Cisco Cisco Administrative Workstation Installation Guide
Chapter 10 Planning for ICM Platforms
ICM Platform Considerations
10-4
Cisco ICM Enterprise Edition Pre-installation Planning Guide Release 6.0(0)
Processor Utilization
As a general rule for all ICM nodes, processor utilization should be kept below 60
percent at the maximum expected call load on the system. This is needed in order
to smooth out call request “spikes” as well as to allow enough reserve capacity to
perform activities such as re-synchronization and background cleanup. Non-ICM
software can make up a part of the 60 percent maximum load. The processor
utilization figure (60 percent) covers all software running on the platform.
percent at the maximum expected call load on the system. This is needed in order
to smooth out call request “spikes” as well as to allow enough reserve capacity to
perform activities such as re-synchronization and background cleanup. Non-ICM
software can make up a part of the 60 percent maximum load. The processor
utilization figure (60 percent) covers all software running on the platform.
In addition to the utilization requirement, it is necessary that no software on the
system run at a priority equal to or higher than the ICM software for more than
100 milliseconds in uninterrupted bursts. In other words, the ICM software needs
to run on the system at least as frequently as once every 100 milliseconds. This is
usually not a problem unless device drivers or other kernel-level software is
installed, or process/thread priorities have changed incorrectly.
system run at a priority equal to or higher than the ICM software for more than
100 milliseconds in uninterrupted bursts. In other words, the ICM software needs
to run on the system at least as frequently as once every 100 milliseconds. This is
usually not a problem unless device drivers or other kernel-level software is
installed, or process/thread priorities have changed incorrectly.
Paging Requirements
The most time-critical component of the ICM system, the CallRouter node, must
not be delayed due to disk I/O (that is, paging). The only disk I/O that should be
occurring on ICM machines is for log file writes and database I/O. The database
I/Os occur on Logger and Distributor AW machines. The simple rule is to provide
enough main memory so that the entire working sets of critical processes remain
in memory. Most customers achieve this with 256 MB of RAM, but larger
configurations may require more RAM. The database platforms (Loggers,
Distributor AWs, and ICM Gateway SQL machines) should have enough main
memory so that all first level index pages are kept in main memory cache.
not be delayed due to disk I/O (that is, paging). The only disk I/O that should be
occurring on ICM machines is for log file writes and database I/O. The database
I/Os occur on Logger and Distributor AW machines. The simple rule is to provide
enough main memory so that the entire working sets of critical processes remain
in memory. Most customers achieve this with 256 MB of RAM, but larger
configurations may require more RAM. The database platforms (Loggers,
Distributor AWs, and ICM Gateway SQL machines) should have enough main
memory so that all first level index pages are kept in main memory cache.
Logger Expansion
The Logger platform you order may include a combination of internal and
external SCSI hard drives. As your call center enterprise grows, your database
requirements will typically grow as well. You might have more services, skill
groups, and routes in your configuration, and you might be routing more calls
each day. This will result in more historical data being stored in the central
database.
external SCSI hard drives. As your call center enterprise grows, your database
requirements will typically grow as well. You might have more services, skill
groups, and routes in your configuration, and you might be routing more calls
each day. This will result in more historical data being stored in the central
database.