Think cloud computing is about getting rid of IT infrastructure to save money? Think again.
“We see very little of ‘let’s move workloads to the cloud so it is cheaper’,” says Gartner analyst Gregor Petri.
“Most of the workloads in the cloud are new ones. The promise of cloud computing for enterprises is that they can become digital enterprises, and move their core propositions to become digital.”
Let's get digital
Becoming digital has two aspects. One is about rethinking business for the cloud and mobile era, and the other is about internal systems.
“They are two separate trends. One means that you will reach out to your customers. The other means that you will do less and less internally because it is no longer practical or feasible,” says Petri.
An example of one of these internal processes is expense management. The old way involved an expense management system in every local office. Now companies can just subscribe to an expense management service like Concur. When you get an invoice, you take a picture of it with your phone. The expense management service picks it up and does all the processing.
A similar logic applies in other areas, such as logistics. “It is not efficient if I take my 200 shipments and optimise routes for them. It is much more efficient if someone in the cloud takes all 6,000 shipments for a certain town and optimises the delivery schedule for that,” says Petri.
“A lot of the electronic manufacturers are working with companies such as GT Nexus, which runs a cloud network that has all the third-party logistics providers. I go there with my loads, saying for example this is what I want to ship from China to New York, and they embed it in their overall optimisation.”
Traditionally, cloud computing has been divided into three types. Infrastructure as a service (IaaS) is about using virtualisation to run computing workloads on cloud infrastructure rather than in your own data centre. Amazon’s EC2 (Elastic Compute Cloud) or Microsoft’s Azure virtual machines are examples of this model.
Platform as a service (PaaS) provides a higher level of abstraction, with your applications running on a provider’s hosted services. If you are not having to think about virtual machines and operating systems, then you are doing PaaS rather than IaaS, though the extent of the abstraction varies.
Software as a service (SaaS) is where you are simply a user of a provider’s application, typically paying per user per month. Salesforce is SaaS as well as PaaS, showing how these categories can overlap.
Gartner and some other analysts add a fourth category, business process as a service, which means business operations such as advertising, payments and procurement are outsourced to cloud services.
So the phrase “adopting cloud computing” means little. Simply moving some workloads to IaaS or even PaaS models might achieve cost saving but has little impact on how a company conducts its business. Where the cloud can be transformational is in enabling new market opportunities or ways of working.
“The venture capital community will no longer give a startup funding to buy IT infrastructure"
There are numerous examples of businesses that could not exist without cloud. “We have many customers that are built in the cloud, people like AirBnB, Foursquare and Supercell,” says Amazon Web Services (AWS) technical evangelist Ian Massingham.
He adds that in the mobile app and gaming sector, many companies are cloud-dependent, and that for startups there may be no choice.
“The venture capital community will no longer give a startup funding to buy IT infrastructure. It doesn’t make sense, because you can bootstrap in the cloud at very low cost,” he says.
Mobile apps are a good use case for the cloud because they need to scale to unknown numbers of users.
“There is a lot of risk if you are a well-known brand. You don’t want to find customers cannot run the app. Building on the cloud is a great idea because it provides elasticity,” says Massingham.
He adds that the value of simple cost reduction should not be dismissed, since it may make sense of projects that were not otherwise viable.
“The Financial Times is running analytics workloads on Amazon RedShift. One is a personalisation workload. If you read a copy of the FT on a tablet or on the web, it will recommend articles based on what you have looked at before.
“Because they can use their SQL skills to build their analytics model, the price of Amazon RedShift is about a tenth of what you would pay for a traditional data warehouse and that makes it applicable. If you had to build it with a traditional warehouse you wouldn’t do it.”
Scaling the heights
Tesco’s movie and TV streaming service, Blinkbox, moved its IT infrastructure from a London data centre to Microsoft Azure. The company claims it will save “millions of dollars” when it unplugs its data centre, which it hopes to do by the end of 2014.
That saving, though, is less important than the ability to scale according to demand.
“Our biggest growth blocker was storage,” says Jon Robinson, group head of IT at Blinkbox.
“We have tens of thousands of customers watching shows simultaneously and it was very difficult for our storage systems to keep up. We were running out of I/O and didn’t have the capital to purchase the amount of storage that we needed.”
Moving its video encoding and streaming application to Azure, where it runs in two Microsoft European data centres, gave Blinkbox the ability to scale its streaming infrastructure on demand, as well as the geo-redundancy that would otherwise have been unaffordable.
