People

At the GigaOM Structure08 conference in San Francisco, we had the opportunity to question Amazon’s CTO Werner Vogels on his virtualization experience, while building the Amazon cloud. He confirmed Amazon Web Services are still powered by Xen hypervisors.

It is remarkable to hear the CTO of a multinational openly thank the open source community for their active support on Xen and hear him claim this to be the main reason for having chosen Xen as a crucial Amazon cloud-enabling building block.

Werner Vogels CTO Amazon.com from Toon Vanagt on Vimeo.

As we reported earlier, Amazon is also very open on its performance and welcomes independent companies to measure and report on parameters for public virtual computing facility such as security, availability, scalability, performance and cost.

Werner finished our video interview by explaining why cloud computing is even disruptive outside of the datacenter and transforms unexpected industries. Venture capitalists seem upset about side effects, such as start-up funding independence, as these fast growing tech companies are no longer in need to burn lots of VC-money on hardware platforms and technologies upfront. They can now scale their offering dynamically, driven by organic growth, while generating the necessary revenues to cover the extra cloud cost.

At Virtualization.com we like to think that “shift happens” and look forward to the upcoming VC-riots on Sand Hill Road against these unthankful self-sufficient start-ups 🙂

A full transcript of the interview is below. If you are interested in Amazon Web Services, you might also want to participate in our contest to win a free book, dedicated by Werner Vogels.

(00:00) Werner Vogels, welcome on Virtualization.com. You are the system administrator of a small bookshop. Could you tell us something more about yourself and on how you virtualized your infrastructure to such a dimension.

“I am the Chief Technology Officer for Amazon.com and I am responsible for the long term vision for technology within Amazon as well as how we can develop radically new technologies to support that business. But also the kind of businesses Amazon could move into, because of the unique technologies that we have developed.”

(00:33) Werner, I am a bit puzzled, because I did an interview with Xen founder Ian Pratt and he told me that Amazon is using this extensively. In your keynote here at the GigaOm Structure08 conference you just claimed you’re using no more third party applications. Did you refer to Xen in that respect?

“My remark about third parties applications was more about our enterprise stuff, where you look at databases and middleware… We do use some third party software and Xen is one of those. But we use them in the mode everybody in this world is using them. We don’t put these types of technologies to the extreme, because we want to make sure their vendors can support us, in a way they support any other customer they have. The remark I made this morning was more about when you really start pushing technology to the edge, we cannot blame vendors for not being able to support us.”

(1:30): How hard was it to integrate the Xen hypervisor into your cloud platform?

“I think Xen is a great product. It is easy to use. But most importantly is the very active community around it. I would not say many ‘issues’ around using Xen, but ‘challenges’ are addressed there with the things every virtual machine has to deal with. Things such as: I/O-issues, guaranteed scheduling issues, domain zero security concerns,…The community out there is very helpful. That was a very big reason for us in selecting Xen.”

(02:15) With “Security”, you just mentioned one of the big Virtualization issues at stake. How do you make absolutely sure that VM’s are isolated in a mixed customer cloud environment? Is Amazon using VLans to do achieve that or did you design proprietary solutions or techniques you can share with the community?

“It is our policy not to discuss specific security techniques. Except for that we have done extensive software development. To make sure that we can audit, maintain and manage the security issues.”

(02:45) You see this as one of your competitive advantages?

I like to believe that security is one of the main concerns and you have to address those upfront. There is no excuse. In this world of cloud computing the most fundamental promise needs to be that it is secure!

(03:10) Yesterday CloudStatus was launched and I imagine you are aware of this? Is Amazon happy about that?”

Absolutely, we love them. But I want to take a step back there. It is very important with things like CloudStatus, that they are actually reporting on things that make sense for our customers. So we are looking forward to working with them and to bring them into contact with our customers and to make sure that the things they are reporting on are useful to our customers…”

(03:40) You would like to advice CloudStatus on the Amazon parameter set they should be reporting on?

“It is up to them off-course. This is not going to be a winner take all business as there will be many cloud providers in the future. As I mentioned in my talk, we will be measured on security, availability, scalability, performance and at cost. So it is very important that we have independent companies measuring these kind of things.”

(04:18) When you talk about independent companies and open alternatives, one of the general concerns remains vendor lock-in. With Eucalyptus there is an open source equivalent, which sort of reverse engineered your APIs (Application Programming Interfaces) and is compliant with Amazon. Do you think that these options of knowing you can in-source your cloud if needed, helps to comfort prospective companies in selecting a cloud provider?

“Let’s first start of with the notion of vendor lock-in. As I mentioned in my talk, I like to believe that Amazon works very hard to provide APIs, which are so simple that there is hardly any vendor lock-in. We use standard techniques to give people access to our APIs. If you look at Eucalyptus, their need came out the schools, involved in high performance computing, on the one hand want to use the public cloud for doing parallel computing, but on the other hand one to keep a similar interface internally. I think they have been very successful to actually make sure that all these schools  adopt this same model.”

(05:32) A last question on your disruptive cloud platform. Could you explain how this technology also disrupts start-up funding cycles and the move from the CAPEX to OPEX expense models? [A capital expenditure (CAPEX) is the cost of developing or providing non-consumable parts for a product, service or system. Its counterpart an operating expense (OPEX) is an on-going cost for running that product]

“Last night I was at a reception, where a venture capitalist walked up to me, who said he hated Amazon, because we killed his business. After we talked for a while, he actually had to confess they also have to adapt to this new world. Where in the old world, they could lock themselves into a company; get their hand on a large part of the equity, because those companies had to spend a lot of money on resources upfront. What we see now is that the availability of these services makes companies start to think differently. Before start-ups maybe had the idea that the only way they could be successful was to have a very big exit. For that they needed a lot of hardware and lots of investments. Many companies based on the fact that these services are available are now moving to a model, where they think they can build a sustainable business. Maybe we can build great products and charge our customers for it. And if you then attract more customers, you spend more on the (development) of these services. Which is just fine as your income follows your customer needs.”