There is a use case where an external tool is polling a device. That device also needs to be polled in DataMiner, but then we are "double-polling" it, which should be avoided. To avoid this, we could enable the SNMP agent on the DataMiner element that is polling that device. Then the external tool could poll the DataMiner element assuming the external tool communicates using SNMP. However, if that's not the case, are there any other possible solutions? Next to that, are there use cases in practice where double polling happens (e.g. if no other workaround is possible)?
Thanks!
Hi,
There is no specific feature in DataMiner that would prevent "double-polling" since it happens on the device/data source and we are not in control of that.
In therefore depends on the implementation of the data source if this can be blocked or not.
There have been devices in the past where the device itself could not handle double polling but most are capable of doing this. When another application would be polling the device, DMA would be unable to establish a connection.
It is best practice during an implementation to verify the impact of the driver on the data source to avoid unnecessary high load and could identify problems due to double polling.
Alternatives to offer the customer next to activating the SNMP Agent would be to use the Web API from DataMiner to retreive information from an element.
Is there a specific reason to avoid the double-polling in this use case ? e.g. device cannot handle it ?
Double polling happens in real life. Often DataMiner needs to poll products where older legacy software tools are also still polling it. But you are right to raise this point, as it does require some caution. And this really needs to be looked at case by case, as it largely depends on the type of API / protocol / interface, and on the quality of the implementation on the product-side of it (e.g. I have seen products with SNMP interfaces that easily can deal with this, and others that immediately run into trouble if you do that).
Double polling can be considered, if it is an absolute must, but should be avoided if possible. The other aspect of it is also the control part of the integration, it is definitely recommended to avoid products being controlled by two systems, as this could result in some trouble as well.
The alternative is indeed to have the third party solution use any of the DataMiner north bound interfaces to collect the data that it needs (such as SNMP, but there are various other north bound interfaces). One of the advantages of that is that the third party system does not have to deal with the complexity or quirks of the underlying product, i.e. the interfaces from DataMiner are always standard (e.g. SNMP) irrespective of the underlying API (which could even be non-SNMP).
Hi Koen, thanks for the info. I don’t have more details about the exact use case yet. It was more a generic question that I received during a training session. I will ask for more details during the next training session.