Successful System Recovery using Ignite-UX

Processing speed
You should be aware that when creating a recovery archive on lower-end and uniprocessor
systems (for example: rx1600, rx2600, rp24xx, and rp3410 systems), a considerable amount of
CPU time is required to perform Ignite-UX tasks.
When writing a network recovery archive, the system compresses the tar or cpio archive being
written to the NFS file system using gzip. There is a considerable amount of CPU time spent
compressing this data. The CPU overhead and the impact of NFS traffic on a system can slow
other applications down considerably. On low-end and uniprocessor systems, it is not
recommended that you create recovery archives while running production applications.
Higher-end systems are less adversely affected when creating recovery archives; however, systems
at the edge of performance problems may experience performance problems.
Note:
You should take extreme care when running applications or other
software on a system that requires a significant amount of resources
while attempting to create a recovery archive at the same time.
Input/output (I/O) usage
When running make_tape_recovery or make_net_recovery, either command can create a
significant amount of disk I/O. This can impact an application as the I/O is executed through the
system’s buffer cache. The I/O performed by Ignite-UX competes with other applications for the
buffer cache, thus reducing the application’s performance.
It is possible that in a situation of high I/O usage by the application, contention for physical I/O
can be created between it and Ignite-UX. This is not a problem when there is no application data
or programs in the root volume or disk group, unless the system is also paging to swap spaces in
the root volume or disk group.
Creation time
If there is a fixed time to create a recovery archive, you should review the other topics discussed in
this white paper that can affect the amount of time it takes to create a recovery archive.
Using past experience, you should decide what amount of time is reasonable to complete the
creation of the recovery archive. If the actual creation time extends past this point, you should
evaluate the other topics discussed in this white paper to determine which may be affecting your
recovery archive creation. In addition, if the amount of data included in the recovery archive has
greatly increased, the expected completion time should be adjusted accordingly.
Service level agreements
The most overriding requirement that you have is to adhere to the Service Level Agreement (SLA).
Since part of a normal SLA is restoration time in the event of an outage, you can use this restoration
time as the basis of your recovery strategy.
For example, if your SLA requires that the system be recovered within 12 hours, you must develop
a recovery strategy that can be implemented in 12 hours or less.
9