Tips for choosing a hub server
For the hub server, you must choose a server that has adequate resources and is located for minimal roundtrip network latency.
- Choose a lightly loaded server
- Consider a server that has a light load for operations such as client backup and archive. A
lightly loaded server is also a good choice as the host system for the Operations Center.
Ensure that the server has the resources to handle both its typical server workload and the estimated workload for acting as the hub server.
- Locate the server for minimal roundtrip network latency
- Locate the hub server so that the network connection between the hub server and the spoke
servers has a roundtrip latency that is no greater than 5 ms. This latency can typically be achieved
when the servers are on the same local area network (LAN).
Networks that are poorly tuned, are heavily used by other applications, or have roundtrip latency much higher than 5 ms can degrade communications between the hub and spoke servers. For example, roundtrip latencies of 50 ms or higher can result in communication timeouts that cause spoke servers to disconnect or reconnect to the Operations Center. Such high latencies might be experienced in long-distance, wide area network (WAN) communications.
If spoke servers are a long distance from the hub server and experience frequent disconnects in the Operations Center, you can increase the value of the ADMINCOMMTIMEOUT option on each server to reduce the problem.
- Verify that the hub server meets the resource requirements for status monitoring
- Status monitoring requires extra resources on each server on which it is enabled. The resources
that are required depend primarily on the number of clients that are managed by the hub and spoke
servers.
Verify that the hub server meets the resource requirements for processor usage, database space, archive log space, and I/O operations per second (IOPS) capacity.
A hub server with high IOPS capacity can handle a larger amount of incoming status data from spoke servers. Use of the following storage devices for the hub server database can help meet this capacity:- An enterprise-level solid-state drive (SSD)
- An external SAN disk storage device with multiple volumes or multiple spindles under each volume
In an environment with fewer than 1000 clients, consider establishing a baseline capacity of 1000 IOPS for the hub server database if the hub server manages any spoke servers.
- Determine whether your environment requires multiple hub servers
- If more than 10,000 - 20,000 client nodes and virtual machine file spaces are managed by one set of hub and spoke servers, the resource requirements might exceed what the hub server has available.Consider designating a second server as a hub server and moving spoke servers to the new hub server to balance the load.