DragonMedicalServer 3.6 InstallationGuide EN

Nuance ®
Dragon Medical
Server
Installation Guide
Dragon Medical Server 3.7
Table of contents
Introduction 4
Deployment overview 4
Information checklist 5
Prerequisites 6
Hardware requirements 6
Software requirements 8
Network requirements 10
Load balancing 11
Dragon Medical Server health check 11
Nuance Management Server health check 12
Permissions 13
System availability 14
Critical components 14
Non-critical components 15
Remarks 15
Installation procedure 16
Installing the Nuance Management Server 17
Launching the Nuance Management Center 18
Server Cluster Wizard 19
Modifying a Dragon Medical Server Cluster 20
Organizations and sites in the Nuance Management Center 21
Installing Dragon Medical Server 22
Silent setup 22
Installing Dragon Medical SpeechKit Browser edition web server 23
Dragon Medical SpeechKit 24
Microphones 24
Virtualized environments 24
Licensing 25
Users and Groups in the Nuance Management Center 26
Upgrading from versions lower than 2.0 27
Prerequisites 27
Procedure 27
Remarks 27
-2-
© 2011-2020 Nuance Communications, Inc. All rights reserved.
Upgrading from version 2.0 or higher 29

Prerequisites 29
Procedure 29
Downgrade restrictions 30
Remarks 30
Network guidelines 32
Performance optimization 32
Network bandwidth 32
Network latency 33
Antivirus software 33
Proxy servers 34
Troubleshooting 35
Appendix 36
Speech recognition topics 36
Nuance Healthcare Support 36
Servers and ports 36
Acknowledgments 38
-3-
Introduction
This document will guide you through an on-premise installation.
Deployment overview
A typical installation has the following components:

n Speech recognition servers.
n Nuance Management Server (NMS).
n Dragon Medical and NMS databases. These can be created on the same or separate Microsoft SQL Server
installations.
n A file share: A network location available to all servers.
n Nuance Management Center (NMC): A web browser interface for administration of your speech recognition
system via NMS.
-4-
n Dragon Medical SpeechKit: COM, .NET, Browser, iOS and Android editions.
n Optional: Load balancer.
n Optional: A web server to host Browser edition.
Information checklist
During installation, you need the following:
Licensing information
n An Organization ID.
This is unique to your organization and required for NMS installation.
n A license key.
n The partner GUID for your speech recognition application vendor.
If you are using an application developed by Nuance you do not need to enter a partner GUID.
Network information
n An Active Directory/Windows domain user name under which to run both Dragon Medical Server and NMS.
See also: Permissions.
n The name of your SQL database server.
n The UNC root path to the system file share.
Speech recognition information

n The locations of one or more Data Packs.
-5-
Prerequisites
The hardware, software and network requirements of a speech recognition system are detailed below. Contact
our support team for more information on sizing your system.
Hardware requirements
Dragon Medical Server
Dragon Medical Server initializes a speech recognition engine that allocates RAM to each user session, whether
or not the user is recording. When the user is recording, the speech recognition engine consumes both RAM and
CPU.
When sizing your speech recognition system:
n Calculate RAM requirements, taking into account the maximum number of users who will be connected to a
server at the same time.
n Calculate CPU requirements, taking into account the maximum number of users who will be recording at the
same time.
In our experience with both mobile and non-mobile users, the ratio of non-recording users (consuming RAM) to
recording users (consuming RAM and CPU) is approximately 3 to 1. This ratio should be taken into account
when sizing your system.
The following requirements are the minimum recommended for Dragon Medical Server activity:
n Memory: 4 GB service overhead for each Dragon Medical Server plus 2 GB per Data Pack and 800 MB per
session.
These memory requirements are in addition to the operating system requirements (minimum 2 GB,
depending on the operating system version).
The memory requirements per session are based on session lengths of approximately 30 minutes.
Important: Dragon Medical Server uses all of the available memory on the server, on which it is installed, for
caching. This behavior is not configurable. If there are other processes running on the same server, Dragon
Medical and those processes will have severely reduced performance, therefore Microsoft SQL Server and
Nuance Management Server (NMS) must be deployed on separate servers.
n CPU: Dragon Medical processes require two physical CPU cores per server, independent of expected usage.
In addition, each recording user requires 40% of a 2.7 GHz latest-generation, physical Intel core (equivalent).
We recommend high-end Intel Xeon CPUs. For example, 30 recording users can be served by 12 physical
cores that are exclusively for speech recognition. Considering the two physical cores needed for service
overhead plus reserving one additional core for the Microsoft Windows operating system, a total of 15
physical cores are required in a server, to serve 30 users speaking at the same time.
Only consider the number of physical cores when sizing your speech recognition system; hyper-threading is
already taken into account in the performance value specified above. A hyper-threaded core does not
provide the capacity of a full CPU core, even though the Windows Task Manager displays two CPUs per
hyper-threaded core.
If older or less powerful CPUs are used, these requirements must be scaled up appropriately.
-6-
Virtualized environments impose an additional CPU load. For example, on a server with a latest-generation
2.7 GHz Intel CPU, each recording user in a virtualized environment requires a full vCPU. Virtualized
environments usually assign a vCPU to half a hyper-threaded physical CPU core; this means that the
vCPU’s performance will be approximately 40% of that of a single physical core. Virtualized environments
also require fast local hard drives for local data storage.
Do not assign surplus virtualized CPU resources on a speech recognition server. Dragon Medical Server is a
real-time system and it must be able to use full CPU capacity without delay.
If a virtualized speech recognition server shares hardware resources with other virtualized server
infrastructure, performance (both speed and throughput) can be affected.
It is recommended not to move virtualized speech recognition servers between physical servers while they
are active; if you do, users will experience delayed results.
If your load balancer does not support the three-state model, then the load-balanced system should have
enough surplus CPU resources to ensure that no single server ever becomes overloaded. As load balancers
usually do not succeed in distributing system load evenly, we recommend sizing the system for
approximately 30% more sessions and recording users than anticipated.
n Each server requires 30 GB of hard drive space for installation binaries and temporary data storage, plus 4
GB additional space for each Data Pack, plus 300 MB multiplied by the maximum number of user sessions
the server can handle for dictation data and user profiles.
System file share

n Each user requires approximately 7.5 GB of storage for dictation data files and approximately 200 MB of
storage for user profile data.
n In addition to the hard drive space needed per user, a minimum of 12 GB of storage is required for log files.
The oldest log files are deleted automatically from the system file share when the maximum size of the
central log folder is exceeded.
n 200 GB of additional hard drive space is required for system data.
Nuance Management Server

