Content Copyright © 2017 Bloor. All Rights Reserved.
Also posted on: IT Infrastructure
If your CIO comes to you on Monday and says you need to replace your old data centre, you should tell him or her that the primary strategy should be to outsource as much of the data centre environment as possible. Therefore, he or she needs to make a very strong case for building and managing a new data centre, as opposed to having to make a very strong case to outsource. There may well be compelling reasons for having your own facility, but I believe we have reached that tipping point where building and managing your own data centre facility rarely makes sense.
The hyper-scale cloud operators have changed the economics of building and running data centres. They have reduced the costs of managing servers by a factor of 10, 50 or a 100 depending on which stories you read. They have perfected the art of squeezing the last drop of cost savings from energy utilisation through innovative cooling technologies, attention to detail in layout and customisation of servers that strips out unnecessary power utilisation. They are building or leasing data centres at a huge rate. According to Research and Markets the Mega-scale data centre market is forecast to grow from $869.7 Million in 2016 to $359.7 Billion in 2023.
The growth in cloud deployment and the advances made by the likes of AWS, Microsoft Azure and Google have also had a galvanising effect on the co-location data centre sector. Their data centres have become far more efficient and operators like Equinix, Interxion, NTT et al have built interconnectivity and direct public cloud access facilities that deliver hybrid-cloud securely, quickly and without the challenges of cost and latency of using the public internet.
Technology is changing at such a rapid pace that data centres need to be built with flexibility and agility in mind. Data centres are actually 15-20 year investments, therefore there is a real need to factor in modularity. Form factors, power requirements, cooling and networking considerations will all be impacted, not only by new technologies in themselves, but also by the way DevOps, software defined infrastructure, Big Data, analytics and the Internet of Things are changing. Do you as a CIO want to be responsible for the challenges of building and maintaining your own data centres to meet these challenges when skills are scarce and frankly, you won’t be able to get anywhere near the operating price points of the cloud and service providers?
But what about hybrid-cloud? For sure, analysts like Gartner are pointing to Enterprises majoring on hybrid-cloud in the next few years. But hybrid-cloud does not necessarily equate to a combination of on-premise and off-premise. The private part of a hybrid-cloud can run quite happily in a co-location or outsourced data centre. Given the growth of off-net direct connects to the public cloud, co-location provided interconnectivity platforms and metropolitan area networks, I would suggest that the in-house, Enterprise data centre is the last place to run a hybrid cloud.
Security and data privacy has often been a big factor in enterprise decisions to keep data centres in-house. The failure of the Safe Harbor provisions has prompted a surge in the building of local data centres to help deal with data sovereignty concerns and legislation. If security is a big issue, then you can’t get more secure physically than nuclear bunkers or data centres built into mountains that are proliferating across Europe. Combine that with managed infrastructure and software security services that are the equal of, and often superior to, in-house security technology and processes, there seems little reason to discount outsourcing on these issues, unless of course you are GCHQ.
Then of course, there is the mainframe. I won’t reprise the reasons for the refusal of the old dears to fade away and die. They are often cited as a reason for maintaining your own data centres. I can’t talk for IBM, but I know that Unisys mainframes now come in industry standard 19″ cabinets and use absolutely standard disk sub-systems. So, they are hardly a physical problem. I now understand that, much to the horror of many in the IT world, whole mainframe applications are being containerised which will radically reduce operations and maintenance headaches.
So, there are no generic, general reasons why you can’t ditch your ageing data centre. As an Enterprise, there are certainly very strong reasons why you shouldn’t contemplate building and owning your own new data centre. If your CIO knocks on your door and suggests you did need to build and run a new one, then make sure you set the bar very high for him or her to get approval.
This post first appeared on the old Cassini Reviews website.