Dell EMC Solutions Enabler 9.2 SRDF Family CLI User Guide

Write-pacing behavior varies by the type of pacing, the SRDF topology (2-site, cascaded, concurrent), and OS version.

Group-level pacing considerations

• Only the group-level pacing values configured for the SRDF group on the R1 side of the SRDF/A session are used.
• In a cascaded SRDF environment:
  • With Enginuity 5876 Q4 2012 SR and later, group-level write pacing is supported on both the R1->R21 and R21->R2 hops of the relationship.
• In a concurrent SRDF/A environment, group-level pacing is supported on both mirrors of the concurrent R1. In this case, write pacing calculations are performed independently for the two SRDF/A sessions, and the host write I/Os sessions are subject to the greater of the two calculated delays.

Device-level pacing considerations

• Only the device-level pacing values configured for the SRDF group on the R1 side of the SRDF/A session are used.
• In a cascaded SRDF environment:
  • With Enginuity 5876 Q4 2012 SR and later, device-level write pacing is supported on both the R1->R21 and R21->R2 hops of the relationship.
• There is no exemption from device-level pacing as there is for group-level pacing, and the R1 group-level exempt state does not affect device-level pacing.
• In a concurrent SRDF/A environment, device-level pacing is available on both mirrors of the concurrent R1. In this case, write pacing calculations are performed independently for the two SRDF/A sessions, and the host write I/Os sessions are subject to the greater of the two calculated delays.
• If both group-level pacing and device-level pacing are active for an SRDF/A session, the group-level and device-level delays are calculated independently, and the greater calculated value is used for pacing. Note that as many as four different calculation results may be taken into account for a concurrent R1 device with both mirrors operating in asynchronous mode (group-level pacing for each mirror, device-level pacing for each mirror), using the greatest calculated delay in the calculation.

Operations

SRDF/A write pacing bases some of its actions on the following:

• R1 side cache usage
• Transfer rate of data from transmit delta set to receive delta set
• Restore rate on the R2 side

SRDF/A group-level write pacing can respond to the following conditions:

• The write-pending level on an R2 device in an active SRDF/A session reaches the device's write-pending limit.
• The restore (apply) cycle time on the R2 side is longer than the capture cycle time.

The enhanced group-level write pacing feature can effectively pace host write I/Os in the following operational scenarios:

• Slower restore (apply) cycle times on specific R2 devices that are managed by slower-speed physical drives.
• FAST operations that lead to an imbalance in SRDF/A operations between the R1 and R2 sites.
• Sparing operations that lead to R2-side DAs becoming slower in overall restore operations.
• Production I/Os to the R2 side that lead to DAs and/or RAs becoming slower in restore operations.
• Restore delays during the pre-copy phase of TimeFinder/Clone sessions before activation.

The configuration and management of group-level write pacing are unaffected by this enhancement.