Very large customer network

This customer is a very large global enterprise company with a simple network architecture but very large numbers of devices. The purpose of this installation is to manage this customer network by alerting the operations staff to major failures and to support short-term capacity planning.

The following sections describe this network in greater detail and provide suggestions for a Network Manager deployment to meet the needs of this network.

Description

Network management is done from a central location and from regional locations. The network is very large and contains over 15,000 network devices and critical servers. Network devices come from multiple vendors. The devices fall into two categories:
  • Network device infrastructure with interface counts in the range of 30 or more per device.
  • Managed devices with 1-2 interfaces per device.
Most of the devices are in the second category, managed devices. To manage a network of this size, the network is partitioned for management on a geographical basis.

Within this environment, the following example conditions apply:

  • There are 5 - 20 active GUI clients.
  • Polling:
    • Chassis ping polling at two to 5-minute intervals.
    • SNMP polling at 15 minutes or longer.
    • SNMPv1 data collection
  • Other major Tivoli® products that are to be integrated with the system, other than the required Tivoli Netcool/OMNIbus:
    • IBM® Tivoli Business Service Manager
    • IBM Tivoli Application Dependency Discovery Manager

Network Manager deployment

Assistance from an experienced IBM services group or qualified IBM Business Partner is highly advisable for a successful deployment. Multiple domains are needed, supported by a collection of individual servers, or running together on a very large system. After you survey the network to be managed, break up the network into sections that can be managed, and then assign each of the sections to a domain. In addition to the multiple-server deployment description provided elsewhere, the following deployment settings are appropriate for this type of environment.
  • Multiple network domains.
  • Platform selections: Linux® and UNIX.
  • Large systems (many processors and very large amounts of memory) can host multiple domains if the memory allocations and processor counts are acceptable.
    • Memory: 32-64 GB per domain
    • Processors: 4-8 per domain, depending on workloads
  • Db2® or Oracle RDBMS used for the NCIM database.
  • Two polling engines for each domain:
    • Use the default ncp_poller process for chassis ping.
    • Create a separate ncp_poller for the SNMP polls.
  • Individual process memory limitations are a factor in this environment. If you are using AIX®, enable large memory access.
  • Client system: single processor, 4 GB of memory, supported JRE and Internet browser
  • IPv6 dual stack support is required if workstations or network devices have IPv6.