CJ Goldsmith
07-28-2009, 05:57 PM
If you log into Traverse as a super user and navigate to:
SuperUser -> Global Config -> Data Gathering Engine
There is an optional parameter there to enable 'Strict OID Grouping'. I wanted to take a moment out to explain in detail what this option actually does and what the trade off involves when you decide to enable it.
First it is important to understand an important Traverse optimization. A central part of the Traverse product is the internal scheduler which is responsible for dynamically scheduling all tests (ping / snmp / wmi) in a way that they don't burden the network and at the same time provide test results in a fashion that allows for clear and accurate reporting on the status of your network. Under normal conditions we want the scheduler to send out tests over the network in a very evenly distributed fashion.
The scheduler normally does tests in batches and optimizes within those batches so all tests grabbed for device A are sent out at the same time within that batch, all tests for device B are sent out at the same time...
etc. The key concept being that batches are a fixed and evenly distributed size.
For Example:
Lets say that we have 4 devices (Devices A, B, C, & D) where:
Device A has 115 tests
Device B has 60 tests
Device C has 50 tests
Device D has 30 tests
And lets say that the scheduler has 5 batch periods in which to provision all tests for all devices. It might do it in a way that looks like this (note that the #tests provisioned per batch never exceeds 55):
http://forums.zyrion.com/images/posts/strict_oid/EvenDist.jpg
Once you turn on the 'Strict OID Grouping' you are forcing the scheduler to place all tests for a single device into a single batch, this forces all communications to a device within a period to be sent at the same time; which can improve performance with devices that are experiencing network lag.
Because you are removing control (disengaging the network optimization) from the scheduler and can now end up with test queue build up and periodic network congestion coming from the DGE. Which is why this option is turned off by default.
Again in our example lets now assume that all tests the 'Strict OID Grouping' option has been turned on and all tests for each device need to be provisioned within the same batch (Note that batch 1 is now performing 115 tests and batch 4 is performing zero):
http://forums.zyrion.com/images/posts/strict_oid/GroupDist.jpg
SuperUser -> Global Config -> Data Gathering Engine
There is an optional parameter there to enable 'Strict OID Grouping'. I wanted to take a moment out to explain in detail what this option actually does and what the trade off involves when you decide to enable it.
First it is important to understand an important Traverse optimization. A central part of the Traverse product is the internal scheduler which is responsible for dynamically scheduling all tests (ping / snmp / wmi) in a way that they don't burden the network and at the same time provide test results in a fashion that allows for clear and accurate reporting on the status of your network. Under normal conditions we want the scheduler to send out tests over the network in a very evenly distributed fashion.
The scheduler normally does tests in batches and optimizes within those batches so all tests grabbed for device A are sent out at the same time within that batch, all tests for device B are sent out at the same time...
etc. The key concept being that batches are a fixed and evenly distributed size.
For Example:
Lets say that we have 4 devices (Devices A, B, C, & D) where:
Device A has 115 tests
Device B has 60 tests
Device C has 50 tests
Device D has 30 tests
And lets say that the scheduler has 5 batch periods in which to provision all tests for all devices. It might do it in a way that looks like this (note that the #tests provisioned per batch never exceeds 55):
http://forums.zyrion.com/images/posts/strict_oid/EvenDist.jpg
Once you turn on the 'Strict OID Grouping' you are forcing the scheduler to place all tests for a single device into a single batch, this forces all communications to a device within a period to be sent at the same time; which can improve performance with devices that are experiencing network lag.
Because you are removing control (disengaging the network optimization) from the scheduler and can now end up with test queue build up and periodic network congestion coming from the DGE. Which is why this option is turned off by default.
Again in our example lets now assume that all tests the 'Strict OID Grouping' option has been turned on and all tests for each device need to be provisioned within the same batch (Note that batch 1 is now performing 115 tests and batch 4 is performing zero):
http://forums.zyrion.com/images/posts/strict_oid/GroupDist.jpg