Open Stack

Hi everyone,

I'm a system administrator at my University's research lab, IIIT-H.
We've been using OpenStack for over 6 months now. The installation is
good old multi-node with one controller(network, horizon), 7 compute
and one storage(volume) node. With ever increasing storage
requirements, we are planning to move to distributed storage.

Openstack installation is currently concentrated on a single rack and
I want to experiment with the possibility of having storage and
compute nodes on the same hosts. The arrangement will look similar to
the diagram below. Here is the link to the image for better
visualization

http://www.asciiflow.com/#3254548095955234713

                                       Distributed
                                        Storage
                                        +-------+
                        +---------------|+----+ |-+
                 Node1  |+-compute-+    ||    | | |
                        +---------------||disk| |-+
                 Node2  |+-compute-+    ||    | | |
                        +---------------|+disk+ |-+
                 Node3  |+-compute-+    ||    | | |
                        +---------------||disk| |-+
                 Node4  |+-compute-+    ||    | | |
                        +---------------|+disk+ |-+
                 Node5  |+-compute-+    ||    | | |
                        +---------------||disk| |-+
                 Node6  |+-compute-+    ||    | | |
                        +---------------|+----+ |-+
                                        +-------+
This distributed storage will hold volume storage, swift, glance image
store and VM disk storage. I would like you to share your thoughts
regarding this layout being suitable for a production level
environment.

Note that, our computations primarily include huge data processing
through hadoop. As a result, data on a single server was turning out
to be a bottleneck.

Thank you.

--
Shashank Sahni