Not official scaling advice
The numbers given here are estimates from real environments. However, they may vastly differ depending on your setup. We strongly recommend that you engage our professional services department if scaling is an issue.Terminology
These estimates are based on "checks". A check is a defined plugin that runs on a schedule, assumed to be five minutes. In terms of scaling, it does not matter if this check belongs to a service, or a host. Checks also have a variable amount of bandwidth usage since it depends on the amount of output they generate.
Estimates
Master <-> Poller
Given 2,800 checks in total and a five minute check interval, you will likely need bandwidth in the range of 0.5 - 1.0 Mbit/s.
As a conservative estimate, you can therefore expect a need of about 0.36 Mbit/s of bandwidth per 1 000 checks.
Peer <-> Peer
Given 23,223 checks in total and a five minute check interval, you will likely need bandwidth in the range of 3 - 6 Mbit/s.
As a conservative estimate, you can therefore expect a need of about 0.26 Mbit/s of bandwidth per 1 000 checks.
Caveats
These estimates were calculated using the iptraf tool on port 15551, which means that only Merlin traffic is included. Any bandwidth consumed by object configuration or other file synchronization is not included.
SSH is hard to estimate since there may be any amount (or none) of additional files being synced, that may change on any interval, meaning that the payload can vary vastly in size.
Comments
0 comments
Please sign in to leave a comment.