By default, SQL Server will use jumbo sized IOs (typically of 1MB) to perform backup and restore operations. These can adversely affect the IO latency for your array. Whenever possible, use MAXTRANSFERSIZE=65536 on your BACKUP or RESTORE statements. This could make your backups and restore take a tad longer, so test accordingly. If you are leveraging Transaction Log Shipping, you might have no way to tune this, as sqllogship.exe doesn’t offer a means to tune this parameter. If log shipping backups/restores are affecting your performance, considering rolling out scripts to perform log shipping manually.
If you are not achieving good throughput from your backup/restore operations after using the MAXTRANSFERSIZE=65536 parameter, considering bumping up the number of buffers available to the process by using the BUFFERCOUNT parameter. In our testing, a 1.5TB database backed up in 24 minutes, 38 seconds at default settings, and 25 minutes, 6 seconds with 64Kb IO block size and BUFFERCOUNT=2048.