Input sound file:
Session Name: AWS: Amazon’s winning strategy
On the off chance that there is someone in this room that doesn’t know who is up next, the way I would describe it is, I’m not sure there would be much of an IT industry these days without the cloud, and I’m not sure there will be much of the cloud without Amazon Web Services acting as a disruptor and innovator, and I’m not sure there would be much AWS without Werner. Without further ado, we’ve got a fireside chat with Jo Maitland from GigaOM and Werner Vogels from Amazon Web Services.
Jo Maitland 00:42
Werner Vogels 00:43
Jo Maitland 00:44
So we’ve only got 20 minutes, so it’s going to go really fast for me, so I’m going to jump straight into my questions.
Werner Vogels 00:50
Because I’m tired, I’ll speak slowly.
Jo Maitland 00:53
I want to kick off asking you about AWS’ strategy for hybrid cloud. It seems like the model that the enterprise at least wants to consume cloud resources in is this hybrid model, which is a mix of public cloud resources and private cloud by which they mean in their data center potentially private infrastructure– within their own data center or within a host of facility. That seems to be the dominant model which the enterprise is kind of clamoring for. From AWS’ perspective, how are you going to reach those peeps?
Werner Vogels 01:34
First of all I think we’ve been doing quite a bit of these kinds of things, helping people that look for that. I think we always know that data centers will remain around for a long period of time. I just think there will be less of them owned by corporations. There will be more things moving to the cloud over time. Many of our enterprises are not immediately moving all of their service on to this crap or to eBay or whatever. There’s a flow over time and these projects and resources will flow on to the cloud. I think we’ve been investing quite a bit of effort in the past year, two years, making it easier for our customers doing things in the cloud, as well as on premise. Direct Connect allows them to build direct connections between their own data centers and AWS without having to go for the public Internet that can pick 1 or 10 GB connections. We’ve been building VPC which allows them to actually cordon off a piece of the cloud assigned and there are some blocks that would connect them back to the wrong data centers with VPN connection. Storage gateway allows them to run software of ours on a virtual machine on premise and touch their own storage blocks do it and then back them up into the cloud. Our workflow engine allows you to execute task on premises, what was in the cloud, identity and access management with sports federation with LDAP, as well as active directory. We’ve been really listening closely to our customers what are the kind of tools that they want to make sure that they can work on this transition of more and more resources into the cloud, but while still keeping things on premise as well.
Jo Maitland 03:25
What can you do further without giving away future products obviously, but what would you imagine you can do if it extends that comfort zone shall we say?
Werner Vogels 03:37
The comfort zone, I think there’s a lot of focus around encryption. For example, many of our customers have been asking us for a more extensive encryption tool, some of these are coming out of our ecosystem and I actually talked about this last year as well when Om asked me, ” What would see happening in the coming year?” Coming two years, I think honestly the rise of encryption as an end-to-end tool. Many of our customers are already encrypting either whether it’s sensitive business data or whether it is personal identifiable information. We’re offering what’s called the CloudHSM and define a hardware security model such that they can keep their keys in the cloud fully protected, fully tamper-proof. We’ve enabled visit to transparent data encryption into RDS, Oracle as well as network encryption, so customers can actually really protect them. We would give them fine grain tools so that they can protect themselves in a way that they would like to do. You should expect us to continue to draw them out, the range of security tools that we give them such that they can protect themselves in ways that maybe they’ve never been able to do before.
Jo Maitland 04:46
Thinking about Amazon as a whole and AWS with the numbers and the company prides itself on sort of doing things and making things simple and easy for customers, and I think on many dimension, it is excellent to that. Within AWS at this point, the some of the pricing and particularly pricing around reserved instances is really complex and there are 100s of skews. For customers that want to make longer-term commitments to AWS, it’s really hard to do. I’m curious to see how you see that evolving and highly might make it easier for customers to actually commit to you longer term.
Werner Vogels 05:32
I think on one hand, it’s the level of education, just as helping customers to make easier steps for them to do self-service to really pick the instances that they want to help them build architectures that where they can really make good trade-offs between scale, cost and things like that. We need to make things simpler, simpler to use for our customers definitely is something that we’ve been pushing for a while. But on the other hand, customers also ask us for more skews. If you look at the things that we’ve launched in the past year, the high I/O instance, high memory instances, massive storage machines. We’re extending things in the direction as well across Amazon’s quest force. You can’t have won without the other, but we can do I think a little better in terms of helping our customers making the right decisions.
Jo Maitland 06:29
It’s of honor to related to know. I was at Reinvent in November. I think it was the CEO of Netflix was talking with you guys and he made the point, that the business user right now is completely unable to understand Amazon Web Services. So, for example, if I’m Josh Mo in the marketing department and I need resources from IT. I as the business user would not understand what an M1 large instance is, for example. Will it be Amazon that sort of creates the less that makes business users be able to sort of buy infrastructures or service or is that going to come from somewhere else?
