Doubletake has two versions, one
that works at the guest operating system level and one that works at the VM
host level. The VM host level version is a new product and something that Doubletake have recently ventured
into, the operating system level version of Doubletake is a mature product
and sits within the host guest operating system and copies every single
file that is written to disk, this is placed into a queue that is then
replicated to a different machine.
Traditionally this product was
used on physical servers where there was a one-to-one ratio of the production
server to DR server that would be put in the data centre. In recent times they have adapted this
product to allow for either a physical server or virtual server
(it makes no difference as this is an OS aware back-up strategy) to then
replicate to a single guest called a virtual recovery assistant that is located
on a different host VM in a data centre.
This has a significant advantage as it requires less licensing
costs.
Traditionally, Doubletake would
require a licence for both the production server and
the DR server. With VRA a licence is
required for each source server but only one licences is required for the VRA and if your source servers are virtual then you can buy packs of 5 license.
The virtual recovery assistant works
by attaching multiple virtual hard discs for each source server that is
being backed up. if we have
two source servers, either physical or virtual, and each has two hard discs,
the virtual recovery assistant will have one operating system, but will have all four hard disks attached to it. As files are changed on the source server they are placed into a
queue on the source server, that then get replicated to the virtual recovery
assistant which knows which source server they came from and applies them in
the correct order to the associated hard discs.
In the event that we require bringing up a failed server, the
virtual recovery assistant will detach the hard discs from the VRA and attach
them to the newly created virtual server.
One of the great advantages about using this technology, is that we are
able to actually see the physical files that have been replicated on the VRA as
these are mount points to yhr VHD, the disadvantage with this is that in a disaster, we need to have
the virtual recovery assistant fully functional as we need to run
the proper Doubletake scripts to detach the mount points from the VRA, build
the appropriate server and attach the hard discs. This should not be done as a manual task as
this will generally cause the virtual machine to fail.
One of the great advantages about
this product is that it does not require the host (source) to require
snapshots. As the agent works within the
guest operating system, it takes advantage of the natural fault tolerance that
is designed within many applications.
Furthermore, if we had decided to use snapshotting technologies, we
would not have been able to use these during the day as it is not advisable to
snapshot on high I/O loads and we would not be utilising the bandwidth we have
available during the day and placing all the load on that bandwidth at
night. By being able to replicate to a
DR site during the day, this spreads the load across our line over a
twenty-four hours period and makes best usage of the available bandwidth
without us having to increase it for night usage. One of the disadvantages of this technology
is it is not designed as a back-up technology, purely as a replication. In
newer releases, it is expected that there will be a snapshotting technology
that can be used on the target server and we will be able to take a minimal
number of snapshot. The disadvantage of this
technology is it is expensive compared to Veeam but it may not be necessary to use it on all
production servers. On servers that have
very little change, we would be able to use a different technology to replicate these in the evening.
The biggest disadvantage with
Doubletake, as it is working at the guest level, in the event that a
guest server is rebooted, the Doubletake agent is stopped and at this point
loses track of any changes that are being made to the server during the
reboot. Because of this, when the server
is brought back up, a re-mirror is required.
A re-mirror does not necessarily mean the whole data has to be
retransmitted to the replication site as a check sum is taken against each
individual file, but if a servers that have a significant amount of data, this
re-mirror can take a significant amount of time, sometimes up to a week if you
have a lot of small files. This is not
necessarily a problem because while the remirror is taking place, the data is
still being replicated.
http://www.visionsolutions.com/products/vision-products-overview.aspx
http://www.visionsolutions.com/products/vision-products-overview.aspx
No comments:
Post a Comment