Hi Brains trust.
Currently we want to conserve data by polling a device from one set of DMA servers instead of polling a device from a Prod server, Test Server and a Trending server. Also Polling SNMP devices too much has caused some devices to lock up and require a physical power cycle onsite.
The Replicate feature available when you create or edit an Element is the perfect solution. That way as long as the Protocol used is the same across all platforms, it works a treat.
A question was raised though as to how much load on the DMA's would this produce if I am polling from one set of DMA's to another.
For example, I have some 500 elements on our trending server that I want to replicate the data from on the Production servers. The Trending Elements only poll for data at 1-5 min intervals whereas the Production elements are polling the Devices with Params every 10 secs, 60 secs and 1 hour intervals.
Hi Nathan,
it's hard to give you exact numbers, as there are a lot of variables involved. But a replicated element should be on average lower in load on the DMA it is hosted, as compared to standard element on a DMA (that needs to be polled by that DMA).
The main reason is that if you replicate a DMA, all the values for the metrics for that element are transferred once onto the target DMA that has the replicated element, and from then onwards the communication exchanged is based on so-called change-events. Meaning that updates from the source DMA to the target DMA are only happening when the value of a metric effectively changed.
So let's say that you have an element with 5 metrics, which are polled each 5 seconds by the host DMA. When you replicate that element, all values for those 5 metrics will be send to the target DMA. After that, the target DMA will send new values to the host DMA only when a value of one of those 5 metrics changed. Let's say that 4 metrics are static (e.g. setting of the device that rarely change), then this will cause no further traffic. And let's say that one metric is a level, that changes now and then. That metric will be polled each 5 seconds by the host DMA maybe, but only when its value effectively changes compared to the previous poll, only then it will be also send to the target DMA to update it over there (and that could be depending on the nature of the metric maybe only once each few minutes or so, albeit being polled each 5 seconds).
In conclusion, the load of a replicated element should be much lower, but it also depends on the nature of the elements and its metrics (i.e. how quickly do the values change over time).
Hope this helps answering the question. If not, let us know and we can further delve into it.