Network calculator for vR Ops, not right this moment. It will depend on the object count and collection volume (# metrics). But it also depending what, exactly, each MP is actually pulling. Some MP are heavier than other with regarding to API calls and what they're gobbling up from the data source. However, if we're talking just the data we're keeping, call it 24 bytes per metric. Assuming you're collecting 1-million metrics, it'll end up being about 22.8MB per 5 min. Once you throw an RC in there, it'll handle the dirty part of the data collection and normalize the data down to something efficient for transfer to the Analytic nodes. So, an RC certainly can't hurt at a remote site that's got a so-so WAN link between Analytic cluster nodes and data source. Here are some guidelines -
Between the Analytics cluster nodes and Remote Collectors, make sure your WAN latency is less than 200ms.
From a throughput perspective, the Analytics Node-to-RC network utilization will be a bit lower than if you're running Analytics Nodes-to-data sources. This is because the RC performs the raw data collections from a data source, then normalized and cleans up the data before it sends it back to the Analytics node.
Another consideration is the queuing.. the remote collectors will queue data and retries for 2min if it loses connectivity to the Analytics nodes. After that 2 min, it'll drop the queued data and pull fresh from data source on next collection interval and repeat the cycle. This gives you a little buffer on those unreliable WAN links.
So, no downside for RCs in WAN scenarios, we do them as rule of thumb if you've got anything other than a high quality, fat WAN link.
If you find this or any other answer useful please mark the answer as correct or helpful.
