When it comes to hosting your IT infrastructure, the number of available options has increased significantly in the last decade, particularly with the rise of the big four public cloud platforms – Amazon Web Services, Microsoft Azure, Google Cloud Platform, and AliCloud. Many large enterprises have announced they have gone or are going ‘all in’ on public cloud, with 68% of enterprise IT teams already using public cloud in some way shape or form according to this NetEnrich report. https://www.youtube.com/embed/h6BzHP-oIKsYet datacenter providers cannot build capacity quickly enough to meet the relentless demand for co-location space, which is showing no sign of slowing any time soon, despite the unstoppable rise of the public cloud. According to this Wall Street Journal article, the number of datacentres is expected to jump from 7500 to 9100 this year and then continue to 10,000 next year. While these numbers are in no doubt boosted by public cloud operators and their datacenter builds, there is no question that the market for managed co-location space is buoyant.
So, which is right for you? The answer may not be one or the other, but let’s explore some of the advantages and disadvantages of each.
Cost Reduction: The biggest driver for public cloud adoption is cost reduction. The big four operate at a huge scale, and with that scale comes the ability to procure hardware, software, and data center space at a lower cost than most businesses can, and that low cost can be passed on to their customers. The fact that you can consume resources as required and pay for what you use means you can align your costs to your business requirements – there is no need to overprovision resources and you no longer have spikes in cost when a new SAN or server enclosure is required – all of your capital expenditure is turned into operational expenditure. Additionally, public cloud vendors are all fighting for market share, so there are many funding schemes available which could be factored into your business case to reduce your costs further.
Availability of Services: Once you have an account on a public cloud, the myriad of services is available to your business. There may be technologies that you can try out on a pay per use basis, which would have otherwise been beyond your budget or technical capabilities. Examples include Machine Learning and Artificial Intelligence, voice to text, etc.
International Coverage: Deploying in a new country or region is as simple as selecting your chosen location from a drop-down menu – no need to worry about sourcing datacentre space and equipment in a new territory, or paying for services in a foreign currency, you can consume services across the globe from the same payer account, making international expansion a breeze.
Complex Billing: While there are cost savings to be made by migrating to a public cloud, you’ll need to stay on top of billing. Public clouds meter and charge for things that you have probably never needed to measure in your datacentre, such as egress bandwidth accurately, storage Put and Get requests and API calls. If your application is not architected to be cloud-native, these costs can soon mount up – you’ll need to keep a close eye on those bills to ensure that you are maintaining your cloud infrastructure optimized. The vendors provide cloud-native tools for this purpose, but there are third-party tools that do it better, and a whole raft of service providers who can help in this area.
Difficult to govern/control usage: Public cloud makes a wealth of services easily accessible to your developers. Shadow IT is a massive problem in larger organizations, where teams have gone outside of the corporate governance structure and procured public cloud accounts on a credit card, claiming the fees on expenses. The IT team can be completely unaware of this. Cloud service providers will collaborate with enterprise IT teams to track down shadow IT so that these environments can be brought back into the governance structure, but this can be a huge undertaking. A well-designed public cloud environment should give developers the freedom they need to innovate while allowing corporate IT to enforce whatever governance is required by the organization to adhere to corporate security policies and compliance standards.
Vendor Lock-In: Welcome to the Hotel California. While it’s super easy to migrate data and server images into the Public Cloud, the service providers make it a lot easier to get into the cloud than to get out of it again. All the funding programs and free tooling make migrating to the cloud relatively painless. But you will be hit by expensive network egress charges if you want to migrate your data away again. Similarly, if you start to consume native cloud Platform as a Service (PaaS) or serverless functions, your application code will need to be re-written to leverage these services, which in many cases have proprietary APIs – so if you want to move those apps elsewhere – guess what – you’ll need to re-write the code again.
Fixed Costs: When you rent rack space in a datacenter and install your equipment in them, you know exactly what your IT infrastructure is going to cost you from month to month. There will be no nasty surprises on your bill. Sure, you may have some small metered line items for bandwidth or power over-usage. However, usually these can be negotiated into your contract at a fixed cost. Some business owners prefer this level of control over their expenditure.
More extensive Technology Choices: It goes without saying that when purchasing your hardware and software and installing it in a datacentre, you can take your pick in terms of vendor offerings. Need a specific graphics accelerator or hardware security module? No problem, go right ahead and buy what you need!
Security & Data Sovereignty: Hosting your IT Infrastructure in a data center gives you the security you can see. Some people like to be able to physically see and touch their own IT infrastructure and experience the physical safety of a data center facility. I have toured many a data center with a curious customer who just wanted to see what the physical security controls were like, from the perimeter fencing, vehicular access, CCTV system, personnel access control, right down to biometrics on the rack door – some people just want to see all of that in order to be comfortable with the level of security in their facility.
And when it comes to data sovereignty, not only can you guarantee the mailing address where your data will be held, you can even point to the rack location and storage array. I have had customers fail data audits where physically dedicated equipment was in a shared rack – some auditors are old school and require a rack with access logging facilities to secure certain categories of data.
Lumpy Upgrade Costs: There’s no question that upgrade costs in the datacenter are lumpier than in the public cloud. Run out of rack space? You’ll need to take another rack. Filled up that blade server enclosure? You’ll need a new enclosure, switch gear, cabling, etc. Filled up that Storage Area Network? Another SAN or shelf is going to set you back a pretty penny.
High Availability is Expensive: When running your datacenter, you essentially need to buy 2 of everything to achieve high availability. Then you need to split your infrastructure between locations and ensure that you have enough network connectivity between sites to handle data replication.
Remote hands: if you’re co-locating in someone else’s datacenter, you’ll need to ensure that they provide an excellent, reliable remote hands service. Sure, most things can be managed remotely nowadays via onboard management tools in hardware, remotely accessible power strips for hard reboots, and console server access for networking devices that are not publicly accessible — assuming, of course, that all the above was in budget.
But even with a belt and braces approach to remote management, there may be times when you cannot get on to the device you need to, and there is no substitute for someone being able to go down to the rack and take a look.
So, as you can see, there are many factors as to which IT Infrastructure option is the right one for your business. And, you’ll likely end up with a mixture of both public cloud and on-premise or co-located datacenter. Some workloads are perfect for public cloud, and some, for the reasons highlighted above, are happier running in a physical datacenter. Despite those going all-in on the public cloud, most businesses will end up with a hybrid setup, and there’s nothing wrong with that.
It all comes down to using the right tool for the right job and aligning with your business requirements. There is also a whole heap of new tools coming on to the market to enable you to manage your hybrid environment from a single pane of glass, so splitting your infrastructure between the datacenter and public cloud does not need to cause you a management headache.