OpenStack Interop Challenge

At the OpenStack Summit in Barcelona, 16 vendors stood on stage and demonstrated interoperability. This was a major breakthrough for OpenStack. It marked a significant departure from just 18 months earlier when the OpenStack Foundation had chided vendors for creating lots of proprietary solutions. Enterprise Times sat down with Angel Diaz, IBM Vice President, Cloud Architecture and Technology to talk about this achievement.

ET:         I was impressed with the OpenStack interoperability announcement. It’s 18 months after the OpenStack Foundation called distributors to account in Vancouver. Everybody had taken their own tools and stuck them into their OpenStack distribution to make it different. The result was that portability had begun to go away. The interoperability test shows that we have resolved a lot of that. How hard has it been, first of all, to clean up all the code to get to the point of having easy interoperability?

Angel Diaz, Vice President, Cloud Architecture and Technology, IBM
Angel Diaz, Vice President, Cloud Architecture and Technology, IBM

AD:         Before I deal with that, let me just give you how we got here. When we launched OpenStack in 2012, I walked into Rackspace and Jonathan [Bryce] was still working on it. I said “Hey maybe we can do something special – how about we actually do Open Source where it’s an open governance, a community not a meritocracy, and not a single man in control.” It ended up being 6 of us that were crazy enough to sign the £500,000 a year cheque to stand up the Foundation.

75 people showed up to the first conference, in Spain this year it was close to 6000 and last year in Austin we had almost 9,000.  It was a lot easier for West coast people to get there back then. I remember saying to the Rackspace executive, and my own executives at IBM, our objective was to create a ubiquitous open source infrastructure service platform. It’s a humble objective. It didn’t seem so humble at the time.

People think Cloud has been around for ever, and it hasn’t. The concept has but the widespread use of it hasn’t. It’s still a growing market. Interoperability was something that we kind of just took for granted. It was about creating ubiquity. Then everyone joined this effort and said we wanted to make it as good as we can. IBM’s got over 400 computer engineers, about 180 dedicated, full-time, focussed on making OpenStack work. It doesn’t matter to the clients if it is our OpenStack or another OpenStack, they want a working code base.

Then we realised that “oh man”, but at the same time it is great to have ubiquity, it’s great for the software to work, but there isn’t one Cloud, not just an IBM cloud or a “this” cloud.  There is not one web, there are multiple webs. So we really need interoperability. Why? Because our client’s business processes aren’t made of cement. They need to change and move, private Cloud, public Cloud.

ET:         But they also change vendor

AD:         They change vendors, right. Vendor lock-in was a concern which we had to focus on. At IBM, our involvement in these communities is very much driven by our clients’ and our clients’ use of the system. We used the code from the community and we contribute code to the community. We started working and focussed on RefStack. We have DefCore and that starts to make things a little bit easier.

As you point out it is still not enough. When you look at interoperability you have semantics of the API definition. You get configuration that needs to happen and all of these things weren’t as well-specified as they should be. They made interoperability harder. You could interoperate, anything could be done, it’s computer science! Trust me, 6 months ago I could make multiple vendors’ regions work together. It would just be a longer process and a more difficult process and not an obvious process.

ET:         It is interesting to see how that has changed.  A couple of months ago I was talking to Rackspace about this. Some of their internal guys were saying “well if somebody wants to go from OpenStack A to OpenStack B, that’s a week of effort.” The interop should reduce that down to just a few hours right?

AD:         That’s right. That week, more or less, will depend on how good the person is. It will also depend on how the API was used, how the configurations or scripts that drive the deployment were written. Is this enough to an enterprise to think maybe it isn’t interoperable? Do I still have to pay for a migration? It becomes a barrier. At the OpenStack Summit in Austin we stood on stage and issued a challenge to the community. What IBM did, on behalf of our clients and on behalf of ourselves frankly, was to say: “hey maybe we should actually demonstrate interoperability.” In the process of standing up we set a target. “Let’s stand on stage on this date.” Honestly we thought we’d just have a couple of vendors maybe 4 or 5. We’d show push-button interoperability and in the process do the work that needed to be done to make the API definitions more precise, to fix the configurations, to write common answerable scripts. All of this for real enterprise workloads.

It turned out that a whole lot of people wanted to do it. So we ended up with 16 on stage, 18 that actually completed it and there is more in the queue, which is great. Moving forward, there is still work to be done. In one dimension we need to get the word out to clients and OpenStack users that it’s not a week’s worth of work, or two weeks, or a month or 2 days. I can be done now, for the majority of the cases, very quickly.

