Is OneAgent selection of AG based on load and latency?
Describing how OneAgents determine which AG to connect to, the page at the link below states that "If more than one such ActiveGate is available, OneAgents will try to switch between the available ActiveGates on a regular basis to achieve proper load balancing."
I had previously thought (assumed) that the OneAgents would be checking network latency to the AGs and would pick the one with lowest latency (within AGs of a specific type, such as Environment AG). Is that not the case, or are BOTH load and latency taken into consideration?
As I know this is only based on AG load, not latency. When we are making deployments, group of agents has access via network to only one AG. If there are some extra considerations here I would be happy to get such informations as well :)
Sebastian
If that is the case then my concern is that agents will often end up not connecting to the "local" AG (i.e., one in the same data center as the agent) but will instead be sending data to AGs that are located in other data centers... not very efficient from a network perspective
According to this doc, OneAgent will prefer AG that is closer. In such case this shouldn’t be issue for you.
Sebastian
I understand the hierarchy of preference between the Environment and Cluster AGs and of these over a direct connection to the cluster, but that view of closeness does not relate to the actual network distance.... so when we have multiple Environent AGs (as we do) the documentation does not appear to state that the one that is actually closest or has the least latency will be selected.
Latency might be very volatile. I think balancing is only dependent on the number of connections or on amount of data.
Perhaps, although I would think that there would often be significant differences between geographically remove sites... so it surprises me that the algorithm does not take that into consideration... But thanks for confirming that it is based on load (connections & traffic volumes) only.
