In a 4-node Hyper-V failover cluster using Cluster Shared Volumes (CSV), you observe high read latency during maintenance windows when a CSV enters redirected I/O. You want to reduce read impact for VMs during these periods and in general. What should you configure?

Increasing the CSV cache size and ensuring it's enabled is the correct solution. CSV cache uses system memory on each cluster node to cache read-ahead data for virtual machines. When a CSV enters redirected I/O mode (e.g., during maintenance), all I/O for that CSV is routed through the coordinator node. A larger, active CSV cache on the non-coordinator nodes can absorb more read requests locally, reducing the amount of I/O that needs to traverse the network to the coordinator node, thereby mitigating read latency. Disabling SMB Multichannel would negatively impact performance, as it reduces available bandwidth for SMB traffic, including CSV I/O. Changing the CSV access mode to File System Redirected permanently would cause all I/O to be redirected, leading to persistent performance degradation. Moving VHDX files to local disks is impractical and defeats the purpose of a failover cluster and CSVs. Enabling write-back cache on the NTFS volume is not a standard or recommended practice for CSVs and primarily affects write performance, not read latency during redirected I/O.