HA: QoS during Synchronization Process
- Explanation
- April 30, 2015
- Download as PDF
Description
When StarWind Virtual SAN operates in High Availability (HA) mode, data synchronization continuously occurs between partner nodes to ensure data consistency. During heavy workloads, this synchronization process may compete with client I/O requests (reads and writes from applications and virtual machines).
To maintain a balance between synchronization and service responsiveness, StarWind implements a Quality of Service (QoS) mechanism through the Synchronization Priority setting. This feature allows administrators to control whether the system prioritizes synchronization speed or client request handling.
- If the priority for all partners is set to Sync process, synchronization will complete faster but may temporarily reduce the speed of servicing client requests.
- If the priority is set to Client requests, StarWind will prioritize client I/O, temporarily suspending synchronization to ensure minimal impact on user experience.
This mechanism provides flexible control over system behavior, allowing synchronization to proceed efficiently without noticeably degrading client performance.
To change synchronization priority, follow the steps in the help page here: https://www.starwindsoftware.com/help/ChangingSynchronizationPriority.html
NOTES
- Setting synchronization priority to Sync process accelerates synchronization but can affect client I/O performance under heavy load.
- Setting priority to Client requests improves responsiveness for users and services but increases total synchronization time.
- Proper QoS settings should be chosen based on workload type and performance requirements.
Conclusion
The Synchronization priority feature in StarWind Virtual SAN enables administrators to fine-tune the balance between synchronization efficiency and client request performance.
This flexibility ensures that StarWind HA environments can be optimized for different workloads — from systems prioritizing uninterrupted client access to those focused on rapid data synchronization between nodes.