Werner Vogels 07:13
I must say there’s two different things. I think on one hand, I see business units going out because they want to do a particular project or they want to develop a new product or service that they want to deliver, often do that together with partners and then select later on AWS resources that are necessary for that. Another thing that these business units do is actually most of the times they’re not really interested in server or an instance, they’re really much more interested in getting, ” Give me a CRM system that is really good, mobile and programmable.” So in the end, all these resources that we have don’t really matter if we don’t do things and if they don’t deliver products, we’ll prove it. Sometimes these business units will develop them themselves, sometimes its partners will deliver these services to them.
Jo Maitland 08:01
Amazon I think you yourself have also talked about AWS being an open set of services and that there’s no lock-in. So that implies that you agree that lock-in is not necessarily a desirable thing. I’m curious as a CTO of Amazon, what sort of principles and strategies you apply within the company in your own consumption of technology to avoid things like lock-in?
Werner Vogels 08:33
Jo Maitland 09:19
That’s how you would advise them, so you use those concepts internally and you would advise them in consuming AWS, I think.
Werner Vogels 09:26
Yes. That’s really focusing on as little, so little dependencies as possible. There’s a saying that [inaudible] gift to many of our customers as they are thinking about how should I automate my services. I think there are a few principles there. Make sure that you decompose things in building blocks over the dimension of A1 to scale up and scale down in, and then make sure that you can actually automate that. You could put APIs on top of that, where it can be business rules that decide how to scale up and scale down your application instead of that it’s an engineer that’s sitting there making those decisions. It’s one of the reasons why for EC2, we decided to go with the mobile that we have as if of a virtual machine, that is. It can be started and stopped in a matter of minutes and everything has an API. Everything has become software, instead of that it is actually read an engineer test to do it, you cannot automate it and have business rules decide. Now I want the latency towards my customer between in this bounds and if that latency changes, I want to scale up. That can be business decisions instead of that it’s a technical decision how you built your systems.
Jo Maitland 10:42
Just on that note starting and stopping machines in minutes, the smallest increment of time though that I can buy an AWS instance as an hour today. I think Google and Microsoft have announced per minute instances. Is that something you can imagine getting to, is that an interesting model from a pricing perspective, charging her minute for an instance.
Werner Vogels 11:04
Well, I think most of our decisions are being made by sort of what is the customer input to us. Until know, I’ve not had much feedback from customers that they want to get to a much finer grain level than the pre-hour one. The only one place where I’ve seen them move away from just short of hourly notice, it seems sort of a higher level services switch as the transcoding service. They just want to pay per job, instead of you just get too worked on. But I haven’t heard much feedback from our customers, so I want that they would like to have a per-minute kind of interface and our customers are the most important thing they’re, more than the pricing model sets off.
Jo Maitland 11:46
Talking about so that we always have to ask you the ambulatory questions around competitive landscape. Last year I think when you were chatting with Om, you referred to the traditional IT guys as dinosaurs. They are starting to move slowly and get with the cloud program a little bit right? HP has launched a public cloud service mow, IBM sort of buying its way in. How does the landscape look to you now versus a year ago and–
Werner Vogels 12:19
We always knew that we’re not going to take whole and take all market, and I think that we’ve always strongly believed in that we should be focusing on our customers, not on any competition that’s entering into the market. The only reason for me to look at any competitors or potential competitors is, why would customers pick another player than Amazon to go with because that might teach me something about my own services. It might teach me why a customer made may choose to go with another provider giving me a shortlist of things that I need to adjust or that AWS needs to address.
Jo Maitland 13:03
Have you seen any of that yet?
Werner Vogels 13:06
Customers in general are pretty vocal towards us in giving feedback about the kind of stuff that they want to address, and in that sense, the roadmap for us doesn’t really change that much. It is really first and foremost still security and operational excellence. That will be forever our number one priority. On one hand, continue to invest in the operational side or security and just general operations in itself. We would rather as I would like to say hold some of the future development and move those resources into there if we feel that, that would be appropriate because security needs to be for everyone, the number one priority. On one hand, the operational side of stuff, but also put in resources towards developing new security tools for our customers such that they can protect themselves and much more finer grain ways than they would ever be able to do in their own data centers. I think the other thing customers ask us about is international expansion. We now have nine regions around the world. We’re not telling what are the other continents that still are left, but it’s obvious that we will continue to work on international expansion. When you mentioned, we’re making it easier to you, so I think launching, for example, OpsWorks last year was clearly addressing some of the request that came from costing us to have an easy application life cycle management environment, which they could be using there.
