Everywhere we look, cloud is the big buzzword. Digital transformation insists that we embrace cloud computing as the next evolution of our enterprise. Microsoft, Amazon, Citrix, and VMware, among many others, are all focused on cloud and mobility as the next logical step forward. Upstart companies like Netflix are the trailblazers, embracing the cloud completely and putting less agile competitors like Blockbuster to the sword.
We are being fed a diet of marketing that tells us we should ignore the cloud at our peril. In response, what are the pertinent questions we should be asking to ensure that we make the right decisions for the future shape of our enterprise IT?
Firstly, is the future irrevocably tied to cloud—the public cloud—as many in marketing would have us believe? Is it good for us, or is it good for the companies providing it? Now that VMs are easy, bandwidth is faster, and hardware is cheaper, do we simply abandon everything we have in-house and move all that expertise to “the cloud”? Is it feasible to send all of our workloads to an online hosting model?
Unfortunately, workloads are tricky things to pin down. For every Netflix, which by its very nature is ideally suited to cloud computing, there are many other industry verticals that aren’t so well suited. Real-time scientific data, super-high-res printing, trading floors, manufacturing—these are just a few examples of areas where the drawbacks of cloud solutions may make them unsuitable for particular workloads. Of course, everyone has some workloads that are suited to cloud computing, email and instant messaging being two of the most fitting. If Azure has an outage and email is offline for a day, then it isn’t the end of the world. Cloud services are ideal for workloads that don’t need to be live 100% of the time. Even for things like backups, if it works 99% of the time, there won’t be a huge problem.
However, if a company has a business-critical manufacturing process and loses that capability for an hour or two, there are serious ramifications for profitability. Anything you need constant access to or that is intensely sensitive to latency is an area in which an outage will have real, damaging effects. For these workloads, it still makes more sense to run them locally.
If you’re in an industry in which some workloads absolutely need to be run locally, then you have an investment in shop-floor IT to make, anyway. If you already have local IT and on-premises skills, then for every workload that could be cloud hosted, you need to ask whether it is more feasible to run it locally or in the cloud.
There are other questions that need to be asked as well.
It’s not a technical consideration, but trust has to be central to cloud adoption. How much do you trust Microsoft, Citrix, Google, VMware, Dropbox, Salesforce, Amazon, etc. with your data and your infrastructure? Do you trust them to act ethically when they are faced with demands from US law enforcement for access to data held in overseas jurisdictions, as both Microsoft and Google have been recently? How much do you trust agencies like GCHQ and the NSA to avoid making cloud repositories high-value targets for espionage?
How much will it cost? There’s not just the cost of bandwidth, but also the cost of running the infrastructure itself (and all the little extra costs that keep being added to it, of which cloud providers seem to be so fond). How will that measure up to the cost of running locally? Is it possible to predict the investment you’re going to have to make with any accuracy? On top of this, what do you do with the existing investment you’ve made in on-premises kit? Do you simply abandon it?
What will the support be like? Will you be tied to the same SLAs you get with local infrastructure? Usually on-premises support can diagnose problems quite quickly, mainly due to familiarity with the systems in use. If there’s a problem with an application, they can usually address it at the source. With a cloud-hosted system, in which support teams invariably must manage multiple tenancies with disparate solutions in place, how long does it take to diagnose an issue? It could be the client endpoint, their router, their ISP, their account, the authentication service, the hosted application service, or the remote data center. Will the added complexity of the cloud-hosted solution translate into longer delays while service is restored?
Is security necessarily better from a cloud provider? While the chance to offload the day-to-day operational workload to someone else is a big draw of cloud computing, it may be what gives enterprises the most cause for concern. Someone else is now handling your data, so will your data be secure? You’d expect a cloud provider’s huge team of security experts to have more knowledge, training, and visibility than your one- or two-person IT security staff, but are they intimately familiar with every aspect of your business and every nuance of your applications? Most security approaches depend on the assumption that you’re actually in control of the data you’re protecting: does a cloud-hosted solution fly in the face of this?
Are your data and your infrastructure going to be held hostage? Once you’re “in” with a cloud provider, can you then get out again? Should you split your infrastructure across multiple cloud providers to ensure you can’t be locked in or have the screws turned on you with price increases? If you do split the infrastructure, or if you’re using something like Citrix’s Azure services, where two companies provide separate layers of the infrastructure, whose throat do you choke when something goes wrong?
The Personal Touch
And this raises a final point. On-premises IT fails at least as much, and possibly more, as cloud IT does. So why do users and management generally accept occasional failures with on-premises IT, but lose their collective heads when there is a small outage on AWS or Azure?
It’s not just a case of “hoping the big kids fail,” although there is possibly an element of that at work. If there’s a planned outage for an on-premises solution, that’s just in-house IT doing what in-house IT does. But if a major cloud host goes down, even for a short period of time, it’s big news, and it invariably generates a lot of coverage. Cloud providers sell themselves on their reliability, and when that fails, it gets publicized.
There’s also an element of the personal touch to in-house IT that is sadly lacking, currently, in a cloud-based solution. A good IT department keeps everyone—users, management, third parties—in the loop. Even in those large enterprises for which regular updates need to be given, a dedicated IT department can give a sincere and business-focused response. However, cloud providers tend to offer canned responses that have been sanitized with the appropriate legal precautions. Their customers feel like they’re being kept in the dark and jump onto social media platforms to vent their frustration, and before long the outage is on the front page of every tech website.
Cloud is great for backups and disaster recovery. Standing up disaster recovery sites is so much easier. It’s great for scalability in environments that need to increase capacity rapidly at particular times (assuming you don’t pay too much for it). It’s also great for moving lower-impact services like email, instant messaging, and monitoring into an environment where you don’t have to worry about operational overheads and hassles. And it certainly has a big part to play in the transformation of modern IT environments that we see on the horizon. Windows 10 brings a new wave of rapid change, and it represents possibly the first iteration of the cloudy OS of the future.
That future is unlikely to be public cloud only. Hybrid is likely to be the way forward for the foreseeable future, although certain industries may find it easier to go all-in on cloud than others. Your deployment decisions should be evaluated carefully in light of your unique priorities; there is no one-size-fits-all answer. Ultimately, most businesses will want to take advantage of both on-premises and cloud-based solutions because they address different important issues. Has any technology since the mainframe truly totally replaced its predecessors? Some may well have supplemented existing tech, but none have yet fully superseded it. Cloud looks to be no different.