Co-authored by Irene Podosenov (firstname.lastname@example.org)
Over the years, retailers have gradually seen a shift in order volumes from physical stores to the online channel. Online sales increased significantly since the pandemic as customers avoided physical stores in these unprecedented times. Some retailers have had their own special events throughout this year to drive sales that have caused an increase in order volumes. Additionally, the holiday season has already begun this year – since October until the end of the year.
As retailers have already ramped up for ongoing holiday season, and are seeing the associated spikes in order volumes, one of the most important question asked is “how prepared are your systems to handle the peak order volumes?”
This blog gives guidance on taking some simple actions to ensure the smooth handling of unanticipated peak orders.
Proactively plan well before the peak
- Start early and involve the most experienced team members in the planning cycle
- Leverage the best practices from the previous planning cycles
Performance test for peak loads
- Outline the anticipated peak volumes and execute end to end testing for all the critical transactions like Create Order, Schedule, Release etc.
- Implement the test recommendations before the peak.
- Ensure housekeeping on system, transaction and databases are complete.
- All patches and critical fixes need to be rolled into production environments
Plan and test failover and disaster recovery
- Test all components of the applications, agents, queues and database components
- Prepare a detailed runbook for execution based on the tests above for controlled stop, switching and start of environment’s
- Be aware of the recovery timelines
Automate monitoring and reporting
- Ensure that statistics collection and reporting on transaction throughput, response times, queue depths, system statistics (CPU/Memory) are readily available.
Plan to support teams during the peak
- Be prepared for bottlenecks
- Include key people from business, infrastructure and application management teams
- Identify critical transactions and workarounds
- Be prepared to disable non-essential processes
Engage with product support team(s)
- Review and share the performance plan and results
- The Sterling product support team is the first point of escalation when there are issues and can support you better when they are proactively kept informed.
During performance tests or in production, when bottlenecks occur while processing high order volumes ensure you monitor resource utilizations (CPU / Memory, JVM Heap Sizes, GC overheads, disk file usage etc.) in all servers that are running the application.
In addition, here are a few pointers for troubleshooting and identifying the root cause in various tiers of the application.
- Review performance reports (Oracle AWR/ADDM report or DB2 Optim Performance manager report)
- Monitor the frequent redo log switches
- Review and ensure statistics are maintained through peak loads
- Use a passive database for all reporting / analytics
- Monitor and resolve all queries that take up long execution times
- Monitor locks and high row lock contentions
Agent/Integration Server Layer
- Monitor Long running transactions
Application Server Layer
- Monitor thread and JDBC connection usage
- Uneven load distribution between application server instances
- Monitor application server logs for exceptions
- Monitor queue depths
- Chose appropriate data store for message persistence (File vs Database)
- Review YFS_Statistics data as part of monitoring
- Application housekeeping (purges, health monitor etc.)
- Ensure all APIs outputs are appropriately pruned with correct templates
I hope you have a better understanding to prepare and execute for Holiday Readiness. While this is not a comprehensive list, you can watch the replay of our webinar IBM Expert Labs Sterling OMS - Best Practices for a smooth Peak Holiday Season for a more detailed overview of the best practices for holiday readiness.
Here are some tips for retailers to successfully navigate the extended holiday shopping season.