Unsupported content

 

This version of the product is no longer supported. However, the documentation is available for your convenience. You will not be able to leave comments.

Disk IO Performance Guidelines

BMC publishes performance data to help you determine if a particular system used to host BMC Discovery is powerful enough to discover the estate at the desired frequency. The performance of a BMC Discovery appliance (whether Physical or Virtual) is dependent on a number of characteristics (such as number of CPUs and CPU speed), but the performance profile of the I/O subsystem is particularly important and is harder to quantify. When actively scanning, BMC Discovery places a heavy read and write load on the database. Consequently the performance of BMC Discovery benefits greatly from giving it the fastest disks possible; the slower your disk I/O, the slower BMC Discovery will be. As a rule of thumb, if you are spending money on infrastructure to support BMC Discovery, then choosing to spend it on fast storage first is a good strategy. See appliance performance.

The metrics published on this page are all gathered on appliances available in BMC datacenters, or hosted by public cloud vendors.  It is not possible to exactly mimic the hardware characteristics of any particular appliance a customer may use. These metrics are therefore guidelines only.

A BMC Discovery appliance where the I/O subsystem does not match or exceed the performance shown below may not achieve the same levels of discovery, reporting, and consolidation performance.

IoZone

To provide guidance on the necessary I/O performance, performance profiles have been produced based on the powerful disk I/O performance measurement tool called IoZone. IoZone is pre-installed on the BMC Discovery appliance. You can configure IoZone to run various customized I/O tests. The following benchmark data is produced using an IoZone test that attempts to mimic the disk access profile of the BMC Discovery product as closely as possible. 

BMC Discovery automatically scales the number of ECA Engines based on the number of CPUs and RAM available. Whilst actively scanning, each ECA Engine will drive database load. Consequently the specific IoZone command run should reflect the number of ECA Engines present on your appliance; our recommendation is to set the number of threads in the IoZone test to be twice the number of ECA Engines. 

You must ensure there is sufficient free space on the drive under test

The Io Zone test will concurrently manipulate many 1Gb files, the number of which depends on the number of threads selected.  The amount of free space required will therefore vary according to this formula: 2 * ECA Engine Count * 1Gb


  1. Stop the tideway services
  2. As the tideway user, create an "iozonetest" directory in your datafile drive and cd into it
  3. Execute the iozone test command tuned to your specific appliance configuration. For example:

/opt/iozone/bin/iozone -t <number of threads> -s1g -r32k -i0 -i2

where number of threads is based on the number of ECA Engines on your appliance.

 Stopping the BMC Discovery services before running the iozone tests avoids them being impaired by parallel disk activity.

The results of the command are returned to stdout on completion of the test. For more information about the options used, see the IoZone Documentation (PDF).

The IoZone test in turn executes the Consecutive Write, Consecutive Read, Random Write, Random Read tests. The random tests reflect the BMC Discovery IO access profile. The Consecutive tests are not used but cannot be disabled.

An example of the Disk I/O performance seen on an Amazon Web Services (AWS) instance is shown, which is indicative of the throughput you might expect. The instance uses 40GB Amazon Elastic Block Store (Amazon EBS) gp3 volumes.


Number of threads

Initial Writers

Avg throughput per process

Kb/sec

Initial Re-writers

Avg throughput per process

Kb/sec

Random Readers

Avg throughput per process

Kb/sec

Random Writers

Avg throughput per process

Kb/sec

30139655138123

124212

137835


Was this page helpful? Yes No Submitting... Thank you

Comments