HP StorageWorks XP24000 Continuous Access Software User and Reference Guide, v01 (T5278-96002, June 2007)

Table21OptimizingContinuousAccessOperationsandStorageSystemPerformance
Condition
Description
Recommendation(s)
Write-intensive
workloads
Write-intensive workloads, such as
database logging volumes, ca n have a
significant impact on storage system I/O
response times.
Spread write-intensive data a cross several
volumes to minimize queuing.
Large block size
Workloads with large write block sizes,
such as DB deferred writes, can impact
performance.
Spread workloads with large write block
sizes across several volumes.
High host
channel
demand
The demand on the MCU’s host channels
can affect performance.
Spread the workload across several
storage systems to utilize additional
channels.
Sequential write
operations
Continuous Access operations can have a
negative impact on workloads with a high
percentage of sequential write operations,
such as batch processing operations (for
example, dump/restore, sor t operations).
Avoid performing restore operations to
volumes which belong to Continuous
Access pairs. Instead, restore data to
a
scratch volume, and then create the
Continuous Acc ess pair.
Cache si ze
A large cachesizeimprovesreadhit
performance, which allows more storage
system resources to be devoted to write
operations. The resulting performance
improvement can offset some or all of the
performance loss due to the Continuous
Access
remote c o py operations.
Continuous Access Asynchronous (and
Concurrent Copy) requires additional
cache for sidefile data. Insufficient cache
resources can result in command retries,
state-change-pending (SCP) notifications,
and
puncture conditions.
Consider increasing the cache size of
the Continuous Access storage systems to
handle sidefile operations and to improve
overall storage system performance. For
best results, the cache and NVS capacity
of the
primary and secondary storage
systems should be the same (the RCU
sidefile requirements for Continuous
Access Asynchronous are twice that of the
MCU) to enable theremotesitetofunction
adequately during disaster recovery.
RCU capacity
Theperformance of theRCUs directly
affects the performance of the MCUs.
If an RCU becomes overloaded with
heavy update activity, MCU and system
performance can also be degraded.
Distribute Continuous Access remote copy
operations among several remote storage
systemstoavoid overloadingany one
RCU.
Paths
An inadequate number of paths may
decrease storage system performance.
Performing Continuous Access
Synchronous operations over long
distances can also degrade performance.
Continuous Access Asynchronous is
recommended for long distances.
Make sure to install an adequate number
of
paths between the primary and
secondar y storage systems. This is
especially important for storage systems
which contain both P-VOLs and S-VOLs.
Discontinuing Continuous Access Operations
If you plan to use Continuous Access for nondisruptive data m igration or duplication (see
“Using Continuous Access for Data Migration a nd Duplication”onpage123), you willconfigure and
establish Continuous Access operations, allow Continuous Access to synchronize the volumes, redire ct
application I/Os (if migrating), and then discontinue Continuous Access operations. When you are ready
to discontinue Continuous Access o p erations, you will need to perform Continuous Access operations in
the
correct order to avoid g enerating error messages. For example, Continuous Access will not allow you
to delete an RCU path until you have released all Continuous Access pairs still using that path, an d you
cannot
delete a group until you have released all Continuous Access Asynchronous pa irs in that gro up.
To discontinue all Continuous Access operations, perform the following actions in the specified order:
1. Release all Continuous Access pairs from the MCU(s) (pairsplit-S, see
“Releasing Continuous Access Pairs ” on page 120) . For Continuous Access
XP24000ContinuousAccessSoftwareUserand ReferenceGuide
97