The following resources are recommended for NMS to support 250 concurrent Dragon Medical Server user
sessions:
n 10 GB of memory in addition to the operating system requirements.
n Two physical CPU cores.
n 40 GB of hard drive space per server.
These requirements are in addition to those of the operating system and other software on the same server. For
systems with more than 250 concurrent Dragon Medical Server user sessions or systems that use NMS for
purposes other than Dragon Medical Server, resources must be scaled up accordingly.
-7-
Software requirements
Important: For more information on supported versions of the software mentioned below, see the release
notes for the Dragon Medical Server version you are deploying.
Dragon Medical Server

n Operating system: Microsoft Windows Server
n Microsoft .NET Framework
n Database environment: Microsoft SQL Server
See also: Microsoft SQL Server
Nuance Management Server

n Operating system: Microsoft Windows Server
n Microsoft .NET Framework
n Database environment: Microsoft SQL Server
Server collation: Use SQL_Latin1_General_CP1_CI_AS; other collation settings are not supported.
Make sure the operating system and the SQL server have the same regional and language settings.
See also: Microsoft SQL Server
n ASP.NET
n Microsoft Internet Information Services (IIS)
Note: Dragon Medical Server complies with the Federal Information Processing Standard (FIPS) but NMS
does not. Do not enable FIPS Compliance Mode on the server that hosts NMS.
NMC
n One of the following desktop browsers:
Microsoft Internet Explorer 10 or 11
Google Chrome 41 or higher
Apple Safari 9
Microsoft SQL Server

Different SQL Server editions have different functionality, scalability and performance envelopes; make sure to
use an edition suitable for your deployment. For more information, see the corresponding Microsoft
documentation (for example, Features Supported by the Editions of SQL Server 2012: Cross-Box Scale Limits
or Editions and Supported Features for SQL Server 2016: RDBMS Scalability and Performance).
Support for Nuance products using Microsoft SQL Server is valid only when it is used within the functionality,
scalability and performance envelope stated by Microsoft.
-8-
Note: By default, Microsoft SQL Server is allocated all of the available memory on the server where it is
installed. If your SQL server shares hardware resources with other services, limit the maximum memory
allocation to the SQL server.
-9-
Network requirements
The network on which you run Dragon Medical must have the following characteristics:
n Dragon Medical SpeechKit connection to the speech recognition servers:
Upload: at least 10 kB/sec per recording user
Download: 2-8 kB/sec per recording user
See also: Network guidelines.
n Within your data center the following ports must be available for server to server communication:
Dragon Medical Server to NMS: 80 or 443 (HTTP or HTTPS).
Dragon Medical Server to Dragon Medical Server: 4714 (TCP/IP).
SQL databases: 1433 and 1434 (TCP/IP).
Dragon Medical system file share and NMS file share: 137-139 and 445 (TCP/IP).
n Within any server the following port must be available for server-internal communication (no external traffic
must use this port):
The local instance of a reverse proxy (for example, the Apache web server) to Dragon Medical Server: 4714
(TCP/IP).
- 10 -
n The following ports must be available for communication between client applications and your servers:
Dragon Medical SpeechKit to Dragon Medical Server (default): 3712 (configurable, HTTP or HTTPS; third-
party client applications might have specific requirements).
Dragon Medical SpeechKit Browser edition to your web server: 80 or 443 (HTTP or HTTPS).
Dragon Medical One Desktop Application, Dragon Medical Directand Dragon Case and Care to NMS: 80 or
443 (HTTP or HTTPS).
NMC to NMS: 443 (HTTPS).
Optional: TCP access from Dragon Medical SpeechKit iOS edition to the Nuance Text-To-Speech server at
nsa.nuancehdp.com on port 443 (SSL encrypted).
Make sure your firewall does not block these ports.
See also: Servers and ports.
n Load balancing.
n You must have an SSL certificate bound to your domain to support HTTPS.
Load balancing
A typical speech recognition system involves multiple servers. A third-party load balancer must be used to
distribute the workload among all speech recognition servers. Depending on the size of the system, hardware or
software load balancers can be used.
Note: Nuance cannot take responsibility for your load balancer configuration; you must ensure that it is
configured to support your system efficiently. For more information, contact your load balancer vendor.
Load balancing between Dragon Medical SpeechKit and Dragon Medical Server
We generally recommend concentrating the required hardware resources on fewer, more powerful servers rather
than a larger number of less powerful servers. This reduces the chance of overload on any single server.
For systems that use load balancing, cookie-based client affinity is required. The lifetime of the load balancer
cookie must be set to session. If you are using HTTPS, cookie-based client affinity requires SSL offloading on
the load balancer. Configure the load balancer to execute the Dragon Medical Server health check regularly
(depending on your availability requirements; for example, every 15 - 60 seconds) and react as required to the
different responses.
Load balancing between Dragon Medical Server and NMS

Traffic between Dragon Medical Server and NMS is unidirectional: Dragon Medical Server sends requests to
NMS; there is no traffic from NMS to Dragon Medical Server.
For Dragon Medical Server connections to NMS, no client affinity is required. Configure the load balancer to
execute the NMS health check regularly (every 60 seconds) and react as required to the different responses.
Load balancing between NMS and NMC

For NMC connections to NMS, no client affinity is required. Configure the load balancer to execute the NMS
health check regularly (every 60 seconds) and react as required to the different responses.
Dragon Medical Server health check

