• Data is transferred
host and target—
• May not support all
file system types
(can also be sent
and OS platforms
Virtual Storage Pool • Adds a potential
point of failure
With the out of band, asymmetric solution, whether it is Ethernet or Fibre
Channel attached (it doesn’t matter) we have to install an agent on each host
for our block I/O request. Because we have to put an agent on each host it
may not support all Operating Systems.
When our block I/O request comes in the agent intercepts it. It redirects it to
our metadata engine (virtualization engine, abstraction layer). It is a single
point of failure so now we are running it as a clustered pair. These can either
be software or hardware based solutions.
Our metadata request comes in to the virtualization layer, it goes back with the
real metadata location, we then do data I/O direct. By putting it into a
clustered pair it mitigates the risk of potential failure. The data I/O is not going
through our virtualization engine so it also alleviates the potential bottleneck.