Big Open Source Week for EMC - Part 1: OpenStack

This week is the OpenStack summit in Tokyo - EMC is there with bells on :-)

I’m not there, but there are many people on the team there… It was interesting to watch the keynote and reflect on the nature of an Open Source ecosystem at work...

In particular - the rapid movement of projects from emergent into deployment, and how the community iterates around them.

The Project Navigator (screenshot below, link here) is interesting in how it shows maturity and adoption of various parts of the project. Something useful for the rest of the market (us included) to emulate, IMO.

NewImage

Now, if you scroll down further - you can see where more work is needed (Ironic, Sahara, Trove, Ceilometer) - but again, like it or not, the community will keep cranking on it.

I’ve always used StackAlytics to look at how people (including us are contributing) - which is great, but it’s great to get a summary view across the entire community how things are being used and adopted.

In my experience - I still see customers struggling with OpenStack more than I see customer flying - but the ratios are turning as the ecosystem keeps improving the stack, tightly packaging (interesting to see what @blueboxjesse and the IBM crew are doing). I think we can keep doing more (within EMC, and within VMware) to keep making getting down to business faster and easier.

So - what is EMC doing at Tokyo (distinct from the VMware contributions which are extensive, including but not limited to VMware Integrated Openstack aka VIO)?

A lot:

Core contributions. You can see what we’ve contributed to Cinder and more in Liberty - there’s a lot!
We’ve always done OK with platform drivers for Cinder for things like VMAX and VNX, but a big step forward is that that ScaleIO drivers are embedded in the Liberty release. ScaleIO has been supported, but these things don’t fly until the drivers are directly included. In ScaleIO’s scale, this means that important part of ScaleIO are simply… included versus needing to get them from us.
Based on customer requests - the always awesome EMC{code} team we’ve created two Mirantis Fuel Plugins for ScaleIO - one for pure installs, and one for attaching to pre-existing ScaleIO SDS clusters. Yes, for those of you using Canonical, Juju charms are coming...

To highlight how strong the ScaleIO SDS transactional block stack is, we did a little stunt (bending/breaking my “never go negative” stance UPDATED: watched the video - the guys did a balanced job, they didn’t go negative, was very balanced, goal was the delta between layering transactional on object, vs. a purpose built stack) with a live comparison of ScaleIO and CEPH’s block implementation on top of RADOS. In this case, we are the underdog (the “David” to the “Goliath” that is CEPH). Most people in the community are well aware of CEPH as an object, block and NAS implementation - but it’s early days of people knowing about SDS alternatives like ScaleIO. We used a live scale-up test of Ceph and ScaleIO in AWS to illustrate how simple, and how efficient ScaleIO is for any transactional use case - you can see the head-to-head here:

Look - everyone likes a dramatic head-to-head, but the beautiful thing here is you can try for yourself. Don’t listen to anyone (me included). You can download ScaleIO here - and it has NO time-bomb, NO feature limit, NO capacity limit. Go. To. Town.

Two over-arching thoughts:

Openstack overall is suffering from the “buzzword bingo” factor that vSphere does re: “Containers!” - where everyone want to paint things in opposition. Yes, like vSphere, Openstack has historically had an kernel-mode virtualization IaaS focus. Projects like Magnum embrace container management, and are roughly analogous to vSphere integrated containers. It will be interesting to see (over time) how many customers decide to layer containers ON TOP of platforms that are mostly kernel-mode/physical IaaS centric (containers on OpenStack/vSphere, vSphere integrated containers, Mesos on top of OpenStack/vSphere, Cloud Foundry on top of OpenStack/vSphere/AWS)… and how many customers will go in different directions (things like the Photon Platform, or Mesos on bare metal)
EMC/VMware - we are viewed with some scepticism by the communities - and I get that. People need to deeply understand (and I suspect this won’t change should the Dell/EMC deal conclude) - we recognize that we operate in an open ecosystem. Sometimes We must embrace OpenStack (and clearly are), we must embrace Mesos (and we clearly are - see all the Mesos contributions).

What do you think? What more should we/could we do? What have been your experiences with OpenStack?