I am currently re-architecting our Service Desk Infrastructure. One of the challenges I am facing is deciding is in what network I should place my application servers. I have the choice between two different networks, they are:
1.Place the servers in the same network as the clients. This will facilitate speedy communication between the clients and the servers. 2.Place the servers in the same location as my monitoring servers. The Service Desk agent is installed on my monitoring servers and this will facilitate communications between SD and monitoring.
The issue I am most confused about is the communications that occur between the Service Desk servers and the Service Desk agent. HP support tells me that the agents need to be able to communicate with the servers via broadcast, when I pressed them for details on this they were not able to answer my question satisfactorily. I have searched through the manuals for information on broadcast and multicast and I have not come up with anything. Can anyone point me in the direction of some detailed documentation regarding the communications between the Service Desk agents and servers? How have other organizations handled this? Thanks in advance.
HI Agent need for execute some task from Application server, not from monitoring system, on Monitoring system need set Event system form SD. move Agent to Applic server and place to client subnet, because request/traffics from client much more, then doing events from monitoring systems.
The agent to server communication is multicast udp (NOT broadcast) by default. So if the server and agents are not in the same subnet, your router needs to pass multicast information. Server to agent is always unicast tcp.
But you can configure the agent-to-server direction to be unicast as well, by specifying the server parameter on agent startup. But if you have multiple servers, this may not work.
Why not use one app server for incident monitoring/integrations (don't accept client connections) and one app server for client connections = less config on agent/server end and good performance for client/server end... Just a thought.
The load from client activity will most likely be more than the agent/monitoring side of the house if you have tuned your policies and are not seeing thousands of un-ness incidents created on an hourly basis etc...