I have some feedback from EMC, their assessment is both the READ and WRITE IO sizes are causing the high latency on the LUN. This is the summary of the solution:
For time-sensitive application, as a workaround increasing the tier for
LUN573 will improve its performance to a certain degree.
Right click LUN and select Tiering Policy->Highest Available
The final resolution is to adjust IO size to make LUN more responsive.
We are experiencing a similar problem wherein a VNX LUN allocated to a ESX host is shared with 5 VM's. One of the VM is hosting SQL database and the data is this SQL database is updated by SQL server Transactional Replication. The VNX LUN is experiencing high response time ( ~50ms ) but having nice service time (under 10ms). SQL host is getting IO timeout errors. Your above comment suggests that some tuning needs to be done on SQL side if SQL Server Transactional Replication is in place. COuld you please huide further on this, What exactly needs to be done ? Or if you have some white paper/best practices for it. Appreciate your help. Thanks