Jo Maitland 14:34
On your point about international expansion, would it be interesting doable for Amazon to launch, for example, clouds in Europe where there’s sensitivity around particular regulations, around data privacy.
Werner Vogels 14:51
We continue to evaluate those scenarios, what we call them as more community clouds or members only cloud. Where they have a Gulf cloud here, we have a Fin cloud here, that is actually members only dated by NASDAQ and managed by NASDAQ. If there are opportunities, I will start with the US, whether it might be definitely a path wheel, we would explore.
Jo Maitland 15:15
Have you been asked to do that?
Werner Vogels 15:18
We get asked to put a cloud in every country [laughter]. Everybody wants Amazon to be everywhere.
Jo Maitland 15:27
Werner Vogels 15:29
However. [laughter] I think it shows a lot of the passion that many of our customers have for the products that we deliver, the way that we help them interface. They want us to really help them succeed. From some of the countries, where latency is still larger, they would like us to be closer and we continue to evaluate those.
Jo Maitland 15:58
I’m not sure how much you can talk about this, but I’d love to hear any kind of learning that you’re sensing from this experience around the bid over the CIA contract? It’s now you guys have won it and then IBM caused a big fuss and maybe that was sour grapes. They’re kind of getting knocking on the door and making their point about various details in the contract. What currently and it’s ongoing obviously, but what’s your sort of learning so far from this process? Is there anything that surprised you about this or what’s been your learning from it so far?
Werner Vogels 16:37
I think my main learning is that if you have a technological superior solution, that customers who would do deep dive on it, they will actually pick the right solution.
Jo Maitland 16:48
The fascinating thing for me is that Amazon costs more than IBM. That just blew my mind in the contract. IBM is, “Hey the CIA is willing to pay, but it was just fascinating to see that the AWS thing was more expensive than the IBM thing. It was good for you and it was like, ” This costs more than IBM.” I never imagined that happening, but–
Werner Vogels 17:12
[laughter] It’s an ongoing process.
Jo Maitland 17:19
Good luck. It strikes me as–
Werner Vogels 17:22
Again, I want to come back to the point that I think we’re really focusing on building the absolute best technology that almost [inaudible]. I think in this case, this was a very long process where the customer went through many months and when we really deep dive from their requirements and the technologies that we have presented to them by many different vendors and they picked Amazon’s because we could get the job done for them with superior technology than what they are able to get anywhere else.
Jo Maitland 17:58
I think we’re slowly running out of time. I’ve maybe got one more question here for you. Tell us about like enterprise production workloads on AWS. We hear a lot about test and dev workloads and there are enterprises, lot of enterprise workloads running on AWS, but then there’s also a conversation that we hear where at some point it gets more expensive to run it on AWS and people bring it back in house.
Werner Vogels 18:25
Jo Maitland 18:25
Werner Vogels 18:26
Give me an example.
Jo Maitland 18:28
I don’t necessarily want to–
Werner Vogels 18:31
So it’s not the conversation that we would know that and to be honest, I would be really truly honest about this if that were to be the case because that will be great learning for us. I don’t know about any of those cases, that to start off with. There is almost no vertical these days that is not making use of AWS in a variety of forms. If you look at life sciences, yes they’ll be doing genomics, but they would also be doing collaborative workloads and they will be doing clinical trials and they’ll be doing HPC. Pick any other for it, hospitality. We spend lots of time in the hotels you know, four seasons wins all the digital assets on us. Intercontinental Hotel Group, we’ve close to 65, 000 hotel rooms means all of their reservation system. Kempinski, the high-end hotel chain just decided to move all of their core IT, 100% of their core IT, ERP, everything into the cloud because they don’t want to be an IT company. Hotel Logic is a software service provider that does property management for boutique hotels. Here, you look the whole gamut from successful startups like does successful startups like B&B, Hotel Logic, the biggest hotel chains in the world, all making the use of this but for different use cases. From really core IT, reservations systems to digital management, the whole gamut is being run on AWS.
Jo Maitland 20:06
So do you think the Zynga story which is old now, but do you think they were an anomaly that sort of good’s idea that they figure out what the capacity for their new games is going to be on AWS once they figure out the capacity, they decided it was cheaper for them to build–
Werner Vogels 20:21
I think you have to talk to the Zynga guys. I’m not an expert–
Jo Maitland 20:25
But not from a workload
Werner Vogels 20:26
It’s not a pattern that I’ve seen it coming back anywhere else with any of our customers. If so, we would have to address that of course, because the customer happiness is the most important thing here.
Jo Maitland 20:39
I’m out of my time.
Werner Vogels 20:40
Have a seat.
Jo Maitland 20:41
Werner Vogels 20:42
Is that all.
Jo Maitland 20:43
I should be going, but what a talk. Thank you Werner.