To check the status of a speech recognition server, use this HTTP GET call:
http(s)://<your.server:port>/SpeechAnywhereServices/IsAlive
- 11 -
Depending on the response to the call, configure your load balancer to do the following:
HTTP 200 OK: The Dragon Medical Recognition Service is ready for new session requests. New sessions
and existing session requests should be routed to the speech recognition server.
HTTP 503 NOT READY: The Dragon Medical Recognition Service is not ready to accept new sessions, but
still able to process requests for ongoing sessions. Continue to route traffic to the server for existing sessions
(incoming requests with an affinity cookie) but do not assign new sessions (incoming requests without an
affinity cookie) to the speech recognition server.
No response or other error: The speech recognition server is currently not able to accept traffic. Do not route
any requests to the speech recognition server.
Nuance Management Server health check

To check the status of an NMS server, use one of the following HTTP or HTTPS GET calls:
http://<NMS–Server-Name>:80/NMS/Platform/ConfigurationSvc/v1/Status
https://<NMS–Server-Name>:443/NMS/Platform/ConfigurationSvc/v1/Status
If the NMS server is running, it returns code HTTP 200 and an XML response to the load balancer. The first lines
of the XML response indicate the server status:
<PlatformStatus xmlns="http://schemas.datacontract.org/2004/07/NMS.Platform.Objects"
xmlns:i="http://www.w3.org/2001/XMLSchema-instance">
<Status>Running</Status>
<ServerDateTimeUTC>2016-10-26T06:13:45.0906088Z</ServerDateTimeUTC>
If the load balancer makes this HTTP or HTTPS GET call and receives anything other than the expected response,
the NMS server is currently not able to accept traffic. Configure the load balancer to reroute the traffic.
- 12 -
Permissions
You need a single Active Directory/Windows domain user account as the service user for both Dragon Medical
Server and NMS. This account must have:
n Log on as a service rights on all servers.
n Read and Write permissions on the system file share UNC root.
n Public server role on the database server.
n Local system administrator rights on NMS.
Remarks
n To install Dragon Medical Server or NMS, you must have administration rights on the server operating
system.
n If you create the database during server setup, the user installing the server must have the sysadmin server
role on the database server.
n If you do not create the database during server installation, the Dragon Medical service user must have the
dbcreator server role on the database server. The database will be created the first time the speech
recognition service runs.
n Administration of your speech recognition system via NMC does not require administration rights on your
local operating system.
Note: It is recommended that the Dragon Medical service user has the dbcreator server role on the
database server.
- 13 -
System availability
To ensure that the speech recognition system is highly available, you must take into account the critical and non-
critical components. It is important to design your system to support automatic failover; all critical components
should be highly available or redundant.
Critical components
The following components are critical for the speech recognition system:
n Speech recognition servers
n Nuance Management Server (NMS) instances
n NMS Database
n Load balancers between speech recognition servers and NMS instances
n Load balancers between speech recognition servers and clients
- 14 -
Each critical component must be highly available or redundant. The system must have an automatic failover
procedure in place for all critical components.
Important: If a critical component fails and there is no failover procedure in place, the system will no longer
provide recognition results.
To make speech recognition servers and NMS instances highly available:

n Add as many servers as you want for spare capacity. For example, to allow one server to fail with no impact
on end user performance, add one spare server.
n Configure load balancers to automatically redirect all traffic from a failing speech recognition server or NMS
instance to other servers. For more information, see: Load balancing.
To make the NMS Database highly available, see the Microsoft documentation on high availability solutions; for
example: https://msdn.microsoft.com/en-us/library/ms190202.aspx.
To make a load balancer highly available, see the documentation delivered with the product or contact your
vendor.
Non-critical components
The following components are non-critical for the speech recognition system:
n Dragon Medical system database; if it fails, Dragon Medical Analytics will not work and users will experience
latency and reduced speech recognition quality.
n System file share; if it fails, users will experience latency and reduced speech recognition quality.
n Load balancers between NMS and NMC; if they fail, NMC will not work.
Remarks
n When designing your system for high availability, be aware that it must have enough spare capacity to
support peak usage even after a component failure.
- 15 -
Installation procedure
Once you have in place the hardware, software and network prerequisites and a domain user with appropriate
user permissions, proceed with the installation as follows:
1. Install NMS.
2. Open NMC and configure an Dragon Medical Server Cluster.
3. Optional: Configure organizations, users and groups in NMC.
4. Install the speech recognition servers.
5. Optional: Configure your load balancer.
- 16 -
Installing the Nuance Management Server

The Nuance Management Server (NMS) is the central metadata storage database and administration
component used by Dragon Medical Server and Dragon Medical SpeechKit. NMS provides a web browser
interface, the Nuance Management Center (NMC), that enables you to install licenses, manage users, etc.
Make sure to install the NMS version delivered with the installation package, start the NMS_Installer_
DragonMedicalServer_On-premise.exe; it will guide you through the installation procedure.
Enabling HTTPS
During NMS installation you can specify HTTP or HTTPS communication. To use HTTPS, create a server
certificate for the NMS web server. The certificate must be signed by a trusted authority known to both the client
and the server and must be bound to port 443 (configurable in IIS); for more information, see:
https://msdn.microsoft.com/en-us/library/ms733043.aspx.
Configuring the NMS file share locations

Select multi-node deployment for a scalable and highly available system that supports automatic failover; you
will be able to deploy upgrades while the system is running. Select single-node deployment if you plan to have a
single NMS instance and no support for automatic failover; you will not be able to deploy upgrades while the
system is running.
NMS file share locations are required for multi-node deployment only.
The NMS file share locations are hard drive locations used for data storage and shared by all NMS instances in
your system. During NMS installation you must specify UNC root paths for the following NMS file share
locations:
n A file share location for the NMS Platform. You do not need to specify a user account; the NMS Platform
user account is used by default.
n A file share location for NMC (hosted in IIS). You must specify a user account that has create, read, update,
and delete permissions at the selected location; this can be the NMS Platform user account or a separate
user account.
Note: The same file share locations must be configured for each NMS instance in your system.
Optionally, configure a single file share for the NMS Platform and NMC.
Remarks
n On the Logon information page, use an Active Directory/Windows domain user account with the
appropriate permissions.
n During installation, set a strong password for your NMC administration user. The administration user has
access to sensitive data (that can include recorded sound and recognized text from speech-enabled
documents as well as template text) via Dragon Medical Analytics.
n NMS_Installer_DragonMedicalServer_On-premise.exe installs several packages, including the Nuance
Management Service (for compatibility with Dragon Medical Server versions lower than 3.0) and the NMS
Platform (for Dragon Medical Server 3.0 and higher). You can see all installed packages via the Control
Panel, in Programs and Features.
- 17 -
Important: Do not add or remove any of these packages manually. Custom configurations are not
supported.
Launching the Nuance Management Center

