NOTE: This documentation is currently being updated for IV 5.0 release on June 16, 2021.

The Intelligent Voice system is a batch processing system designed to process hours of audio recordings as efficiently as possible.  We also offer systems designed for real-time transcription, low latency key word spotting, IoT hub devices and embedded applications and more - for details on running these solutions on Azure please contact us.

5.0 requirements

A basic installation of Intelligent Voice 5.0 with GPU acceleration on Azure requires a singe GPU instance type with 128GB RAM and 4 CPUs.

At time of writing in June 2021, the Standard_NC16as_T4_v3 type is recommended. Other supported sizes:

Standard_NC64as_T4_v3

Any NCv2-series

Any NCv3-series

Intelligent Voice 5.0 has full support for installation on Red Hat Enterprise Linux 7, and preview support for Ubuntu 20.04

The minimum storage requirement is 500GB.

5.0 production deployment example

An example of a full production system deployment with autoscaling.

This system uses a single application server, with an Virtual Machine Scale Set deploying processing node images from a Shared image gallery.  The current state of the gearman queue is sent to Application Insights using Telegraf and scaling rules created based on the number of jobs.

This system was benchmarked with different counts of processing nodes, and average throughput of each processing node was 67 audio hours per hour.

Application server

Standard B12ms (12 vcpus, 48 GiB memory)

Azure image: Red Hat Enterprise Linux 7.8 (with latest updates applied before starting install)

OS Disk: 610GB Premium SSD LRS, partitioned as follows:

Filesystem                Size Mounted on
/dev/mapper/rootvg-rootlv 40G  /
/dev/mapper/rootvg-usrlv  28G  /usr
/dev/sda2                 494M /boot
/dev/mapper/rootvg-optlv  190G /opt
/dev/sda1                 500M /boot/efi
/dev/mapper/rootvg-homelv 6.0G /home
/dev/mapper/rootvg-varlv  150G /var
/dev/mapper/rootvg-tmplv  35G  /tmp

(tomcat8 temp files are in default location /opt/apache-tomcat-8.5.61/temp/)

Processing node images

Azure image: Red Hat Enterprise Linux 7.8 (with latest updates applied before starting install)

OS Disk: 610GB Premium SSD LRS, partitioned as follows:

Filesystem Size Mounted on
/dev/mapper/rootvg-rootlv 40G  /
/dev/mapper/rootvg-usrlv  20G  /usr
/dev/sda2                 494M /boot
/dev/mapper/rootvg-varlv  310G /var
/dev/mapper/rootvg-tmplv  25G /tmp
/dev/sda1                 500M /boot/efi
/dev/mapper/rootvg-homelv 6.0G /home
/dev/mapper/rootvg-optlv  20G  /opt

4.13 Minimum requirements

A basic installation of Intelligent Voice with GPU acceleration on Azure requires a GPU instance type with 64GB RAM and 4 CPUs.  At time of writing in May 2020, All NC-series, NCv2-series, NCv3-series, ND-series, NDv2-series, NV-series and NVv3-series meet these requirements.  NVv4-series are not suitable.

The recommended minimum instance type for an all-in-one installation is the Standard_ND6s.

The minimum storage requirements are for an managed disk volume of 60GB.  This will need to be increased if more than 100 hours of audio data is processed. 

4.13 Low volume batch processing

To provide the lowest cost per audio hour, we recommend that Intelligent Voice is installed as a minimal auto-scaling solution.  This would comprise at least:

• 1 application server, online whenever the API or review platform are required
• 1 or more processing nodes (GPU optional), online only when processing new data

The minimum requirements for an application server are 16GB RAM and 2 CPUs.  Low cost instances types such as the Standard_E2a_v4 are suitable for this.  The minimum storage requirements are for a managed disk volume of 60GB.

The minimum requirements for a processing node without a GPU are 16GB RAM and 4 CPUs. The Standard_E2a_v4 is a suitable instance type for this.  The minimum storage requirements are for an EBS volume of 60GB.

The minimum requirement for a processing node with a GPU is a Standard_NC6_Promo instance type, recommended Standard_ND6s.   The minimum storage requirements are for a managed disk volume of 60GB.

Typically processing batches of audio under 4 hours in total size will be cheaper without a GPU.  Batches over 4 hours will be cheaper with a GPU, and much faster.

IV has validated this base VM images in Azure:

Red Hat 7.7

High volume batch processing

More details to follow - please contact us for information