HP XP P9500 Storage - Manual (Page 23)

1. Using your performance monitoring software, collect the following:
•
Disk-write bytes-per-second (MB/s) for every physical volume that will be replicated.
•
Data should be collected over a 3 or 4-week period to cover a normal, full business cycle.
•
Data should be collected at 5 minute intervals. If you use averages, shorter intervals
provide more accuracy.
2. At the end of the collection period, convert the data to MB/second, if needed, and import
into a spreadsheet tool.
Sizing journal volumes
You calculate the size of your journal volumes using write-workload and RPO.
To calculate journal size
1.
Follow the instructions for “Measuring write-workload” (page 22) .
2. Use your system’s peak write-workload and your organization’s RPO to calculate journal size.
For example:
RPO = 2 hours
Write-workload = 30 MB/sec
Calculate write-workload for the RPO. In the example, write-workload over a two-hour period
is calculated as follows:
30 MB/second x 60 seconds = 1800 MB/minute
1800 MB/minute x 60 minutes = 108,000 MB/hour
108000 MB/hour x 2 = 416,000 MB/2 hours
Basic journal volume size = 416,000 MB (416 GB)
Journal volume capacity and bandwidth size work together. Also, your strategy for protecting your
data may allow you to adjust bandwidth or the size of your journal volumes. For a discussion on
sizing strategies, see
(page 41) .
NOTE: If you are planning for disaster recovery, the remote system must be large enough to
handle the production workload, and therefore, must be the same size as master journals. If not
planning a disaster recovery solution, remote journal volumes may be smaller than master journal
volumes.
Planning journals
Cnt Ac-J Z manages pair operations for data consistency through the use of journals. Cnt Ac-J Z
journals enable update sequence consistency to be maintained across a group of volumes.
Understanding the consistency requirements for an application (or group of applications) and their
volumes will indicate how to structure journals.
For example, databases are typically implemented in two sections. The bulk of the data is resident
in a central data store, while incoming transactions are written to logs that are subsequently applied
to the data store.
If the log volume “gets ahead” of the data store, it is possible that transactions could be lost at
recovery time. Therefore, to ensure a valid recovery image on a replication volume, it is important
that both the data store and logs are I/O consistent by placing them in the same journal.
The following information about journal volumes and journals will help you plan your journals.
•
A journal consists of one or more journal volumes and associated data volumes.
•
A journal can have only P-VOLs/master journals, or S-VOLs/restore journals.
•
A journal cannot belong to more than one storage system (local or remote).
Sizing journal volumes 23