To launch NMC, go to: https://<your.NMS.server>/nmchtml
If you have not yet configured a Dragon Medical Server Cluster, the Server Cluster Wizard will open
automatically when you log on to NMC.
- 18 -
Server Cluster Wizard

Before you install Dragon Medical Server you must configure a Dragon Medical Server Cluster. Depending on
how you have designed your system you can have one Server Cluster for each organization or one Server Cluster
for all organizations.
Note: If NMS is to be configured for HTTPS, this must be done before the first Server Cluster is set up.
You can add a Server Cluster at any time; in the Nuance Management Center (NMC) open the Server Cluster
tab, click Add and follow the wizard. Make sure you have all the relevant information.
Partner GUID
If you are using a speech recognition application that is not produced by Nuance, you need the application
vendor's name and partner GUID.
Dragon Medical license

Your license defines how users can use speech recognition.
1. Paste in the license key when prompted and click Show Information to see details.
2. Click Import; the license key is imported into the Nuance Management Server (NMS) database.
Note: If you are creating additional Dragon Medical Server Clusters based on your existing license then this
page does not need to be filled in.
Organization token
All organizations associated with your license are listed; each organization must have a unique organization
token.
The organization token is used by Dragon Medical SpeechKit applications to identify which license should be
used.
To add an organization token, select an organization in the list and click Add/Modify; an organization token is
generated; add a comment if necessary. Click Save .
Note: If an organization token is used by unauthorized personnel you can modify it; the old organization
token will no longer grant access to speech recognition functionality.
Speech recognition system

Select Dragon for your system's speech recognition technology and enter a name for your Dragon Medical
Server Cluster.
Dragon Medical Server Settings

The information you enter here depends on your NMS and Microsoft SQL Server setup. Hover over the icon
next to each field for information and examples.

External NMS URL : Dragon Medical One Desktop Application, Dragon Medical Direct and Dragon Case and
Care require direct access to NMS via an external URL on port 80 or 443. This URL can be different from the
- 19 -
URL you use for the connection between the speech recognition servers and NMS. In this case enter a public
URL for NMS in the External NMS URL field. Otherwise, leave this field blank.
Data Packs
To install Data Packs on your system, do the following:
1. Click Install; a dialog box is displayed.
2. Browse for the Data Pack and click Install. Installing the Data Pack will take a few minutes.
3. Click Close .
Repeat this procedure for all Data Packs your users require.
Completing the Server Cluster Wizard

Click Finish to apply your settings.
Modifying a Dragon Medical Server Cluster

After a Server Cluster has been created you can modify it at any time in NMC. Do the following:
1. Open the Server Cluster tab and click Search .
2. Double-click the Server Cluster you want to modify.
3. Make your changes and click Save .
Important: The Dragon Medical Recognition Service, Dragon Medical Adaptation Service and Dragon
Medical Analytics Service must be restarted for every Dragon Medical Server after any modifications.
- 20 -
Organizations and sites in the Nuance Management

Center
The Nuance Management Center (NMC) is used to configure your speech recognition system at the level of
organizations and sites. It can be useful to do this before you start installing Dragon Medical Server. For more
information, see the documentation delivered with NMC.
Multiple organizations
To set up Dragon Medical for multiple customers, use a separate NMC organization for each customer.
To add an organization, do the following:
1. Open the Organizations ribbon and click Add .
2. Enter information in the Organization Details screen.
3. Import the license key and add an organization token. For more information, see 'Licensing' in the Dragon
Medical Server Administration Guide.
- 21 -
Installing Dragon Medical Server

Make sure you have configured a Dragon Medical Server Cluster before starting the setup.
Dragon Medical Server supports HTTP and HTTPS communication channels and can be configured to listen on
a custom port. To take advantage of the secure communication support, you must have an SSL certificate
available; bind it to Dragon Medical Server port.
If you are using SSL offloading on the load balancer, you must install the certificate on your load balancer (not
speech recognition server).
If you are not using SSL offloading, you must configure the SSL certificate on the speech recognition servers.
For more information, see: http://msdn.microsoft.com/en-us/library/ms733791.aspx
To install Dragon Medical Server, execute DragonMedicalServer.msi and follow the installation wizard.
Remarks
n When you enter the Nuance Management Server (NMS) location, make sure you include the protocol (http
or https) in the URL.
n Make sure that the clocks on all servers and databases are synchronized and that their time zones are
correctly configured.
n Create Databases: If you create the database during server setup, the user installing the server must have
the sysadmin server role on the database server. For more information, see: Permissions.
If you do not to create the database during server installation, the Dragon Medical service user must have
the dbcreator server role on the database server. The database will be created the first time the speech
recognition service runs.
Note: It is recommended that the Dragon Medical service user has the dbcreator server role on the
database server.
Silent setup
You can install Dragon Medical Server via the command line. For example:
"<path>\DragonMedicalServer.msi" NMSSERVICEURL=https://<your.NMS.server>
NMSUSERNAME=<user> NMSPASSWORD=<password> WINDOWSPASSWORD=<password>
SASCLUSTERNAME=<cluster.name> SASDATAROOT=<path> /quiet
Remarks
n SASDATAROOT is an optional parameter. The default value for SASDATAROOT is
C:\ProgramData\Nuance\DragonMedicalServer.
n WINDOWSPASSWORD is the password of the Microsoft Windows account that NMS and the speech recognition
service run on.
n To upgrade Dragon Medical Server via the command line, /quiet is the only required parameter.
- 22 -
Installing Dragon Medical SpeechKit Browser edition

web server
To deploy the Browser edition, copy and extract Nuance_DragonMedicalSpeechKit_Browser_Edition_
Server_<version number>.zip to a folder in your web server's document root (for example,
C:\inetpub\wwwroot\Dragon Medical).
Important: Dragon Medical SpeechKit Browser edition uses relative URLs to locate files; do not change the
folder structure when you deploy the files on your web server.
Open the Nuance.SpeechAnywhere.js file and modify the NUSA_service variable. For example:
NUSA_service="http(s)://<your.Server:port>"
This variable must point to the URL that the Browser edition will connect to for speech recognition services.
Remarks
n Web applications that embed the Dragon Medical SpeechKit Browser edition need to access the
Nuance.SpeechAnywhere.js file. For example, http(s)://<web.server>/<server path to root
folder>/scripts/Nuance.SpeechAnywhere.js
n How this URL is configured depends on the specific third-party web application.
- 23 -
Dragon Medical SpeechKit

