I sit down with data center managers and other IT professionals all the time and one of the things we start undoubtedly talking about is ‘the future’ of IT. In fact everyone in IT loves to talk about the future of IT since there is simply SO MUCH transformation going on these days. Now the vast majority of us have been in IT for 25 years or so and I can genuinely say that I have never seen such a breakneck pace of change for all aspects of the data center and IT infrastructure in my generation.
While many of those ‘future’ discussions start with the state of the industry very generically, these discussions quickly zero in on the state of THEIR IT FUTURE. While the state of the industry is curious and exciting to most, it is how that change will affect their own world that is of primary importance. And all of that change is being viewed with a fresh new set of ‘business colored glasses’. So while it’s still fun to talk about Ethernet Port speeds and new capabilities of Windows Server (oh boy) a higher priority discussion revolves around the costs to deliver IT “products” or as sometimes referred to as IT services. (e.g. a user’s email capability). And with worldwide spending on IT in 2014 of more than $3.7 Trillion dollars, you can bet there are a lot of people involved and a lot of financial analysts asking some pretty tough questions.
So that brings me to the topic of “computing styles”. 15 years ago, we didn’t really have ‘styles’ of computing. Prior to that time, everyone had a data center if they wanted to do computing. Any organization that relied on IT built big data centers. They did so on land that they bought, in buildings that they constructed, and they filled these structures with gear that they purchased. This was pretty standard fare for most of the 80′s and 90′s.
Then the we watched in horror throughout the troubling years of 2000-2010. We took a double-whammy in that decade; 1) The DOT-COM melt-down in 2000 and 2) The ECONOMY melt-down of 2008. What these did is provide extreme motivation to develop and promote alternative and much more cost-effective business models for computing. Putting aside the related changes in end-users business needs themselves for a moment, these Enterprises now had a handful of choices or ‘styles’ for computing.
- In-House – characterized by brick and mortar construction, large upfront costs, and complete responsibility for all aspects of its operation and usage. In many circles, this in-house capability is being re-tooled to behave as a ‘private cloud’.
- Co-Location – much like an in-house Data Center, but the cement and MEP gear is provided, essentially as a service. Enterprises ‘rent’ space that comes with power and cooling and network connectivity.
- Cloud – the hyperscale data centers with tens or hundreds of thousands of servers, running specialized software for self-service and quick provisioning which provide the ability to purchase computing by the transaction, eliminating all other operational concerns. Usually “Cloud” is the shorthand for “Public Cloud”.
- Modular – think of a small in-house data center that can be transported to site in 500kW increments, stood-up in just weeks rather than years, and can be tuned for specific cookie-cutter business needs without impacting other portions of the IT structure.
Most importantly, IT professionals that get the big business picture realize that their own infrastructure WILL NOT be changed entirely overnight. In fact all four of these styles will exist in various combinations across their own span of control for years to come. If asked, most IT professionals will say something like “I am going to the Cloud” but what they really mean is that their strategy is to utilize a growing percentage of their computing needs via transactional cloud-oriented computing, focusing more on transactions and transactional costs rather than floor tiles or servers. It’s not black and white change, it’s about the changing mix of these styles over time. And the forward-looking decisions are centered on the value of using each type.
Now the beauty of computing styles is that there are a number of startups and public companies alike that are dealing with the TRANSITIONS and MIGRATIONS between these styles and ways to leverage each style for it’s core value. Say for instance a company already owns 100,000 square feet of data center space that is less than 3 years old. But say they need to augment their capacity with more transactions twice a year. Why not just add public cloud transactions to their own transaction handling capability? It turns out you can! Startups and open source projects make the migration of workloads as simple as clicking a mouse, or in some cases completely automated based upon demand. Workloads can be added or moved from across in-house servers as easily as between private and public clouds… and back again!
Or consider the concept of Disaster Recover planning in the context of computing styles. Everybody should have a rock solid DR plan , but justifying that plan and it’s costs usually revolves around the costs to stand-up a ton of warm standby gear. What if you could use one of the other styles, say the cloud for instance, as the ‘flick of a switch’ style DR site by moving workloads to it when the main site is being impacted? Once again, you can!
Today’s challenge for data center and IT professionals is really is about properly setting a value to each of the IT workloads that they commit to provide and doing it with the level of resiliency needed for the business. It’s ALL about the business and the number of options available to run the business has increased!
Workload management is the key to business efficiency, and the most critical facets of workload management is identifying the cost to deliver work. With $3.7Trillion in spending on the line, do you know what your costs are to process each unit of your work, and do you know if it would be more or less per unit of work to shift from one style to another? Do you have transition plans and technologies to move workloads dynamically to leverage each style based upon demand, status or time? Can you handle demand peaks effectively or have you over provisioned resources which sit idle for most of their life? Do you have a DR plan that makes your IT contingency strategy both low-risk and low cost? These are the tough questions on the table today.
Anyone want go back to simpler times when we all talked about raw networking switch throughput or the capacity of the latest spinning media disk drives?