30	Yes	Yes	28%
90	45	45	No	No	22%
90	45	45	Yes	No	38%
90	45	45	No	Yes	25%
90	45	45	Yes	Yes	43%
120	60	60	No	No	30%
120	60	60	Yes	No	51%
120	60	60	No	Yes	32%
120	60	60	Yes	Yes	58%