I’m seeing more and more Openstack in my interactions with customers. Personally, it increasingly “feels” like a two horse race at most of the customers where they are looking at/evaluating/using the vCloud Suite and/or Openstack (along with Puppet/Chef) as the management, orchestration, automation and self-servie portal. IMO, the vCloud Suite is more of a “enterprise product” (with the good and the bad that implies).
BTW – If you’re interested in Openstack and what EMC is doing, I’d follow Nikhil Sharma’s blog here.
Ok, when it comes to Openstack, we’re committed to making all our drivers open and available, and you can get the Havana drivers at Github here:
- FC SMI-S driver for VMAX and VNX: https://github.com/emc-openstack/smis-fc-cinder-driver
- iSCSI SMI-S for VMAX and VNX: https://github.com/emc-openstack/smis-iscsi-cinder-driver
- iSCSI “Direct” driver for VNX: https://github.com/emc-openstack/vnx-direct-driver
- ViPR driver: https://github.com/emcvipr/controller-openstack-cinder
They come in a three flavors:
- Platform-specific SMI-S drivers. Best choice for quick tactical deployments for customers with VNX and VMAX platforms.
- iSCSI “direct” driver. Interacts directly with the array using the array APIs. Best choice for customers with just VNX.
- ViPR driver. If it’s me, I would go this route. It’s missing a little bit, and you need to import the ViPR virtual appliance to get started (which today is vSphere only – sometimes an issue in Openstack environments), but it is the strategic direction of EMC for this sort of purpose, and will get a lot of dev and iteration. ViPR becomes a universal “abstract, pool, automate” layer of EMC and non-EMC storage.
NOTE: working to make the ViPR Controller and Data Services easily accessible to all – please stay tuned…
It’s notable that these are the “Cinder” (oriented for block volume) drivers. We are also jumping in working with other industry leaders like NetApp and Mirantis (in partnership!) to improve the NAS support in the Manila project (see here). As of Grizzly, you can use NFS but it’s really missing a lot of things to make it all it can be (check it how to use NFS now here).
Beyond the volumes that are used by Nova (compute) instances, object storage models are the thing that increasingly the compute nodes are accessing, and in Openstack that’s Swift. ViPR is a SWIFT compliant software-only object store (and also supports S3, the ATMOS APIs). Also commonly accessed from compute instances are HDFS storage models. Hey – the ViPR object stack ALSO supports HDFS, and something really cool and unique – which is simultaneously accessing data via object and HDFS protocols!
We want some of your feedback (EMCer’s, EMC Partners, Cutomers) on what you’re doing (or thinking of doing) when it comes to Openstack… PLEASE LET ME KNOW – THIS WILL STEER THE SHIP! Openstack Survey