|Article Number: 503583||Article Version: 3||Article Type: Break Fix|
Host read and write response times extending into hundreds of milliseconds, on a single engine VMAX 450F.
Investigation with inlines showed that both FAs showed continuous 100% busy, although the FAs were doing only a few hundred IO/s.
The long response time was caused by extremely deep FA queuing and low IO execution rate.
Customer had set host limits of only 500 IO/s and 10MB/s on the storage group that contained almost all devices in the system. When this was adjusted to the normal 40,000 IO/s and 3000 MB/s setting for the SG, the performance returned to normal.
Customer had observed that a large periodic IO/s burst on the main storage group in the frame affected the response time of another small but critical SG sharing the same FAs. They had progressively reduced the host IO/s and host MB/s limits on the main SG, attempting to reduce the impact on the second SG.
The host IO/s limits for the main SG were returned to the original 40,000 IO/s and 3000 MB/s. Performance returned to normal. The IO/s approached 40,000 only during the periodic burst and was limited only briefly then, causing no performance issue.
Solution is to increase the MB/s limit using the Solutions Enabler symsg -bw_max command.