Configuring Dragon Medical SpeechKit for on-premise deployment is described in the documentation delivered
with each SDK.
How these configuration options are exposed by third-party applications that integrate Dragon Medical
SpeechKit depends on the application. For more information, see the documentation delivered with the product
or contact your vendor.
Microphones
Speech recognition requires a microphone which records audio data with a bit depth of 16-bit and a sampling
rate of at least 16 kHz.
Button controls
Dragon Medical SpeechKit .NET and COM editions support microphone button controls. For information on
supported devices, see the documentation delivered with the product.
Virtualized environments
Dragon Medical SpeechKit .NET and COM editions can be deployed in a Citrix XenApp/XenDesktop, Microsoft
RDS or VMware Horizon View environment.
To reduce the network bandwidth for audio data from the client PC to the virtualized application, install the
Nuance virtual extension for Citrix, RDS or VMware on the client PC. Download the extensions from the .NET
SDK section of the Nuance Healthcare Development Platform (https://www.nuancehealthcaredeveloper.com);
click Download Citrix and Windows Terminal Server drivers.
You do not need to install the Nuance virtual extension on the server/virtual desktop where your application is
hosted. If the Nuance VMware Server & Virtual Desktop Audio Extension, Nuance RDS Server & Virtual
Desktop Audio Extension or Nuance Citrix Server & Virtual Desktop Audio Extension is already installed, we
recommend you uninstall it unless it is still used by another product.
For more information, see the documentation delivered with the Nuance virtual extensions.
Button controls in virtualized environments

To enable Nuance PowerMic buttons in a Citrix XenApp/XenDesktop, Microsoft RDS or VMware Horizon View
environment, install the corresponding Nuance PowerMic extension on the client PC.
You do not need to install the Nuance PowerMic extension on the server/virtual desktop where your application
is hosted. If the server/virtual desktop component of the Nuance PowerMic extension is already installed, we
recommend you uninstall it unless it is still used by another product.
To enable button controls for third-party devices, contact your device vendor for drivers that support your
virtualized environment; deploy them on the server/virtual desktop where your application is hosted and on the
client PC.
For more information on microphone support, see the Dragon Medical Server Administration Guide.
- 24 -
Licensing
Initial configuration is done in Nuance Management Center (NMC) via the Dragon Medical Server Cluster
Wizard. To install additional licenses, do the following:
1. To access NMC, browse for https://<your.NMS.server>/nmchtml and log on with your admin account.
2. Open the Licensing ribbon.
3. In the Partners group, click Add Existing and do the following:
Name : Enter the name of the application/company.
ID: Delete the automatically generated partner GUID and replace it with the partner GUID you received from
the Nuance Order Desk.
Mode : Production is automatically selected; do not change this.
Note: If you use an application developed by Nuance this step is not necessary.
4. In the Administration group, click Import License Key and paste in your license key.
5. Click Show Information to see the number and expiration date of new licenses and/or the new expiration
date of existing licenses.
6. Click Import.
7. Log out of NMC and log in again.
If the organization associated with the new license does not have an organization token assigned, add one.
Proceed as follows:
1. Open the Organizations ribbon and click Search .
2. Enter search criteria. If you do not enter any search criteria, all your organizations are displayed.
3. Select an organization from the list and click the Details icon.
4. Open the Organization Tokens tab.
5. Click Add (the token is automatically generated) and then Save .
To view a report of the licenses used in your system, do the following:
1. Open the Reports ribbon and click License Use .
2. Enter search criteria and select a License Type from the list.
3. Click Search .
4. To export the results as a CSV file, click .
Remarks
n After applying a Dragon Medical license you must log out of NMC and log in again before the Dragon
Medical Server Cluster functionality is available.
n The organization token allows access to speech recognition functionality. It must be configured in third-party
applications that integrate Dragon Medical SpeechKit. For more information, see the documentation
delivered with the product or contact your vendor.
n The License Use report identifies licenses; you can reassign licenses that have not been used recently.
- 25 -
Users and Groups in the Nuance Management

Center
It is recommended to create a Nuance Management Center (NMC) group that only has the rights to create auto-
texts. Proceed as follows:
1. Log on to NMC with your administration credentials.
2. Open the Groups ribbon and click Add .
3. Select the new group's organization from the drop-down list.
4. In the Role field, select NMC Administrator .
5. Select Dragon Medical as the existing group on which to base the new one. The Dragon Medical group is
created when the first user connects to the speech recognition system from any client. Click OK.
6. Enter a Group Name and select a Site .
7. In the Privileges list, enable: Product Manage Dragon Medical Server
8. Click Save .
Adding members to a group

1. Open the Groups ribbon and click Search .
2. Enter search criteria. If you do not enter search criteria, all your groups are displayed.
3. Select a group and click Add User Account to Group in the Membership ribbon.
4. To add a user account to the group, enable the check box in the Member column in the corresponding row.
5. Click Save .
- 26 -
Upgrading from versions lower than 2.0

