Microsoft’s cloud strategy four years on: Q&A with Azure GM Steven Martin
This month, Microsoft’s public cloud and platform as a service business turns four years old. Microsoft has been fairly cagey when it comes to revealing user numbers but Windows Azure general manager Steven Martin told Business Cloud News that the overall compute footprint has consistently doubled every six to nine months since the service began running. BCN spoke with Martin to learn more about how far Microsoft’s cloud business has come over the past four years, and where it’s heading.
How would you characterise the Azure platform’s growth over the past four years? Are you seeing a tendency towards certain kinds of workloads or applications more than others?
It’s been a lot of fun watching this business grow. When we got started we had five people working on this, now we have a much more substantial team. Having worked at Netscape at the height of the web, I’ve been fortunate enough to see phenomenal growth before. Demand for Azure continues to double every six to nine months, and we’re still waiting for that point in this hockey stick where growth dips. So we’re still in the early days of this growth.
The focus that we saw early on was with things you would imagine – organisations that needed bursting capacity for short periods of time: actuarial customers doing things in insurance, media companies doing video encoding where it was time sensitive and they needed a lot of capacity but not for particularly long periods of time. Over the last year more line of businesses have been using cloud computing as a default IT shop, and so we saw quite a bit of transience early on, and now we’re seeing much more sustained usage.
IaaS seems to have come after the initial Azure service launched. Is Azure still primarily focused on PaaS in terms of long-term strategy?
First thing I would say is our focus is on IaaS and PaaS. In terms of our evolution, we had a very specific, very deep focus on PaaS from the outset. There are lots of companies all over the world that stand up servers and do a great job of hosting Windows and others, lots of options for customers there that want to rent capacity from providers. But PaaS takes unique IP, we wrote a ton of new technology for developers that wanted to move up the stack and not think about how to configure load balancers and patch things and whatnot. The market tells us that PaaS is ultimately going to win.
Our view of the world was, there are lots of places where people can get IaaS directly. But it was an area where we were really pulled into by our customers.
Are there any big inhibitors to cloud-based PaaS uptake in the near term?
The truth is, I don’t necessarily want every customer in my cloud. I want them to bet on a Microsoft stack, or a stack that uses a variety of Microsoft technologies to help solve particular problems. If they find the economics of cloud something that meets their needs, I’m glad to host their work. If they want to use it as a dev and test environment, I’m glad to meet their needs. I haven’t met anyone who said they’re going to turn off their datacentres and move everything to the cloud, it’s going to be a long term transition. My view? Let the economics decide – if it makes sense, great. If not, we’re delighted to participate in helping customers run things on premise. Ultimately, we want the economics to be the deciding factor, not the technology.
Often the problem with PaaS is that it forces developers to use proprietary APIs linked purely to that platform. Any time that companies have constructs that force architectural decisions on developers, developers begin developing around business models and only bad things happen. We just want to take that out.
What have you seen so far with Azure in terms of hybrid cloud deployments?
It’s still early days for hybrid cloud. In most cases what we see are organisations using some of our finished services, taking say, Azure Active Directory service and using it in conjunction with applications that might be hosted in other cloud environments. But this will likely change over the next few years.
Given the enterprise focus on mobile and flexible working these days, coupled with Microsoft’s recent reorganisation and acquisition of Nokia, it seems mobile will figure front and centre in much of the company’s future activities. How will mobile fit with Microsoft’s cloud strategy moving forward?
Mobile creates some interesting opportunities. Most of the mobile interaction is in low latency configuration. So you may have a mobile application, could be a video or a game, something highly interactive where that particular workload needs to be relatively proximal to the customer. That’s where the datacentre strategy matters a lot. It was the reason we put a datacentre in Tokyo specifically – there are more cost effective places to put datacentres in Japan, but we know that this area focuses on mobility and proximity to a major metropolitan area was critical.
We also know the majority of developers developing for mobile understand the mobile environment well. But what we find is that they’re not particularly great at building back-end applications. We see a lot of focus on UI and experience but the backend piece is a bit of an enigma to most, which is where the Azure backend piece comes in. We want to continue making it easy to build these applications and add sophisticated mobile services without having to be particularly sophisticated on backend technologies.
Where does Nokia fit into that? Do you see a bigger opportunity in driving more integration between software and hardware?
We do think we can provide some unique value for people using our front-end and back-end simultaneously in terms of developer productivity because we do participate on both sides of that. So those two organisations working closely together can tell the developer what it takes to build the front end and the back end, and it’s becoming a very cohesive environment.
We also recognise that this is a highly fractured market and we don’t want to make developers feel that making a choice on Microsoft locks you in to a particular device, front-end or back end. This is and will remain a heterogeneous market for the foreseeable future.
How do you see Azure growing over the next year?
We’re going to add 14 new datacentres over the next nine months. A bunch of that is going to be in existing locations, but we now regularly have customers using fifty thousand cores simultaneously – so the equivalent of 6,000 servers will go up and down for some customers. We will add significant capacity in Japan, Australia and Brazil this year – a physical presence in those areas for the first time. There will be more, and we’ll announce those closer to the time. We’re also going to keep building integration between various cloud and on premise products and services.