Service Center is resource-intensive. This means application servers should be dedicated to Service Center web applications and Windows Services. Sharing an application server with other applications results in contention for resources, particularly memory and disk throughput, which can significantly impact the performance of all applications on the server.
Hosted Application Server Capacity Guidance
Hardware Minimums | Capacity |
---|---|
8 GB of RAM, 4 vCPU |
|
16 GB of RAM, 8 vCPU |
|
32 GB of RAM, 8 vCPU |
|
For maximum performance, the following hardware configurations are also recommended:
- A paging file that is equal to at least the total amount of RAM, which is recommended for stability.
- Application servers should be scaled up to a maximum of 32 GB of RAM before adding more application servers. This maximum is recommended due to the overhead for hosting the infrastructure code that launches and houses the VSC application domains, and because it is more efficient to have fewer scaled-up servers. Additionally, having fewer servers reduces the time to upgrade Barracuda RMM. However, if an application server has 32 GB of RAM and more RAM is needed, it is recommended to add a server. This ensures the server stays within the Windows Standard Edition memory limits, and spreads the load across multiple servers.
- If you create virtual application servers, ensure that memory ballooning is not enabled.
- If you are using more than on application server, you must set up HTTP/HTTPS load balancing with sticky sessions configured.
- A Gigabit Ethernet LAN connection is required for connections between application servers, SQL Server, and the SQL Report Server.
- It is not recommended to have more than 7 application servers; if more are needed, you should create another environment.
Server Center Worker Processes Management
The following worker process configurations are recommended:
To avoid memory fragmentation, ensure that Service Center worker processes do not exceed 2.5 GB of RAM. When a Service Center worker process surpasses this limit, to minimize the number of VARs impacted by issues or performance problems, you can move the VARs to another worker process or create a new worker process.
- As a general rule, the number of worker process per server should not exceed twice the number of logical CPUs. For example, if a server has 4 logical CPUs, there should be a maximum of 8 worker processes.