Dr Marc Hanheide, a researcher in the Lincoln Centre for Autonomous Systems (L-CAS) research, has now joined the Kumo to help setting up OpenStack on the hardware and also to make Kumo’s service available to the research in L-CAS. L-CAS is one partner in the multi-million Pound project STRANDS, developing robots that learn from long-term experience.

STRANDS_logoLong-term experience in terms of robotics means: Tons of data, and high demands on computing to process all this data. Hence, L-CAS is an ideal early-adaptor  for Kumo’s services. Marc is chipping in his Linux expertise to help setting up OpenStack and the network, to accommodate the requirements for STRANS, but also to help to develop the OpenStack installation further towards the aims outlined in this post.

So, what the status then?

Finally, OpenStack is running on the controller node, based on the devstack installation routines. Devstack has been customised for our local needs and if now successfully deployed and running in a test installation on the hardware. Test instances can be launched (with limited resources at the moment, due to the single node configuration) and the can be assigned floating IPs, allowing them to be integrated into the university’s networking infrastructure. Right now, this is all running in a small network island (subnet), but in the new year, we are looking to

  • integrate the two other computing nodes, to have more computing resources
  • make the floating IPs accessible from within the whole university network
  • integrate the NAS storage into cinder, OpenStacks volume manager, to accommodate large data storage requirements, e.g. for STRANDS
  • get it all up and running also for others to use (enter production stage)

We spent a few days last week trying to install OpenStack on RHEL6, but hit a brick wall in the documentation. We’re now waiting for iLO access to our servers so that we can install Ubuntu on them (the reference platform for installing OpenStack), and we expect to deploy it via Puppet. We’ve been using Puppet for other projects and got further in 30 mins with Puppet installing OpenStack than we did over several days wrestling with RHEL.

In the meantime, here’s a nice video.

Following on from a summer of mind-boggling delays and discussions (for a number of reasons), today our first server was hooked up to iLO in readiness for installation of the OS. As soon ask this is in place we can start rocking out our OpenCloud installation.

Watch this space. It’ll be awesome. We promise.

For those of you interested in this kind of thing, we’re using the following hardware to build our initial Cloud.

  • 30TB QNAP TS-1079 Pro NAS Storage
  • 2 x HP DL380P servers with 146GB 15K SAS drives on board. These are used for OpenStack compute nodes. These have 12 CPU cores and 64 GB RAM each.
  • 1 x HP DL 360P server with a 146GB 15K SAS drive, 4 cores and 12GB RAM. This is used for the OpenStack Controller.

It may be of interest to readers of this blog that the budget we had for this hardware was around £17,000.

At the point of writing, the following diagram shows how we think everything will click together. This will almost certainly change, but we’ll keep you updated when it does.