Welcome
...to our weekly Great Debate. Are you ready?
David Chernicoff
Overhaul
Incremental
Adrian Kingsley-Hughes
The moderator has delivered a final verdict.
David Chernicoff: It’s just too easy to take a piecemeal approach to the upgrade of your datacenters rather than taking big bites out of the problem and moving forward at a faster pace. Business change simply happens too fast for you to lock your company into a glacial upgrade path for your datacenters. As new business model evolve and new technologies appear, that’s the time to consider overhauling your datacenter as a business opportunity.
If your budgets can handle it, taking advantage of technologies such as modular or containerized datacenters means that you can make large scale updates to improve the performance and efficiencies of your datacenter while initially running in parallel with existing legacy systems, retire the legacy equipment as the new systems fully integrate, and take advantage of all the benefits such a major upgrade can bring.
A few large scale upgrades will get you a more efficient and powerful datacenter in time to take advantage of the potential business advantage that technology can bring.
Adrian Kingsley-Hughes: During tough economic times it is natural for organizations to look for ways to reduce costs, and one way to do this is to review existing IT operational procedures in order to identify possible efficiency improvements. But what approach to take?
While some may advocate a comprehensive, holistic approach that optimizes all attributes at once, I believe that a more conservative approach -- where optimizations are handled in a more controlled way -- is far better.
The problem with the holistic approach is not only that it represents considerable disruption -- not to mention possible downtime -- but it is a solution that comes attached to an eye-wateringly large price tag.
In an ideal world, it would be great to be able to throw everything out and starting from scratch, but most of us live in the real world.
...to our weekly Great Debate. Are you ready?
Set to go
I am for Overhaul
Let the games begin
I am for Incremental
To some, the term "overhaul" means replacing servers or reworking cables. To others, the
term implies more significant changes, such as energy-efficient cooling solutions and modular data centers. To ensure we have a clear understanding of how our debaters characterize an overhaul, what is your definition of a data center overhaul?
I define overhaul as making a significant change to any part of the data center infrastructure or IT component. This means replacing, consolidating, adding, or significantly changing the capabilities of the data center.
I am for Overhaul
An overhaul to me is any action carried out that is designed to improve performance, increase reliability, or reduce costs. These are in my mind the three tenets of data center overhaul, and everything should come down to one or more of these.
With that in mind, overhaul becomes a very broad term and can range from small changes
such as reworking cabling, to more significant actions such as installing faster drives,
switching to SSDs, or adding new processors, all the way to starting from a totally clean slate at a new location with all-new hardware.
I am for Incremental
Likewise, individuals can interpret the term "incremental" differently. What is your definition of an incremental data center change?
I see incremental change as maintaining standardized replacement cycles for infrastructure or IT hardware components. Changing at perceived end-of-life or on a set schedule as an ongoing process, something like the apocryphal stories of painting of the Golden Gate Bridge, a never ending process done at a slow and steady rate.
I am for Overhaul
Small changes over time, concentrating specifically on getting more out of the hardware infrastructure that's already installed. Again, these will revolve around the three tenets of improving performance, increasing reliability, or reducing costs.
I'm not for one moment suggesting that a complete overhaul of all the major hardware isn't necessary from time to time, but incremental improvements help keep things running at peak efficiency the rest of the time.
I am for Incremental
The benefits of completely reinventing one's data center can be significant, but the potential pitfalls are also very real. David, what are the three or four top benefits of a total data center overhaul? Adrian, on the flipside, what are the three or four top hazards?
First, I don’t think that you can do a “total” data center overhaul in the sense of it being a single sweeping redo. I think the advantage of the overhaul approach is to examine the business processes operating ion the data center and overhaul the equipment that is used be each process. Or, as new business processes and model are adopted, to add hardware and infrastructure capabilities that will deliver the best performance and ROI for that process, while also being able to support other activities in the data center.
This goal-based approach allows you to make sure that the newer processes are running on the most optimized environment, while also adding capacity that will allow, over time, the migration of older business processes to new equipment that is more effective and efficient.
If I was to attempt to define the benefits in three or four items I think it would be this:
1. Measureable improvements in energy efficiency
2. Measureable improvements in application or process performance
3. Faster delivery of capabilities to users
4. Shorter period to achieve measureable ROI
I am for Overhaul
There are huge potential pitfalls revolving around undertaking complete overhauls. Here are
just a few:
Cost – If you work for a company that's happy cutting blank checks then complete overhauls are awesome, but in the real world that rarely happens, in a tight economy companies have to make the most of what they already have.
Time – Overhauling a data center is nothing like replacing a PC, and it takes considerable time to plan and execute.