Note: For specific version numbers, see the release notes delivered with the package.
Prerequisites
n 360 | SpeechAnywhere Services Server 1.4 or higher. If you are currently running 360 | SpeechAnywhere
Services Server 1.3, it is not possible to upgrade directly to the latest version.
n All NMS servers must be upgraded to the version supplied with the new version of Dragon Medical Server. If
upgrading from NMS lower than version 5.0, make sure that all NMS prerequisites are fulfilled before
starting the upgrade.
Important: Do not uninstall NMS before installing the new version, when upgrading from NMS lower
than version 5.0; upgrade directly to the version delivered with the package.
n NMS/NMC URL and port numbers have changed with NMS 5.0; for more information, see: Installing the
Nuance Management Server.
Procedure
A workstation
Install the supported Data Pack versions via the Nuance Management Center (Server Clusters ribbon,
Dragon Medical Server Cluster Details screen, Data Packs tab).
No changes to any other Dragon Medical Server Cluster or user information settings are required.
The Dragon Medical Server

Browse for your Dragon Medical Server installation package, double-click DragonMedicalServer.msi and follow
the installation wizard.
The web server hosting the Dragon Medical SpeechKit Browser edition
Optional: Delete your current Dragon Medical SpeechKit Browser edition server files and then follow the steps
for deploying the new version.
Alternatively, extract the Browser edition server files for the new version to the same location as the previous
version. This way, Browser edition clients that partially cache Browser edition server files will also be upgraded
without service interruptions.
Remarks
n NMS and Dragon Medical Server are not available on the server being upgraded. If you have more than one
server in a load-balanced environment, you can upgrade servers one at a time so that all servers except the
one being worked on are active.
After removing a Dragon Medical Server from the load balancer pool, wait approximately 30 minutes, for
remaining sessions to close, before performing the upgrade.
n Upgrade all instances of NMS before you upgrade any Dragon Medical Server.
- 27 -
n After upgrading, Dragon Medical Server can perform administrative follow-up tasks that consume some
server resources for an extended period of time. However, the speech recognition system is fully usable
immediately after the upgrade has been performed successfully.
- 28 -
Upgrading from version 2.0 or higher

