Skytap

AIX Best Practices for Skytap Environments

This guide covers recommended settings, diagnostics, and best practices for AIX LPARs running in Skytap, especially for environments using Oracle ASM and high I/O workloads.

1. Network Configuration and Tuning

Gather current network settings:

no -a | grep rex
no -a | grep sack
no -a | grep sb
no -a | grep pmtu_di
no -a | grep pmtu_ex
no -a | grep node
ifconfig en0
lsattr -E -l en0

Recommended settings for optimal network performance:

# no -p -o tcprexmtthresh=100
# no -p -o sack=1
# no -p -o sb_max=1048576
# no -p -o tcp_pmtu_discover=1
# no -p -o pmtu_expire=10
# no -p -o tcp_nodelayack=1
# chdev -P -l en0 -a rfc1323=1
# chdev -P -l en0 -a tcp_sendspace=1048576
# chdev -P -l en0 -a tcp_recvspace=1048576
# chdev -P -l en0 -a mtu_bypass=off
# chdev -P -l en0 -a mtu=1240
# chdev -P -l en0 -a remmtu=576

Note: After making these changes, validate performance in the guest OS.

Skytap VPN MSS Recommendation:
Set the MSS in Skytap VPNs to 1200 for improved performance. The current N/A value should not cause issues, but setting it to 1200 can enhance performance. Changing this requires disabling the WAN to update.

2. Disk Tuning: `rw_timeout` and Oracle ASM

Check current disk devices:

lsdev -Cc disk

Check current **``** value for each disk:

lsattr -El hdiskX -a rw_timeout
# Replace hdiskX with your disk name

Best Practice:
Set rw_timeout = 120 and queue_depth = 10 for each hdisk.

3. Oracle ASM and Disk Layout

Example of a Workload profile:
- 95% reads with large block size.
- Majority of workload goes through 5–6 ASM raw disks.
- 70% of workload passes through a single vscsi adapter (e.g., vscsi1).
Best Practice:
Distribute ASM disks more evenly across controllers to increase resilience and reduce bottlenecks. This change requires a maintenance window (LPAR must be powered off). Skytap Support will need to know which hdisks are used by Oracle ASM for each LPAR involved.

4. Troubleshooting & Follow-up

Apply network and disk tuning as described.
Test results in the guest OS after making these changes.
Document which LPARs and hdisks are affected, and coordinate with Skytap Support for infrastructure-level changes if needed.
If issues are found on multiple LPARs with different configurations, document each occurrence for analysis.

5. Additional Recommendations

Do not use manual PVID assignment for Oracle ASM disks—let the system manage disk identifiers.
Increasing rw_timeout and queue_depth can decrease the risk of disk disconnections.
Record all current settings before making any changes.
Use the tuning values provided as the default unless your workload analysis suggests otherwise.

References

For further support or architecture guidance, contact Skytap Support.

This site is open source. Improve this page.