IBM has sent an email about this to all our clients and we are not the only one. We have had press and analysts and everybody knocking on the door. There are some analysts that said OpenStack would never be interoperable. We need to get the word out. However, there is also technical work to be done. While we hardened the default options, the semantics and all that stuff, there is lot of work that we need to do in RefStack to ensure that this stays in place. Over the next 6 months to a year we will improve RefStack, adding test cases that essentially demonstrate and ensure that we remain interoperable.

ET:         This current step is just about for deployment isn’t it? I see two problems here. The first is the ability to deploy to multiple clouds using the same script and have a reliable outcome.

The second is helping customers on OpenStack A move to OpenStack B. That is more complex and nuanced mainly due to APIs. It has been suggested that the APIs are becoming a blocker and that we need to almost step back and have a middleware solution to abstract those APIs. Where are we on both of these?

AD:         We aspire to do both and we worked on both. Not just deploying but also the API hardening so that there is less ambiguity and less ability to stray. API compatibility is hugely important. When you look at RefStack and what we are going to do with RefStack, you are going to see a lot going on there around the API. So yes, both problems need to be addressed.

ET:         Do you have a timescale? Will it be the next OpenStack Summit when you will be able to say we’ve done 95% it is just dotting the i’s and crossing the t’s?

AD:         We haven’t sorted that out yet. When we issued the Interop challenge, in a lot of interviews people asked “Oh so what is the timeline?” I don’t know! I think the next Summit would be nice. That’s why we’re here. The guys are working on it right now, so I think I have to call them to see what we got. I can tell you that we, at IBM, are putting a lot of energy into API definition, API interoperability and we are putting a lot of work there as well.

There will be a follow on to this, I don’t know exactly what or how, but there are lots of ideas. People have said, well you know we could set up a cloud that has all these things, we could continuously interoperate, there are a lot of things we could do… we’ll see where we land.

ET:         It’s not just about OpenStack. How do you pull in the non-OpenStack clouds? We’ve had projects to try and make OpenStack work with other clouds such as Microsoft and Google. AWS has now come to the party where they are trying to enable their customers to move in and out of OpenStack. The OpenStack community have said it is about freedom of choice. If these other clouds are more about winning OpenStack customers rather than cloud interoperability do you still make it easy for them to work with you?

AD:        Everybody’s welcome to the OpenStack party. Microsoft has been around for a while, mostly because of Hyper V but they have been around. Google is now getting much more active which is great. Amazon has really not done anything in the community. The way we view this is that the APIware is the lingua franca. That is what we would like to see in cloud provider support. To the degree that Amazon decides to support OpenStack APIs that’s awesome, because then you can use that directly. That’s how I think it should go down. Now, will it go down that way? I don’t know. But that’s what the community has been trying to promote.

ET:         When I look at this market as a journalist and an analyst, I see it as being a long way from the sum zero game. The reality is there is so much business out there there’s no need to be fighting over users. It is about grabbing what is waiting to be taken. That leaves a real opportunity in the API space for product partners and system partners to get involved. They could step in and say, we’ll solve that gap between the APIs, between Amazon and OpenStack or even something else. Do you see, at the moment, anybody stepping up, or is that an opportunity for another community driven project?

AD:         There has been, in the past an Amazon API in relation to OpenStack. It is not like it hasn’t been done, or is trying to be done. It is always easier when the actual provider does it because they understand much better what’s on the other side of the API. It certainly could be an opportunity for someone to do something, but what we try to do with OpenStack is to get the actual people who build the clouds getting the API, or using the code or whatever.

ET:         We have all been involved in projects over the years where suddenly somebody said “oh yeah, I remember I have a patent on that.” So far OpenStack has kept control of that. Is it a concern?

AD:         IBM tried to fix that problem in the early days when we helped set up Apache. When we created the Apache licence we created the notion of the contributor agreement that you sign before you contribute code. It essentially gives away your IP. The way we do it when corporations get pulled in is there’s a special agreement that covers the corporation and their intellectual property. In the early days of open source IBM stood up and said: “We’re protecting open source with our intellectual property, so if you want to go mess with them, you gotta mess with us, and IBM has been No.1 in intellectual property for the past 20 plus years.” We’re very focussed on making sure that innovation can stay open and that people feel free in innovating. That’s how we prefer to do things.

ET:         There is one name obviously missing from the list, in terms of big players, that’s Oracle. They are making a lot of noise about where they think their value is in the cloud. You mentioned there are other people who are just starting to come and talk. Are they among those or like Amazon are they standing off a little bit?

AD:         Oracle was part of and has been part of the community mostly because of the acquisitions they have made. They made some acquisitions that have community members who are still very active. You know, I don’t know what Oracle’s been doing, or whatever, but they have been part of the community because they have people who do that, so I can’t say that they are not, but I don’t know who their clients are.

ET:         Thank you



Please enter your comment!
Please enter your name here