This kind of elasticity is at the heart of what cloud can offer, according to Bloor analyst Dai Clegg.
“I worked with a company that built voting technology for TV talent programmes such as the X Factor," he says.
"The capacity they need for the 24 hours around the show, when they are taking votes and comment feeds at tens of thousands of events per second and providing instant feedback in the studio, is maybe 100 times more than what they need for the rest of the week, when they are crunching the data and updating the app.
“If they had to pay for the peak capacity all the time they could not afford to deliver the app at an acceptable price. The business depends on cloud elasticity.”
Another key cloud opportunities is personalisation. It is a touchy subject, since personalisation has privacy implications, but reframe it in terms of building a closer relationship with the customer and it is hard to resist.
Cloud-enabled analytics is transforming marketing and companies such as ExactTarget (owned by Salesforce) specialise in precise targeting of vendor communications. If they get it right, they are more welcome to the recipient as well as more effective for the business.
Infometrics is another cloud service that creates customer profiles based on the contacts a business has with its customers. Global dairy foods company Arla, which supplies brands including Lurpak and Anchor butter, used to lose contact with consumers once its product reached the retailer.
The company now uses an online recipe database to build a relationship with its end customers, resulting in personalised offers that have a high take-up.
One reason for the slow adoption of cloud services is the complexity or unsuitability of certain applications.
“You can get a spaghetti of dependencies between different applications that can make it hard to move that infrastructure to new platforms,” says Massingham.
Petri observes that migrating applications that were not designed for the cloud does not yield much benefit. “They were never built from scratch, like Netflix for example, to scale up from a few users early in the morning to 170 million users at night,” he says.
Migrating existing applications can mean making architectural changes as well untangling dependencies. Cloud applications typically require a stateless design so that they scale easily according to the level of workload.
Massingham points to tools such as elastic load balancing and autoscaling on AWS, which are effective only with correctly designed applications.
But complex migrations are still possible. “National Rail Enquiries was migrated to AWS by one of our partners, Smart421,” says Massingham.
“The UK’s rail enquiries site is the only journey planner that adapts to the current state of the transport network. If there is disruption it will adapt and give you a feasible route.
“It does that in near real time, so it is a complex system with many data feeds reporting on the the rail network. That was successfully migrated last year.”
Besides scalability, another aspect of moving applications to the cloud is the ability to test variants of the same public-facing application to determine the value of new features or designs. Setting up the infrastructure for this could be prohibitively expensive, but using cloud resources you can run such tests for a short time with little extra cost.
Another benefit is rapid deployment. At a recent AWS DevOps event in London, takeaway company Just Eat showed off the agility of its systems.
“It showed a live demo of its development pipeline, where it changed something in one of its codebases and pushed that live during the event,” says Massingham.
"In the background, it committed a change to its source code repository that was built with a continuous integration tool and deployed automatically.
“In this case it was into a QA environment but it could equally have been into a production environment. It took about 10 minutes."
While cloud computing is not essential to continuous deployment of this kind, Amazon’s APIs are now sufficiently well known to be well supported by tools such as Jenkins and TeamCity which make this kind of process possible.
Do the automation
Technical benefits are fine but what matters is not cloud transformation, but digital transformation enabled by cloud, according to Petri.
“The infrastructure cost is minor. Out there in the market, your competitors are thinking about how to become digital,” he says.
“If the IT department starts to worry about how to take its existing stuff to the cloud while the rest of the organisation is thinking about all these exciting things to do, it is focusing on the wrong problem. Organisations are appointing chief digital officers as change agents.”
Digital transformation means automating and integrating processes. Internet of Things sensors create greater possibilities for automation and faster response to incidents, he adds.
As business processes become commoditised organisations need to think about how to differentiate themselves.
“Everyone thinks of Nike, for example, as a shoe maker but the company feels that gluing plastic together into shoes is not what makes it different. What does make it different is the design and all the stuff that sits around it,” says Petri.
“Individual tasks that used to be part of the cost of doing business – an expense management system, a truck to deliver stuff to customers – that will be less and less necessary because you can use these large cloud services to do that.
"These cloud services will be highly automated, because they have scale. Think of Google Maps: you cannot build it for one company, you have to build it for the world.”
Cloud computing is not a new way of doing computing but a new way of doing business. ®
Find out about the largest Amazon Web Services public sector cloud event in the UK here.