Jump to: navigation, search

FA Deployment Architecture and Recommendations for Optimal Performance

The following table describes the high-level dimensions controlling the environment used to achieve the results described in Load-Carrying Capacity.

Frontline Advisor Presentation Object Configuration
Object Count
Agents 30 000
Depth (levels) 6
Multiplicity (refers to the average number of teams to which an agent belongs) 1
Agent Groups 5000 (with agents)
Time Profiles 3

The following diagram shows the environment topology used to successfully achieve 1500 concurrent users of Frontline Advisor (FA). "VM" is a virtual machine. The Aggregation Instance performs data aggregation; the Presentation Instance services the dashboard.

Frontline Advisor Environment Topology

Frontline Advisor Presentation Machine and VM Information

The following table describes the characteristics of the hardware and virtualization environment shown in the figure, Frontline Advisor Environment Topology, which supports 1500 concurrent users of FA.

Server # of Processors Processor Type # of Cores Total Cores RAM OS Application
Server 4:


2 Intel Xeon X5675 @ 3.07GHz 6 12 32 GB VMWare ESXi 5.0 Standard 64bit
Server 4:


2 Same as host 4 8 8 GB Windows Server 2008 Standard SP1 64bit Dashboard simulator
Server 4:


2 Same as host 4 8 8 GB Windows Server 2003 Standard SP1 64bit

NOTE: At the time of testing, Genesys supported Windows Server 2003. In your enterprise, be sure to use a Genesys-supported operating system.

Geronimo (FA Presentation instance on Advisors Platform)
Server 4:


2 Same as host 1 2 4 GB Windows Server 2003 32bit

NOTE: At the time of testing, Genesys supported Windows Server 2003. In your enterprise, be sure to use a Genesys-supported operating system.

Configuration Server
Server 4:


2 Same as host 4 8 8 GB Windows Server 2008 R2 Standard SP1 64bit FA Aggregation instance on Advisors Platform
Server 5:


2 AMD Opteron 2439SE @ 2.8GHz 6 12 32 GB VMWare ESXi 5.0 Standard 64bit


2 Same as host 1 2 8 GB RHEL Server 5.7 64bit Oracle database

Recommendations for Frontline Advisor Performance Improvement

The following table describes settings you can change to improve the performance of Frontline Advisor.

Location Sub-directory or File, where applicable Settings
On each FA node <FA home>\geronimo-tomcat6-minimal-2.2.1\bin\setenv.bat
Starting with release 8.5.2, the path to the setenv.bat file changes to <FA home>\apache-tomcat-<version>\bin\setenv.bat.
  GERONIMO_OPTS=-ms128m -mx1024m


  GERONIMO_OPTS=-Xms4g -Xmx8g 

Starting with release 8.5.2, GERONIMO_OPTS becomes CATALINA_OPTS. In addition, the XX:MaxPermSize setting no longer exists. So, starting with release 8.5.2, you make the following change in the setenv.bat file:


CATALINA_OPTS=-ms1024m -mx2048m


<FA home>\geronimo\var\catalina\server.xml Under the <Connectorname="TomcatAJPConnector"> section, add maxThreads="2000"
Starting with release 8.5.2, you do not change the TomcatAJPConnector setting because it does not exist.
On AGA Releases 8.1.5 and 8.5.0:

<AGA home>\conf\wrapper.conf

Release 8.5.1 and later:
conf\run.bat (Windows)
bin\setenv.sh (Linux)

Releases 8.1.5 and 8.5.0:





Release 8.5.1 and later:
In the JAVA_OPTS parameter, change



<AGA home>\conf\inf_genesys_adapter.properties Change
  Server.addp.clienttimeout = 120
  messagerate.numberofmessages.batch = 500


