IBM DS6000 Computer Drive User Manual


 
182 IBM System Storage DS6000 Series: Copy Services with IBM System z
15.1 Performance
Because Metro Mirror is a synchronous mirroring technology, it introduces a performance
overhead (for write operations) as compared to a similar environment with no remote
mirroring. As part of the planning process for Metro Mirror, you should understand this.
Bandwidth analysis and capacity planning for your Metro Mirror links can help define how
many links you need, and when you need to add more, to ensure best possible performance.
As part of your implementation project you might be able to identify and then distribute hot
spots across your configuration or take other actions to manage performance.
Note that no processor resources (CPU and memory) are consumed by your Metro Mirror
volume pairs (your management solution excluded) because these resources are managed
by your DS6000 subsystem.
15.1.1 Peak bandwidth requirements
When you are researching your Metro Mirror link bandwidth requirements, you must find the
peak sustained write rate. Only writes are mirrored to the secondary volumes. For a typical
z/OS system, this peak is usually during batch processing, when many updates are occurring.
It could be higher still during month-end or year-end processing.
You must make allowances for this peak period when you are calculating bandwidth.
15.1.2 RMF
Resource Management Facility (RMF) can collect and report Metro Mirror link performance
statistics on DS6000 subsystems, and it is used for capacity planning and trending reports.
You can track the utilization of the Metro Mirror links over time and introduce new links in a
planned manner.
You can also produce reports about Extent Pool and Rank performance statistics using the
RMF Post Processor, if you have enabled the collection of this data through the RMF Parmlib
member.
15.1.3 Initial synchronization
When doing the initial synchronization of your Metro Mirror pairs, the DS6000 uses a
throttling algorithm to prevent the Metro Mirror process from using too many primary site
resources. This might prolong the synchronization process if your DS6000 is busy at the time.
You can choose to stagger your synchronization tasks or to run them at a time of low
utilization to make this process more efficient.
As an alternative, you may also choose to do the initial synchronization in the Global Copy
mode, and then switch to the Metro Mirror mode. This will allow you to bring all volumes to
duplex-pending XD state, which has little or no application impact, and then switch to full
duplex state. You can do so through TSO commands, the DS CLI, or one of the other
interfaces.