Upgrading from version 2.0 or higher requires an NMS administrator and is done via the Upgrade
Management page in Dragon Medical Analytics.
You can also still upgrade manually; for more information, see: Upgrading from versions lower than 2.0.
Note: For specific version numbers, see the release notes delivered with the package.
Prerequisites
n 360 | SpeechAnywhere Services Server 2.0 or higher. If you are currently running 360 | SpeechAnywhere
Services Server 1.7 or lower, it is not possible to upgrade directly to the latest version via Dragon Medical
Analytics. For more information, see: Upgrading from versions lower than 2.0.
n All NMS servers must be upgraded to the version supplied with the new version of Dragon Medical Server. If
you are upgrading from NMS lower than version 5.0, make sure that all NMS prerequisites are fulfilled
before starting the upgrade.
Important: Do not uninstall NMS before installing the new version, when upgrading from NMS lower
than version 5.0; upgrade directly to the version delivered with the package.
n NMS/NMC URL and port numbers have changed with NMS 5.0; for more information, see: Installing the
Nuance Management Server.
Procedure
A workstation
1. Install the supported Data Pack versions via the Nuance Management Center (Server Clusters ribbon,
Dragon Medical Server Cluster Details screen, Data Packs tab).
No changes to any other Dragon Medical Server Cluster or user information settings are required.
2. Go to: http(s)://<your.Dragon_Medical.Server>:<port>/SpeechAnalytics and log on with your NMC
administrator credentials.
3. Go to the Upgrade Management page.
4. Click Choose File and browse for your installation package.
5. Click Upload and enter your NMC password.
6. Deploy the new version:
Select the servers you want to upgrade, click Deploy for the version you want to install and enter your NMC
password.
From version 3.2 Release 3 or higher: To upgrade servers one at a time, open the Deploy drop-down menu
for the version you want to install and click Deploy Sequentially. To upgrade servers simultaneously, open
the Deploy drop-down menu for the version you want to install and click Deploy in Parallel.
The upgrade state changes to Pending and then Started . If the state does not change to Done an error
has occurred.
- 29 -
The web server hosting the Dragon Medical SpeechKit Browser edition
Optional: Delete your current Dragon Medical SpeechKit Browser edition server files and then follow the steps
for deploying the new version.
Alternatively, extract the Browser edition server files for the new version to the same location as the previous
version. This way, Browser edition clients that partially cache Browser edition server files will also be upgraded
without service interruptions.
Downgrade restrictions
You can downgrade Dragon Medical Server 3.2 Release 2 or lower only to the version that you upgraded from.
For example, if you upgraded Dragon Medical Server 3.0 to 3.2, you can downgrade to 3.0, but you cannot
downgrade to 3.1.
After a downgrade, you can upgrade Dragon Medical Server 3.2 Release 2 or lower only to the version you
downgraded from or a higher version. For example, if you downgraded Dragon Medical Server 3.2 to 3.0, you
can upgrade to 3.2, but you cannot upgrade to 3.1.
Remarks
n The Upgrade Management section lists all available speech recognition servers in your cluster.
n The Upgrade Packages section contains information on the version that will be deployed and the options
to upload the installation packages.
n Click Delete to delete the uploaded package.
n Click Cancel All and enter your password to cancel all pending upgrades.
n Open the Restart drop-down menu and click Restart All Now to restart the Dragon Medical services on the
selected servers simultaneously or select a time interval to restart services on the selected servers
sequentially. If you have one server only, click Restart.
n We recommend using Mozilla Firefox or Google Chrome for upgrading.
n Installation errors are not handled automatically. You must monitor the upgrade status and make sure that
the upgrade is successfully deployed. To check the upgrade status, refresh the page.
n If upgrading fails, see the Windows Installer logs (C:\Windows\Temp) and Dragon Medical Upgrade Service
logs for information.
n If the automatic upgrade is not successfully deployed in a reasonable amount of time (5 to 15 minutes), you
must manually upgrade on the corresponding servers.
n Rolling back to a version lower than 2.0 is not possible.
n The servers will be restarted automatically if necessary.
n NMS and Dragon Medical Server are not available on the server being upgraded. If you have more than one
server in a load-balanced environment, you can upgrade servers one at a time so that all servers except the
one being worked on are active.
After removing a Dragon Medical Server from the load balancer pool, wait approximately 30 minutes, for
remaining sessions to close, before performing the upgrade.
n Upgrade all instances of NMS before you upgrade any Dragon Medical Server.
- 30 -
n After upgrading, Dragon Medical Server can perform administrative follow-up tasks that consume some
server resources for an extended period of time. However, the speech recognition system is fully usable
immediately after the upgrade has been performed successfully.
n Upgrading Dragon Medical Server also upgrades Dragon Medical Analytics. After upgrading, it is
recommended to clear the browser's cache and refresh the Dragon Medical Analytics page to load the new
version.
- 31 -
Network guidelines
Dragon Medical Server and Dragon Medical SpeechKit provide cloud-based, real-time speech recognition. They
can be used over any network connection: wired, WiFi or 3G/4G cellular networks.
The user’s speech is recognized once the audio data has been successfully transmitted to the on-premise data
center via secure Internet connection. This means that the user experience is affected by the quality of the
network connection.
Network bandwidth, network latency and background noise are likely to have an impact on a user's speech
recognition experience.
This chapter provides guidelines on the minimum network requirements for optimal user experience. Acceptable
performance is often dependent on the user's perception and expectations vary from user to user. Note that the
quality of a network connection cannot be consistently guaranteed over time - a connection that is good one
minute can degrade the next, especially if the user is mobile and roaming between WiFi and/or cellular towers.
Performance optimization
The following features are designed to optimize network-based speech performance:
n Audio compression: Dragon Medical uses a lossy codec to minimize bandwidth requirements while retaining
the high quality audio that is required for speech recognition. Audio transmission from a client to the on-
premise data center requires 10 kB/sec; results returned from the data center to the client require
approximately 2 kB/sec. If text-to-speech is used, audio streamed from the data center to the client will also
require approximately 8 kB/sec.
n Audio chunks and silence detection: Audio is streamed to the server in small blocks to increase
responsiveness. Data is transferred only when audio is being recorded; users who are not currently dictating
do not consume network bandwidth.
n SpeakAhead: Nuance has implemented SpeakAhead technology which enables the user to continue
dictating while the system compensates for network latency. The recognized text is, asynchronously, written
in the field that it was originally dictated for.
Network bandwidth
A user needs 10 kB/sec total bandwidth upstream and 2-8 kB/sec downstream. This means that the network
bandwidth requirements per user are minimal. Network connections are often shared and mobile users often
utilize multiple network connections as they roam from a hospital network to a 3G/4G cellular tower to a home or
public WiFi network.
To test individual connection speed we recommend using a standard tool such as http://speedtest.net. Mobile
users can utilize the native applications for iPhone and Android.
The quality of a network connection is not consistently guaranteed, especially for mobile users. In particular,
consider the following:
n iPhone devices use a weak/poor WiFi connection for data services even in the presence of a strong 3G/4G
signal.
n A strong 3G/4G signal does not guarantee a high bandwidth connection.
- 32 -
Network latency
Network bandwidth is not the only factor in optimal speech recognition. Due to our bandwidth-consumption
optimization, network latency can often play a role in the user’s perception of performance.
In general, Dragon Medical Server converts audio to text in real time; this means one second of audio is
converted to text in one second or less. In order to provide near real-time results to the user, audio is streamed in
small chunks. Network latency is overhead that adds directly to the time it takes the user to see the results.
Network performance tools, such as Speedtest.net, report latency based on a ping of the geographically closest
server. However, since this does not report the latency between the user's device and Dragon Medical Server,
this data is of limited value. If a general speed test reveals high network latency, it is possible that the network
connection in general is poor and the user might experience delayed results. However, low reported network
latency, from a general speed test tool, does not guarantee the network path to the on-premise data center will
also exhibit low latency.
To accurately measure network latency from a desktop device to Dragon Medical Server, use the following
command line tool: tracert <YourServer>.<domain>
Note: Your servers might not reply to pings; it is therefore normal to receive 'request timed out' messages on
the last hop of the traceroute.
To accurately measure network latency from a mobile device to Dragon Medical Server, use an app that
performs a similar traceroute function, such as iNetTools for iOS and traceroute for Android.
The quality of network connections, especially for mobile users is not consistent. A one-time check of network
latency does not guarantee continued optimal performance.
Acceptable performance is often dependent on the user's perception and expectations vary from user to user. If
the user is experiencing high latency, our SpeakAhead feature enables the user to continue dictating while
results are processed asynchronously. However, users can grow impatient or frustrated waiting for results.
High level guidelines for network latency:
Antivirus software
Antivirus software can severely reduce the performance of the system. It is recommended not to run any
antivirus software on the speech recognition servers.
- 33 -
The first step in investigating performance issues in a speech recognition system is to check if antivirus software
is running. If it is running, check if performance issues persist when it is deactivated.
If you use antivirus software, you should exclude the following from scanning:
n All traffic on ports used by Dragon Medical Server and NMS.
n The folders storing the system binaries and data: C:\Program Files\Nuance\DragonMedicalServer and
(by default) C:\ProgramData\Nuance\DragonMedicalServer .
n All system processes (processes executed from binaries in the C:\Program
Files\Nuance\DragonMedicalServer folder).
n The system file share.
n Any other data or network traffic where latency caused by scanning could affect performance.
Proxy servers
You should exclude all traffic on ports used by Dragon Medical Server and NMS from proxying, including
transparent proxy. Proxy servers add latency and can lack the throughput to properly process Dragon Medical
Server traffic. This can result in apparent loss of audio, lack of response by the Dragon Medical Recognition
Service and other problems. The Dragon Medical Recognition Service must not be connected via a proxy server
to other services such as NMS.
- 34 -
Troubleshooting
n Your system suffers from performance issues.
The use of antivirus software on Dragon Medical Server and Nuance Management Server (NMS) can inhibit
performance. Deactivate antivirus software as a first step in investigating performance issues. For more
information, see: Network guidelines.
Do not use any kind of proxy for network traffic on the ports used by Dragon Medical Server and NMS.
n It is not possible to connect to Dragon Medical Server.
Check that the Dragon Medical Recognition Service and the GSPAdapters processes are running.
Check that the Dragon Medical Recognition Service is not blocked by your firewall.
Make sure that the port number is correctly set.
n The Dragon Medical Recognition Service does not start.
Make sure that the clocks on all speech recognition servers and the database are synchronized and that their
time zones are correctly configured.
Make sure that the Dragon Medical Recognition Service user has the correct permissions. For more
information, see: Permissions.
Check the Windows Event Log. For more information, see the Dragon Medical Server Administration
Guide.
n The Dragon Medical Recognition Service does not start after upgrading Microsoft Windows Server 2008 R2
to Microsoft Windows Server 2012 or 2012 R2.
Uninstall Dragon Medical Server and reinstall it after upgrading the server operating system.
n Issues when stopping/restarting the Dragon Medical Recognition Service.
If you get an error stopping the service, try again.
After the Dragon Medical Recognition Service has been stopped, verify that no GSPAdapterHost processes
are left over (manually end any processes that are still running).
n Dragon Medical Server is in an inconsistent state after an upgrade is canceled.
If subsequent attempts to run Dragon Medical Server setup fail with the error: The system cannot find the
file specified , run the installer package from the command line with the SKIP_CHECKS=1 parameter; for
example:
msiexec /i dragonmedicalserver.msi skip_checks=1
n Uninstalling Dragon Medical Server via the Control Panel does not work. The Windows Installer fails with the
error: Error opening installation log file. Verify that the specified log file location exists and is
writable.
Stop the Explorer.exe process and try again. For more information, see Microsoft Support:
http://support.microsoft.com/kb/2564571
- 35 -
Appendix
Speech recognition topics
Dragon Medical Server supports a predefined list of speech recognition vocabularies, called topics. Topics can
be used to customize the set of active words and are configured via Dragon Medical SpeechKit (for more
information, see the corresponding documentation).
The list of available topics is:
n GeneralMedicine
n ClinicalAdministration
n Cardiology
n Emergency
n InternalMedicine
n MentalHealth
n Neurology
n ObstetricsAndGynecology
n Oncology
n Orthopaedics
n Pathology
n Pediatrics
n Surgery
Nuance Healthcare Support

