High Availability Pack - The Art of Optimizing Bandwidth and Storage

REPLICATION SERVER. The most importantas this is ready, probably within an hour, the primary
characteristic of HAP is support for replication of theserver is ready for backups and restores. While the
primary backup appliance. This lets the user copy allsecondary partition is being checked, the incoming
incoming blocks and meta-information to theblocks for backups are stored on the first partition
replication server. Of course this happens withoutuntil it is 95% full. This functionality is available on
interrupting the backup process. The replicationsingle and replicated setups.
server is a Backup Appliance. Just like with everyIf the primary partition is available, but the secondary
other our Backup Appliance, all HAP settings areis not available yet, restores can be performed by
placed on the primary server. The number ofapproaching the binary blocks on the replica server.
replicated appliances is unlimited It is possible toThis is done automatically when the storage on the
connect three, four or even more to each other as aprimary appliance is not available.
backup for the backups of the backups. EachADDING STORAGE. HAP also lets the user add
incoming block is sent to the replica server as soonstorage to a backup appliance. This can be direct
as it is available. If the replica server is not availableattached storage, UNC storage. Fiber attached SANs
temporarily, the primary server keeps track of whichor NASs and even USB disks (not recommended).
blocks have not been replicated yet. The metaAdding storage demands special expertise and is
information is still replicated by the database enginetherefore implemented from your sales contact
that is in use and can also pick up everything afterperson.
temporary non-availability of the replica server. TheNO AUTOMATIC FALLBACK. The HAP is not an
status of the replication can be seen in the console.automatic fallback system. If the primary appliance is
POWER LOSS. One of the most critical events for ano longer available for reasons such as theft or fire,
backup server is power loss. 95% of all appliancesthe replica appliance must be upgraded to a primary
are in a data centre. Although it should not happen,appliance. This can be done by the technicians by
reality has taught: us that power loss in a data centrechanging the BSP code. Although this is only a small
does actually occur. The problem here is that the OSchore, it has major consequences. It is not possible
for the appliance will check all storage before it isto reverse this process without copying the
available for the new backup. This can last 12 hoursmetadatabase. This not only takes a lot of time, but
on a 10TB system with a lot of information. Tomoreover both servers are unavailable during this
prevent this the 7D appliance and 14D appliance havecopying process. Depending on the available binary
two partitions This first partition contains 5% of theblocks on the new replica server, it can also take up
available storage and the second partition 95% ofa lot of time to replicate the binary blocks again. This
the storage. In the case of a power loss or otherdepends on the volume, the delays in the network
cause that forces the OS to check all of the availableand the available bandwidth.
storage, the first partition is checked first. As soon