This is the question Alistair Croll first asked Werner Vogels in the fireside chat session at ECS. Werner admitted he’d been caught off guard by the question but admitted that the future is automation for sure and scripts are powerful tools to achieve this.
An example of an enterprise use case for cloud computing
Werner related the case of the NASDAQ which had a lack of capital, which was restricting innovation and making it difficult for them to solve the technical problems around handling complex historical stock queries.
They solved this by having every ticker symbol for 10 minutes written to a text file in Amazon S3. An Adobe Air application was created which allowed you to specify a symbol and a date range. The app would download the text files for that time period – meaning you can do joins, queries etc. The computation is done by the customer’s desktop which means there is no resource investment. They were able to use cloud technology to keep things “nice and simple”
Cost savings can include people
Werner talked about the idea that when assessing the cost of cloud computing versus in-house infrastructure, you have to think about the total cost of ownership not just hardware. Werner talked about the example of the Indy 500. He said they have a very nice website which offers a flash environment with multiple video streams including views from the cockpits of drivers’ cars with audio feeds and telemetry. This is a high load application but it only runs three times a year. They found that they had to move a lot of engineers into data centers to keep their servers up. When they moved to cloud infrastructure they made 75% cost savings, the majority of which was on the people side; now they can manage everything from their armchair at home.
On Amazon direction and strategy
Amazon created these cloud services because they needed them for themselves. When you go to amazon there are 200-300 services being accessed to create that page. They did a deep dive into looking how the engineering organisation was working and found that 70% of the time that people in those teams where focussing on infrastructure level. Amazon made a conscious decision to virtualize the infrastructure platform – which is what brought about Amazon EC2. Werner said “When we did EC2 it was as important to scale down as well as up – and to do it in a matter of minutes. Scaling down hadn’t been done much before. things were left running “just in case”. This means significant savings can be made by reducing this “hoarding”.
Werner noted that another reason they’ve rolled everything themselves, is because control is important – performance, cost, reliability. 3rd party services become a black box. They were not built for 80,000 transactions per seconds against S3.
On infrastructure as a service versus platform as a service
Amazon has an internal principle of avoiding lock in. “We do not force an engineer to use any particular technology”. They can use whatever works best. He noted that to be able to do that you can’t offer an API forced to one programming language externally. “It’s important to provide an environment where anything is possible, even shooting yourself in the foot. We don’t believe in lock-in.”
Werner cited the example of Stax, which is a J2EE as a platform service – you can move your J2EE standard app into the cloud without any re-coding. He noted that Amazon didn’t build it but it happened because there was an ecosystem. If Amazon had started higher up the stack this wouldn’t have happened. Amazon makes strategic investments in the cloud ecosystem.
Alistair asked how do you decide what to invest in, using the analogy of the roofrack and the speedometer. Some things are better “built-in” and others are better done by third parties? Werner said that the fact it is a growing ecosystem means people will choose what’s best. Choice and competitions is good.
On the future
It doesn’t matter whether you view providers as engineers or systems integrators. There’s nothing bad about helping customers being successful. Werner observed that consumer power is rising. This means there is much less certainty about whether something is going to be a success. He is sure that tremendous value will be delivered. But it has to come from somewhere. We just have to do it. Innovation is key and people should focus on building value add applications.
He gave an example of this: College IT. Many large universities, pretty much every university in the US, manages their own email system. It’s historical. It used to be that students’ first email address was from their University. Now students already have 5 email addresses when they come into uni – they hate having another service (although they do need a University domain for their address). CEOs from Universities want to hand the email provision out into the cloud. They can use the specialists – and re-divert those internal resources. This is a classic example of focussing on what’s important to your customers, deliver more value.
Constraints breed innovation. Cloud research/OS research is not over. Looking at the precursors for cloud computing, there were a number of things that needed to happen: Virtualization needed to mature (CPU, storage, network, security). SaaS needed to mature. Needed to get experience with delivering very large software systems over the internet. Service oriented architectures had to develop – ie. the standardised set of protocols to access a service. Distributed computing needed to make a lot of progress so we could deliver these storage services at this scale – we couldn’t have done this five years ago.
Exciting things are happening, such as voice recognition as a service, location based services.
Another example: Delivery companies lose money because 8-10% of deliveries reach a closed door. They didn’t want to create call centres, but could still save money by automating a call to check if customers were in before attempting to deliver.
Now we are getting a lot of aggregation of different services – offered by people who historically wouldn’t have offered these things because it’s not their area – like mashups but real services that offer real value.
Business-driven technology is good. Business comes to you with real problems to solve. Not because it’s cool technology and we want to see what we can do with it (as happened in the 90s when we bought software and didn’t use it). We are focussing on delivering value to our customers.
IT will be a driver for the adoption of cloud tech. Customers are choosing it because they get a higher level of integrated security than they would be able to achieve themselves. For example, Wall Street. No data center space left in New York. These firms have capacity in 15 different data centers. This is not optimised for security. When a security breach happens its too late. This gives them a solution.
Werner concluded “This is Day One of cloud infrastructure. We will continue to listen to our customers and evolve our services to meet the needs of real customers, and solve real problems today. Day two may be a number of years away, so we will see what happens.