(imported topic written by murtasma91)
BigFix recommends keeping the amount of clients on a relay to 1,000. Currently we are exceeding this across our relays but not by much. I’m not really noticing a performance difference when we started to exceed the recommendation.
How many clients do you typically have on your relays?
What are some good performance metrics to measure to determine if we need to add additional relays to our infrastructure?
(imported comment written by JackCoates91)
One thing to be concerned about is failover behavior. Let’s say that Relays A and B are both running at 85% capacity. Relay A gets shut down for maintenance, and its clients roll over to Relay B. Relay B is now attempting to operate at 170% capacity, which may not work as well as we’d all like.
(imported comment written by MattBoyd)
I think the answer is always going to be “it depends.” There are many factors:
Storage performance
Network bandwidth
CPU
Memory
Number of actions
Number downloads
Number of clients on relay X that need download Y for action Z
If your BES infrastructure isn’t heavily used, you can put more than 1000 clients on a single relay without breaking much of a sweat. However, an increase in deployed actions, analyses, and large client downloads can bring the relay to a crawl, which means it takes longer for clients to finish the actions and report in. Disk IO is tends to be the first bottleneck that relays encounter.
For performance metrics, you could try tracking the following with the performance monitor that’s build into Windows:
Network Utilization - \Network Interface\Bytes Total/sec
CPU Utilization - \Processor Information%Processor Time (per core might be better)
Disk Activity - \PhysicalDisk\Disk Bytes/sec
Current Disk Queue Length - \PhysicalDisk\Current Disk Queue Length
Be VERY careful when configuring data collectors with performance monitor. They can fill up the disk quickly 
If possible, I’d store them on a non-system volume.
You’d need to monitor these counters for a week or two in order to figure out how well your relays are handling the content/traffic that’s thrown at them. Even then, you should take into consideration any times during the year where you may see an uptick in BES utilization. For example, our relay utilization explodes prior to the beginning of the fall semester when we’re rebuilding labs and deploying new software via BigFix.
My 2 cents. I’d love to hear other suggestions for performance metrics or relay performance measuring.
(imported comment written by BenKus)
FYI… I wrote this one a while back:
http://support.bigfix.com/cgi-bin/kbdirect.pl?id=1742
Bottom line is that relays can handle way more than 1000 agents, but we still recommend only 1000 agents per relay to be consistent with the idea that relays are shared servers and the fact that relays should be able to very quickly and simultaneously service all agents.
Ben
(imported comment written by StacyLee)
we have on average 1600 /per relay. However not all clients are connected and powered on all at the same time, performance is still pretty good for us.
Each relay is a server class poweredge server with 15K disks and 1GB network connection.