Does anyone have experience to share about High Availability AF Collective
Specifically with the below setup in mind:
1) using Network Load balancing for two AF Servers.
2) using AlwaysOn for the AF SQL servers(running in a Windows Server Failover Cluster).
All comments are welcome!!
In the PI Live Library it is stated that using the Network Load balancing method for AF is "Not recommended with asset analytics, event frames, and notifications"
- Why is this not recommended? Due to difficulty in setting it up or due to instabillity?
PI AF-based high-availability solutions
The following table shows a comparison of PI AF
-based high-availability solutions.
|PI AF-based solution ||Advantages ||Disadvantages |
| Clustered || No real advantages, other than it can help restart the service if it fails. |
- Requires significant initial investment in cluster hardware.
- PI AF server unavailable during cluster failover period.
- No real advantage over having single server service restart itself on failure.
| Network Load Balancing (NLB) || Allows for load balancing across multiple servers using NLB clusters or round-robin DNS. ||Network Load Balancing can be single point of failure if unavailable. |
|PI AF collective (with static load balancing) || Low-cost entry into high availability. |
- Static load balancing across PI AF collective members requires per-client PI AF SDK configuration.
- As of today, no dynamic load balancing is available.
Not recommended with asset analytics, event frames, and notifications.