The Nuance Healthcare Support Platform is a support website where you can find answers to questions,
download software updates, submit online requests for service and contact our support team about Nuance
Healthcare products. In order to use the Nuance Healthcare Support Platform you need to be a registered user.
To register, go to http://www.nuance.com/healthcare/support and click Register now or email
hc.support@nuance.com and provide the following information:
n Your company name.
n The name of your Dragon Medical product.
n Your customer account number (see your welcome kit).
n Your contact information.
Note: The Nuance Healthcare Support Platform is available in English only.
Servers and ports

The Dragon Medical Server, Nuance Management Server (NMS), Nuance Management Center (NMC) and
Text-To-Speech (TTS) servers communicate in the system. Dragon Medical SpeechKit must be able to access
- 36 -
the following ports on the corresponding server:
Dragon Your web NMS TTS Server

Medical server
Server
Dragon Medical SpeechKit <your.Dragon N/A N/A N/A

.NET/COM edition Medical.Server>
port: <your port>
(default: 3712)
Dragon Medical SpeechKit <your.Dragon N/A N/A N/A

Android edition Medical.Server>
port: <your port>
(default: 3712)
Dragon Medical SpeechKit <your.Dragon <your.web.server> N/A N/A

Browser edition Medical.Server> port: 80, 443
port: <your port>
(default: 3712)
Dragon Medical SpeechKit <your.Dragon N/A N/A nsa.nuancehdp.com

iOS edition Medical.Server> port: 443
port: <your port>
(default: 3712)
Dragon Medical One <your.Dragon <your.web.server> <your.NMS.server> N/A

Desktop Application, Medical.Server> port: <your port> port: 80, 443
Dragon Medical Direct & port: <your port> (default: 80, 443)
Dragon Case and Care (default: 3712)
NMC N/A N/A <your.NMS.server> N/A

port: 443
- 37 -
Acknowledgments
This product includes software developed by the Speex project:
n libspeex: http://www.speex.org/
This product includes software developed by the NSpeex project:
n NSpeex: http://nspeex.codeplex.com/
This product includes software developed by the Apache Software Foundation:
n log4net: http://logging.apache.org/log4net/
- 38 -

DragonMedicalServer 3.6 InstallationGuide EN

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DragonMedicalServer 3.6 InstallationGuide EN

Uploaded by

Copyright:

Available Formats

Nuance ®

Upgrading from version 2.0 or higher 29

A typical installation has the following components:

Speech recognition information

System file share

Nuance Management Server

Dragon Medical Server

Nuance Management Server

Microsoft SQL Server

Load balancing between Dragon Medical Server and NMS

Load balancing between NMS and NMC

Dragon Medical Server health check

Nuance Management Server health check

To make speech recognition servers and NMS instances highly available:

Installing the Nuance Management Server

Configuring the NMS file share locations

Launching the Nuance Management Center

Server Cluster Wizard

Dragon Medical license

Speech recognition system

Dragon Medical Server Settings

next to each field for information and examples.

Completing the Server Cluster Wizard

Modifying a Dragon Medical Server Cluster

Organizations and sites in the Nuance Management

Installing Dragon Medical Server

Installing Dragon Medical SpeechKit Browser edition

Dragon Medical SpeechKit

Button controls in virtualized environments

Users and Groups in the Nuance Management

Adding members to a group

Upgrading from versions lower than 2.0

The Dragon Medical Server

Upgrading from version 2.0 or higher

Nuance Healthcare Support

Note: The Nuance Healthcare Support Platform is available in English only.

Servers and ports

the following ports on the corresponding server:

Dragon Your web NMS TTS Server

Dragon Medical SpeechKit <your.Dragon N/A N/A N/A

Dragon Medical SpeechKit <your.Dragon N/A N/A N/A

Dragon Medical SpeechKit <your.Dragon <your.web.server> N/A N/A

Dragon Medical SpeechKit <your.Dragon N/A N/A nsa.nuancehdp.com

Dragon Medical One <your.Dragon <your.web.server> <your.NMS.server> N/A

NMC N/A N/A <your.NMS.server> N/A

You might also like