The Amazon Web Services summit in London was held earlier this week and CBR got the chance to sit down with AWS’s Carlos Conde, Head of EMEA Evangelism, to discuss why people are so keen on the cloud.
Q: What made Amazon, as an online marketplace, so well-placed to move into cloud?
Before being a marketplace, Amazon was a largely distributed eCommerce website, not only in the US, but also here in the UK and in France, Germany and many other countries. To operate at this scale, we had to operate our own data centres.
So we gained, since 1996, a lot of experience at operating data centres, and not only data centres but highly available and highly scalable websites and extremely secure facilities, because we are processing credit cards around the world.
We always had this philosophy that, as soon as we are doing something in a very efficient way, we open it for our users. As you mentioned, the good example is the marketplace. Amazon was an eCommerce platform and we realised that this is something you could offer to other partners in order for them to sell goods from the eCommerce website, so we opened the marketplace.
Amazon Web Services is the same logic. But it’s important to realise that AWS is not excess capacity from Amazon. It’s a complete service that has been built, and Amazon is a customer of AWS.
To completely answer your question, it’s important to keep in mind the business model of Amazon. Our model is low margins and very high scale. This is important to keep in mind, to put it into perspective with other traditional IT vendors which have the complete reverse business model, of low volumes and extremely high margins.
We definitely feel that cloud is a high volume business with low margins – it’s a retail business. That’s why for us, as an online retailer it is a logical move. Since day one we’ve been operating at extremely low margins with very high volumes, so now we’re also doing it for computing.
It’s these two sides.
Q: Why is flexibility so important to businesses nowadays?
Flexibility and agility are some of the main reasons why enterprises and start-ups are using AWS. It’s not only the flexibility and agility regarding the price. If you compare using AWS to a traditional IT approach where you have to provision and buy machines and so on, you of course have the difference between operational expenditure and capital expenditure, so this gives you a lot of agility in terms of economic perspective.
It’s also important in terms of speed of iteration. This is something that many start-ups, and now big enterprises like the Financial Times, have also learned. The speed of iteration beats the quality of iteration in terms of agility in the business. If you’re able to push data and change the software in your internal applications or in your public applications it makes you more agile and you can do more trial and error.
So, on the one hand you have the business agility that the cloud is way more cost-effective. Imagine you try a new business intelligence project and it doesn’t work, so you pay by the hour, it’s not too bad. You’re not tied to anything.
On the other hand, the speed of iteration comes from the fact that if you start a virtual machine, it takes a couple of minutes, maybe less. If you compare it to traditional organisations requesting a new server, that can sometimes even take a month.
It’s definitely a different story. You have a developer, you want resources, you have them now. You go with the traditional way, you have to do the request, call for tender, rack it, mount it…perhaps after a month you can work.
This flexibility comes from different angles. So you have the economic advantages, the speed – the third one is the auto-machine. In the last two years we’ve talked a lot about dev ops and automating operations; since day one, AWS has been offering a set of services with an API. Everything you do manually on a web console, this is something you can programme as well. So it’s basically a programmable data centre.
Q: Could this fundamentally change the business model for start-up IT?
Particularly for start-ups who base their activity on mobile applications or internal applications or any kind of IT-backed activity, it completely removes the risk. Traditionally you have to go to a founder, ask for money, invest this money in hardware and if your start-up doesn’t survive the first year or even less, well what do you do. You have all of this investment which you basically keep in your hands.
Because you pay only for what you consume and you have only the monthly Amazon bill, first, if your start-up is small you pay only for what you consume. But if you continue to grow, it adjusts automatically to what you need, which is important for start-ups. It’s doesn’t happen that finally your start-up becomes extremely popular and you’re a victim of your own success because you didn’t provision enough capacity. Here we’re going to be able to give capacity automatically to start-ups who need it.
They all have the same scheme – they started small and as they became popular they began scaling to massive loads.
Q: That’s the start-up model – what’s the appeal for enterprises?
More and more start-ups want to have the same processes and quality of operation as enterprises and enterprises want to have the same flexibility as start-ups. So, it’s very interesting to see that even companies like the FT and BT, they have the same problem. Even if you have already the financial capacity to buy any hardware you want, you still don’t have the flexibility. You have those internal processes where you’re not able to provision capacity as you want.
For many of them, the way they’ve started using AWS is for test and development requirements. With AWS we have many different services and features that allow you to establish a hybrid connection between your physical data centre and AWS, so you can do a float capacity. If you look at many companies that are in the financial industry, they have these peaks of load.
This means that their IT usage model is not effective. One classic example is banks – when the trading day closes they download all the information from the intraday and they do risk assessment and analysis. Since the crisis in 2008, many banks have to prove their solvency and not only banks, but also insurance agencies.
They had to get all the data from the intraday and run risk assessment analysis which lasted something like 23 hours. So basically they were one day late. Now they run this computation on AWS and it goes down to 20 minutes. If you have only periodic peaks you have to buy the capacity that matches the top of the peak. When you’re not using this hardware it’s money wasted. But if you can only pay the hour when the day closes you can download the data, provision the resources on Amazon, run your computations, and then you just terminate it.
Big companies not only move to AWS because of the dev and test environment and because people can try new projects through the hybrid architecture, but because they can also do mission-critical applications in a way that wasn’t possible before. Yes they could have computational clusters, but it was extremely expensive to operate. As the amount of computation grew over time they needed more capacity.
Now for companies here in the UK, cloud has become the new normal. It is not a question of whether they are going to use it or not.
More and more CTOs start to understand that operating a data centre and doing backups, maintaining the networks and so on are not part of their core business, they are what people call undifferentiated heavy lifting. It’s things that you have to do but don’t bring any value to your core business – so don’t do it.
Q: Surely the amount of processing that a big company, such as the Financial Times, has to do is fairly consistent?
If you believe that a company stays in the same operational model, you could say that it’s not going to change. But the press and the media industry has been heavily disrupted by the internet.
Now the Financial Times or the NYT and many that we have running on AWS: having their newspaper available online or mobile applications is really important because of targeted marketing operations. So suddenly they have way more data on their hands to be able to intelligently suggest content to the reader.
Last week we announced in San Francisco a service on machine learning. This is the kind of service that media uses a lot in order to suggest content or to personalise content for the reader. So perhaps they will have a slightly stable amount of users but they now do way more IT operations to serve those users.
If you look at not only the traditional press, but also the traditional TV media – take the BBC, which is also using AWS for iPlayer. This was a business core application for them, because by processing the content on AWS they went from ten hours to twelve minutes to serve the content to their users.
So before everything you had on the replay you had to wait half a day, and now it’s available ten minutes after the show. Now the transcoding operations and the processing of the high resolution video they have, they do it on the cloud.
What the BBC does is take, say, a two hour video programme, they chunk it into five minute sections, they parallelise the distribution of each block of five minutes to separate instances, transcode it, reassemble the video, and ten minutes later you have a multi-gigabyte file that is transcoded for multiple formats.