The unexpected – The more new hardware and software you are having to deal with, the greater the scope for things to go wrong, and this invariably leads to downtime, which, in turn, hits the bottom line.
I am for Incremental
A slow-and-steady approach to data center modernization may limit disruptions, but it can
also stifle innovation. Adrian, what are the three or four top benefits of incremental data
center optimization? David, on the other hand, what are the three or four dangers?
The biggest danger is getting trapped in that endless cycle of being behind the technology curve; by the time you’ve worked your way through your entire datacenter it’s time to start over. Second, many of the optimizations, such as those that affect cooling or power delivery efficiency, don’t see real benefit until they are complete, or simply can’t be done on a small scale effectively. And, of course, you will find that you need to constantly be justifying ongoing capital investment without being able to show ROI improvements in the short term; simply projections on benefits when the cycle is complete (and about to start all over again).
I am for Overhaul
While a balance needs to be struck between complete overhaul and incremental change, a slow-and-steady approach has a number of advantages.
Keeping up with the times – Better processors, faster drives, and denser RAM is a fact of life, and incremental upgrades allow you to take advantage of better hardware in a more timely – and cost-effective – way.
Cost – Incremental is always cheaper than a complete overhaul.
Focused spending – Incremental upgrades and improvements allow IT managers to focus spending more effectively depending on need.
Less scope for problems – While small changes can result in problems, which in turn lead to disruption, the scope is far more limited.
I am for Incremental
Data center technology is constantly evolving, but the rate of change seems to be increasing as does the variety of new server, storage, and networking tech. What new technologies are pushing companies to completely rework their data centers and do they strengthen the argument for a complete data center overhaul?
The number one differentiator between the past and present is the data center facilities issue; building design, power, and cooling were formerly the domain of dedicated facilities groups. But the huge increase in the cost of power and the recognition that energy consumption and utilization were factors of the IT workload of the data center, have brought IT and facilities together. With technologies such as cooling and power distribution suddenly moving to center stage, most data centers find themselves well behind the curve and need to address these issues on a facility –wide scale. Plus, of course, the greater public awareness of "Green IT” means that being a good corporate citizen now requires paying heed to the greenness of your facility.
I am for Overhaul
The factors are today pretty much as they were a decade ago:
Better processors
Faster drives
Increased computing density/footprint reduction
Virtualization
Despite being an advocate of incremental change, I'm well aware of the fact that a time will
come when almost every company will need to consider a total overhaul. While a relatively
modern data center will be able to take advantage of new technology without requiring a
complete overhaul, if your data center has that Flintstone feel to it, then an overhaul might be the only thing that saves it!
I am for Incremental
Does the proliferation of cloud services such as Infrastructure-as-a-Service (IaaS), Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS), cloud storage, and Software-Defined-Networks (SDN) favor a the overhaul or incremental approach?
I see this as two different questions. If you are moving services of any type to cloud providers, the most significant impact on the data center is that the money being spent of cloud-provided services is no longer available for data center internal use. If a business has made the cloud approach paramount, then the changes to the data center will need to reflect the new service-provider focused environment. The changes necessary for that will be individual to each business based on what services and how much of the IT infrastructure is moved to cloud providers.
Software defined networking (SDN) is another story. In most data centers the decision to implement a complete SDN environment will require a major overhaul of the networking infrastructure to support the demands of SDN. Few older data centers have the underlying networking infrastructure necessary to support a full scale transition.
I am for Overhaul
The more computing work you can offload on some other company, the more you shift the pressures onto that company, freeing you to take a more leisurely approach to change. This is exactly the reason why companies have embraced the cloud. This way companies can buy exactly what they want, rather than being sold something that they may never need.
I am for Incremental
When organizations calculate the ROI of their data center optimization projects, they often begin by calculating the direct costs (fewer servers, software licenses, personnel, and facilities). Thinking only about the direct costs, is a complete overhaul or incremental update the better optimization strategy--in the near-term and long-term?
Overhaul projects, by their very nature, having more significant upfront costs. But they also have more rapidly identifiable ROI, in terms of their impact on overall IT costs and potentially on identifiable improvements in business workflow.
I am for Overhaul
Again, assuming that the existing data center wasn't around at the time of the dinosaurs, then incremental updates are by far the better optimization strategy because they allow the IT managers to focus spending on what they need more of to get the job done, whether that be more compute power, more storage, more RAM, or whatever.
I am for Incremental
Calculating the indirect cost savings of a data center optimization project, such as energy savings, efficiency gains, and increased scalability, can be more challenging than determining the direct costs. Nevertheless, doing so is still a critical step in measuring the project's ROI. Thinking only about the indirect costs, is an overhaul or incremental update the better strategy--in the near-term and long-term?
