Use > Sizing > Sizing Reference for Automated Service Modeling > Scenario 2. Service Discovery and Universal Discovery

Scenario 2. Service Discovery and Universal Discovery

This section describes the scenario that you already have Universal Discovery and want to introduce Service Discovery.

Objectives

The following table lists the expected time to complete all service discoveries based on the data we collected in performance tests.

Number of Services First Run of Service Discovery Scheduled Rerun of Service Discovery
<100 <8 hours <8 hours
<500 <24 hours <24 hours

Note Scheduled Rerun of Service Discovery means rerun all service discoveries on a periodic basis, and the interval can be adjusted.

Test Environment

The performance tests are performed on an environment that meets the following preconditions:

  • One service consists of less than 30 Running Software instances, which run on different hosts.
  • One host can be discovered by the Host Discovery by Shell job within two minutes if everything works as expected.
  • Existing UCMDB data has less than 30,000 nodes, and less than 20 million CIs and relationships.
  • Service Discovery and Universal Discovery never run in the same time window. For more information, see Discovery Schedule.

Note  

  • This scenario is slower than Scenario 1, because the large amount of existing UCMDB data increases the workload of processing the results sent from the Data Flow Probes.
  • If the actual environment cannot meet these preconditions, the Service Discovery performance may not reach the above objectives. In this case, you can adjust several configuration settings to improve the Service Discovery performance.

  • The numbers of CIs and Relationships reported by Service Discovery are shown in the table below.

    CI Types

    100 Services

    500 Services

    Node

    6000

    30000

    IP Address

    6000

    30000

    J2EE Domain

    6000

    30000

    Running Software

    12000

    60000

    J2EE Application

    6000

    30000

    Service Connection Point

    6000

    30000

    IP Service Endpoint

    6000

    30000

    Consumer-Provider Link

    12000

    60000

    Composition Link

    18000

    90000

    Membership Link

    6000

    30000

    Ownership Link

    12000

    60000

    Usage Link

    6000

    30000

Discovery Schedule

We strongly recommend that you schedule Service Discovery and Universal Discovery, so that Service Discovery, Universal Discovery, and Integration Service can run in different time slots.

Note An overlapped execution of Service Discovery and Universal Discovery may,

  • Slow down both types of discoveries.

    Because the UCMDB Server uses the "first-come, first-served" mechanism, a later-arrived job must wait to be executed until the previous jobs are finished.

  • Further increase the workload of the UCMDB Server and the Data Flow Probe.

For example, assume that you run Universal Discovery jobs once every day from Monday to Sunday before introducing Service Discovery. To avoid the overlapped execution of Service Discovery and Universal Discovery, consider the following policies:

  • Define a job execution policy that disallows the execution of Universal Discovery jobs on weekends.
  • Change the Service Discovery schedule so that Service Discovery is allowed to run on weekends.

For more information about job execution policies, refer to the Data Flow Management section of the UCMDB Help.

Configurations

This section lists the hardware and configuration settings that we used in performance tests. The real numbers may vary depending on the actual environment.