  Server.addp.clienttimeout = 360
  messagerate.numberofmessages.batch = 100000
On each Apache HTTP proxy httpd.conf
  • Uncomment or add the following modules:
    LoadModule deflate_module modules/mod_deflate.so
    LoadModule headers_module modules/mod_headers.so
    LoadModule proxy_module modules/mod_proxy.so
    LoadModule proxy_ajp_module modules/mod_proxy_ajp.so
    LoadModule proxy_balancer_module modules/mod_proxy_balancer.so
    LoadModule proxy_http_module modules/mod_proxy_http.so
  • Add the following block to increase the number of Apache worker threads (note that this is for a Windows-based server):
  <IfModule mpm_winnt_module>
  ThreadsPerChild 512
  MaxConnectionsPerChild 0
  • If you use a Linux server, add the following block:
  <IfModule mpm_event_module>
  StartServer 6
  ServerLimit 32
  MinSpareThreads 150
  MaxSpareThreads 250
  ThreadsPerChild 25
  MaxRequestWorkers 800
  MaxConnectionsPerChild 0
  • Add the following to enable a request response proxy:
    ProxyPass /fa/ ajp://localhost:8009/fa/
    ProxyPass /am/ ajp://localhost:8009/am/
    ProxyPass /admin/ ajp://localhost:8009/admin/
    ProxyPass /am-admin/ ajp://localhost:8009/am-admin/
    ProxyPass /ca/ ajp://localhost:8009/ca/
    ProxyPass /ca-ws/ ajp://localhost:8009/ca-ws/
    ProxyPass /ea-ws/ ajp://localhost:8009/ea-ws/
    ProxyPass /base-ws/ ajp://localhost:8009/base-ws/
    ProxyPass /dashboard/ ajp://localhost:8009/dashboard/
    ProxyPass /nav-service/ ajp://localhost:8009/nav-service/
    ProxyPass /prefs-service/ ajp://localhost:8009/prefs-service/
    ProxyPass /wu/ ajp://localhost:8009/wu/
    ProxyPass /rmc/ ajp://localhost:8009/rmc/
    ProxyPass /gc-admin/ ajp://localhost:8009/gc-admin/
    ProxyPass /ca-xml/ ajp://localhost:8009/ca-xml/

Best Practices for Frontline Advisor Sizing

Use the following notes and best practices for optimizing FA performance:

  • Use Gigabit connectivity between the FA aggregation node and FA presentation node(s).
  • Enable an AJP connection between the Apache HTTP proxy and FA presentation node(s).
  • Allocate as much CPU resource to FA as possible; FA performance is improved if you provide multiple CPU cores and faster clock speeds.
  • Allocate sufficient memory for FA components (Genesys recommends 6GB).
  • You may require multiple AGAs (up to two).
  • Regarding Stat Server performance:
    • Stat Server is a single threaded process. Carefully monitor the CPU usage of your Stat Server(s).
    • Consider adding more Stat Server pairs if a Stat Server is saturating a CPU. You may require up to six pairs of Stat Servers for best performance.

Frontline Advisor Dashboard Age

Dashboard age is a performance measure used in assessing the state and performance/rule processing cycles. It represents the age of the statistics on the dashboard, and includes the following:

  1. Pre-Rollup Delay: The time from the end of the last rollup until the scheduled start of the next rollup.
  2. Rollup Duration: The duration of the rollup + the time to publish to the distributed cache.
  3. Request Response Time (RRT): The 95th percentile of response time for a dashboard request. That is, 95% of the time, a response is returned to a dashboard request after X number of seconds, where X is a constant.

The following table shows results from Genesys’ performance testing. State processing refers to the state metric rollup cycle and performance/rule processing is the performance metric rollup cycle (state and performance metric cycles run independently).

Dashboard Age Results from FA Performance Testing
Measure (seconds) State Processing Performance/Rule Processing
95th Percentile Median Average 95th Percentile Median Average
Dashboard age 18 seconds 12 seconds 12 seconds 70 seconds 62 seconds 62 seconds


Comment on this article:

blog comments powered by Disqus
This page was last modified on April 26, 2018, at 19:32.