The way this question is asked touches directly on the benefits of the overhaul approach. With this approach there are very often direct, quantifiable savings on energy expenditures, which gives you greater efficiency measurement and on the availability of additional IT workload capability, allowing applications to scale or new processes to be added without additional investment. This provides direct measurement of savings, often in both CAPEX and OPEX spending.
I am for Overhaul
Think about it this way – which option gives you the best change to see whether your improvements are having a marked – and positive – effect on indirect costs? Small changes
allow a company to keep an eye on those indirect costs, and even help steer them in a
particular direction. For example, you can incrementally change a specific subset of your hardware and compare that to the baseline.
On the other hand a complete overhaul can be a total leap of faith, and many times doesn't
allow for a "Plan B."
I am for Incremental
David, what are the characteristics of a company that's well suited to optimize their data
center through an overhaul? Adrian, what are the characteristics of a company that's well
suited to optimize their data center through incremental changes?
First, they need to be willing make the investment in technology. Second, there is a compelling business need to have a best-in-class solution to the specific problem they are looking to solve or business process they are looking to add. Third, their existing infrastructure can’t adequately support their planned growth or expansion related to the business issues. And forth, they want to be on the leading edge and reap the business advantage that can provide.
I am for Overhaul
The companies that are best suited to optimize their data center through incremental changes are, rather ironically, those who are already taking this approach. Incremental betterment of their hardware and software means that they've kept up with change over time, and have not allowed their infrastructure to get too old.
Other companies suited to optimization through incremental changes are those who have
recently overhauled. Their overhaul will – or at leas should – have left them with hardware
that they can build upon and improve.
I am for Incremental
For our last question, provide at least one example of a company that's either completely
overhauled its data center or optimized it through incremental changes and is seeing the
benefits.
Almost all the examples that are public knowledge (i.e., that companies are publicizing these changes) are technology businesses or companies that have built entire new data centers. For example, when eBay needed additional capacity in their existing data centers, they simply dropped in data center-in-a- box solutions to upgrade their capability while the existing facilities were then upgraded to their new standards. That can be considered the extreme example of the overhaul approach; the next step up would be to build completely new facilities.
Other Internet-based businesses and technology businesses are quick to announce the major upgrades and changes that they are making to their data centers, but it is very rare to hear a public announcement from businesses in other than the technology field that they have overhauled their data center processes.
Most businesses consider these types of choices to be confidential business information because it defines the way they plan to do business.
I am for Overhaul
One company where incremental changes have been paramount is Facebook, as demonstrated by its Open Compute Project initiative. Anyone involved in optimizing or improving a data center needs to take a look at this because it is a mine of information, and it should convince you that incremental changes are the way forward.
These shared custom data center designs clearly show that incremental changes are at the
heart of how Facebook operates, making sure that that there will always be enough server
capacity for one more LOLCAT picture!
I am for Incremental
You'll agree that both David and Adrian did a great job defending their positions. Come back Wednesday for their closing statements and Thursday for my final verdict.
David Chernicoff
It’s clear that the choice, like many in IT, needs to be tailored to the specific needs of the business. Comments here and those I’ve received in email certainly make the case that incremental is often considered the safer choice. But safe doesn’t necessarily mean best. When you’re spending resources in your data center you need to consider what the best overall return on investment is. Ruling out options because there is a level of risk or discomfort is not going to lead to the best business decisions.
Like the old saw ‘no one ever got fired for buying XXX” you can play things safe, but that adage became untrue when businesses realized that maintaining the status quo might not the best model for long term business success. As such, the option to overhaul large components of your datacenter infrastructure should remain on the table, regardless of the desire to focus on safe, low risk choices.
Adrian Kingsley-Hughes
If your server room is not stuck in the Flintstone days, then taking an incremental approach to upgrades could allow your company to save time and money. It will reduce the risk of downtime which is inevitably associated with any major overhaul.
Also, as long as your server room isn't creaking under the weight of its current workload, taking an incremental view to upgrading allows money to be spent on customizing the hardware to make it better suited to what is being asked of it.
And, if you're still not convinced about taking the incremental approach, consider the costs -- both direct and indirect -- of a total overhaul, and imagine running that past the CFO.
All of a sudden, incremental sounds great, doesn't it,
Bill Detwiler
We've had close votes before, but this was the first time in Great Debate history that the audience was evenly split. And perhaps, we shouldn't be surprised. Deciding when and how your company should upgrade its tech, whether in the data center or on the desktop, is a choice driven by the specific needs of your organization. There's rarely a one-size-fits-all solution.
Posted by Bill Detwiler