Lessons Learned from Openstack SV Summit: Play nice with Containers (Or Else…!)

Posted by Pascal Joly August 25, 2016
Lessons Learned from Openstack SV Summit: Play nice with Containers (Or Else…!)

I attended the Openstack SV summit a couple of weeks ago. While a smaller venue than the bi-annual summit it certainly is the biggest event outside of these and it has the added merit to be local to the San Francisco Bay Area. Once you get pass the massive bottleneck on Shoreline and 101 during commute times, we have a winning combination drawing the brightest minds behind Openstack and great food truck selection at lunch time.

There has been much debate in the last couple of years, since the rise of Docker, about the positioning, or even relevance of Openstack in the world of containers. While not directly in competition, there is certainly not a direct dependency for containers to use Openstack, nor the other way around.

This time around the main topic was very much around the winning combo of Openstack and Google Kubernetes. The whole idea is to use the power of Kubernetes to deploy complex containerized applications to address the challenge of Openstack installation, scaling and upgrade. Since Google open sourced Kubernetes about a year ago, they have cozied up with the Openstack approach, not something we had seen in the past.

Image result

Mirantis has taken the initiative to bring their Fuel platform to the next level and use Kubernetes as their primary underlying technology. Indeed, as a side effect, Google was promoting their solution as the large scale container orchestrator among other well known contenders such as the likes of Mesos/DCOS and Docker. Interestingly enough, the container world was this time around represented by Docker's alternative: CoreOS. While it gives some hopes to solve Openstack still daunting challenges for future green field adopters, it remains to be seen if the existing implementations (such as AT&T) will eventually migrate to it.

Another important topic of discussion was how to address the current short comings of Openstack networking (OVS as in "Open Virtual Switch") with respect to scale and performance. Some companies have been working on network offload solutions both hardware (Netronome) and software (CPlane networks). VMware is betting on a new project (OVN as in "Open Virtual Network" -- let's not be too creative with the acronyms) that will essentially provide the same capabilities as NSX in an open source framework.

Lastly, white boxes vendors such as Quanta, shared their intent to provide more integrated certification with various rack configuration of Openstack (and Kubernetes if needed). This is in response to some feedback from their customers to get higher level of confidence that the generic hardware they buy will indeed perform at scale with specific workload. This is where a Cloud Sandboxes such as Quali's CloudShell will come handy for both the vendor and their customer s, since there are many possible hardware/software/versions combinations that should be certified and validated in a dynamic and automated way.

I’m now looking forward to the November Openstack summit in Barcelona, with this time some tapas on the menu, and certainly some great seafood.

On a related note, expect an upcoming Openstack App deployment option for CloudShell in the coming months that will enable developers to deploy these workloads and automate their CI lifecycle.