Primary Storage

Stackers Summit in Austin 2016: “You did not tell me you attend a concert” / Getting unified

By Archive User posted Sun May 15, 2016 08:45 AM


Originally posted by: slorenz





Quotation is my wife’s reaction on a video I sent her from the summit, more further down.

The first Openstack Summit I attended, was the one in Vancouver Spring 2015, hold in the beautiful Convention Center at the harbor.
I was very impressed about how community work can create great things. Very helpful talks about how to use Openstack components.

As of that I was looking forward to the Spring 2016 Openstack Summit in Austin. And I was not disappointed.
Again a great location, very good hosting in the Austin Convention Center and even more helpful talks.

Compared to the Vancouver Summit, I found the Vancouver one to be slightly more a Techy convention, meaning in the sense, that it was mainly tech talks (how does Openstack help me for my use case), easy formatted slides, ...(also optical, just very view suits ;-) ).
My impression is, that this was a bit different for Austin. Slides for Talks where more professional like (some seemed marketing helped) and overall it seemed more business driven (showing telling we/our product does it this and this way). That’s perfectly fine, as while the core is Open Source, there is a lot Business around it, which keeps it alive.

Monday and Tuesday started with the Keynotes that gave great details how Openstack is used in different use cases.
Well, actually the days did not start with the Keynotes, but as of being in “The Live Music Capital of the World” they decided to have live music and the band Soul Track Mind did a great job at 08:30 am in the morning.

That’s what the video I mentioned at the beginning was about.

I’m particularly interested in Swift, Cinder and Manila Openstack Components, as I’m working in the IBM Spectrum Scale Team.
We are working on the highly integration of Openstack components with the IBM Spectrum Scale product.

To also help users to combine the great features of a fast, highly available and scalable shared file system with the Openstack Swift (Object), Cinder (Block) and Manila (File) features, we released a Redpaper "IBM Spectrum Scale in an OpenStack Environment" on Day 1 of the Openstack Summit.

It is great to see how we can combine in example Swift (Object) features with features of a scale and performance proven shared file system to provide solutions for a lot more use cases.
With the help of SwiftonFile, we can access the same data from either object or file side, without moving data or calculations on where to find it.
Means you are able to upload data via Swift and you can directly access it via file protocols. Same works vice versa, store data as file and get it via Swift.
This enables in example analytic workloads directly on object data. It also helps a lot in digital media workloads.

We were able to give 2 talks on the Unified topic:

The 1st “Amalgamating Manilla And Swift for Unified Data Sharing Across Instances” explained, how also Manila can be combined with Swift, SwiftonFile and Spectrum Scale.
We gave a demonstration of an example use case for Live Video Production Systems or Video Surveillance setups.
A sample Video (my Catamaran Trip to the Summit ;-) ) was upload as an object. It then was edited with a file based video editing software. The video was loaded from a NFS share that was created on the Object Store path, means right from the spot where the object store saved it. We rendered the edit and saved it back into the NFS share and were able to show the result (a nearly flipping Catamaran) directly after downloading via Swift.

With the 2nd talk “Write a File, Read as an Object” we went deeper into the design and how SwiftonFile is used. As Object and File have their strength in different areas, those strengths are in some cases not easy to combine. During the talk we also mentioned some of the challenges such as accessing object data directly using file interfaces (SMB/NFS/POSIX) and common user identity across file and object. We also described how those challenges can be addressed.

During the summit I learned, that Unified (File and Object access to the same data) becomes more and more important, as there are lots of use cathe ses. As another example, that was also stated at a talk: In the past, Applications were file based and many won’t be changed to use the object interface.
Being able to use Unified access will provide the solution to this.
Additionally the summit showed me, that IBM Spectrum Scale with its “Unified file and object” feature fits in here very well. Object is known to scale very large and usually filesystems don’t. That’s different with Spectrum Scale (built upon IBM GPFS). It is able to scale very large and as such works great in combination with Object. We are able to combine the best of both worlds and can provide solid and performing unified solutions for a lot of use cases.

It was a great opportunity to attend the summit, being able to present and working with so many enthusiastic stackers, colleagues and customers.

I’m looking forward to the next Summit which will be in Barcelona and I hope I will be able to attend.

(These are my personal views and not of my employer’s)