Data Center Consolidation and Efficient IT: Strategies to streamline and reduce wastage in your server estate
Earlier this summer (July 2011) the US federal government announced plans to shut down 40 percent of its data centers over the next four years. With more than 2000 data centers, the Obama Administration realized that data center consolidation was part of a broader strategy to become more IT efficient and make substantial financial savings (predictions run into billions of dollars a year) through managing information technology.
Many large organizations today find themselves in the same boat, albeit on a smaller scale, where they are choosing to make economies of scale by running fewer data centers and reducing their overheads.
Over the past 20 years, the commoditization of computing in the datacenters has seen physical servers shrink in size- initially they took up the space of a room, which reduced to the size of a fridge to today’s ‘pizza box’. Added to this, for the past 40 years processing capacity has been doubling approximately every two years and this trend is expected to continue until at least 2020.
These factors of power density and increased processing capacity have contributed to the higher power and cooling demands in data centers. In fact the demand can become so focused on small areas jammed with compute resource that spare space becomes ‘dead’ as there’s no more available power. Data center utilization can stall and the irony is that it becomes cheaper to build a new data center to accommodate energy demands than to retrofit an old one to ‘unlock’ the space.
The route to virtualization and server rationalization
Server commoditization has led to a proliferation of standalone servers. The main reason why companies provisioned so many individual servers was that performance and capability could be compromised by running multiple applications on a single server. It also made maintenance easier since servers were relatively cheap when compared to monolithic mainframes.
As companies have grown either organically or through mergers and acquisitions so have their server estates. This server sprawl has led to inefficiencies in the use of data center resources, hence many companies have aimed to mitigate it through virtualization.
It is worth noting where costs can creep through use of virtualization. The standard specification of a server purchased to host virtual machines (VMs) is usually much higher than a standalone server. Hosting multiple virtualized servers requires more CPUs and significantly more memory both of which contribute to higher power requirements and heat output.
If you virtualize 100 percent of your physical servers today then you will be running more operating system instances than you started with. The physical and environmental footprint has changed but you still need to monitor, patch, secure, backup and license the virtualized servers and now the host servers as well. Requirements such as high availability further increase infrastructure costs. Yet the publicity of energy and space savings has created a mindset that virtualized servers are cheap. Consolidation through virtualization therefore is only a partial answer to becoming more IT efficient.
Increasing your IT efficiency
Server rationalization should be part of a virtualization project and a business-as-usual process within a virtualized environment. New tools designed for IT efficiency help identify what is useful and when. This knowledge makes it easy to establish an on-going process to reclaim unused resources and avoid unnecessary expenditure.
Traditional tools designed for systems and operations management struggle to deliver reports on the useful work a server is doing. Often a report of how busy a CPU is over time is used to represent how much “usage” took place. Most servers are loaded with standard software that has to be on every corporate server – antivirus software, systems management, backup and event monitoring – so each server reports a specific amount of usage but they fail to reveal when or if that usage provided business value.
Although virtualization has clear benefits, including a reduction in energy consumption and floor space, to get the most out of virtualization you should be prepared to invest in tools which monitor efficiency, specifically the amount of useful work your IT assets are doing.
Software licensing inefficiency
The overprovisioning of servers in data centers is rife and has led to an inefficient use of hardware. Additionally, there’s software license waste through unused and partially used applications on organizations’ servers. Anecdotal evidence echoes analyst opinions that the total cost of licensing and running servers could be as much as 80-90% of total software spend.
Businesses buy billions of dollars of enterprise applications such as ERP, CRM and finance systems to name a few. This includes an enormous outlay on the cost of the infrastructure software that enables virtualization. Many of these applications have different and complex licensing rules. It is important to understand actual usage and whether, for example, a premium license could be replaced by a standard one or even cancelled if it’s not used.
There are many different models of licensing for example, per CPU, per RAM, per user. Sometimes this complexity is so complicated that organizations tend to overprovision as they can’t understand what they need to do. So no one truly understands how to be truly efficient or feels comfortable that they have exactly the right number of licenses and getting true business value.
Saving Your Server Estate
Your data center energy demands aren’t going to go away any time soon. In fact, they are going to continue to get more complex as inefficiencies abound in physical and virtual environments, resource usage, and software licenses. If you factor in the current trend of moving some of your business services to the cloud, you have an even more daunting landmine of potential misuse. It is time to take control of your server estate by implementing software that can help you efficiently consolidate your data centers and eliminate energy waste. The longer you wait, the more energy, time and money you will throw away.
Written by by Andy Hawkins, Product Manager at 1E.