Every business is looking to save money. Newer servers with faster optical data center connections within and between data centers are a great way to do so while punching the “green” ticket, as numerous enterprises are demonstrating.
The path to “green” IT is often confusing, layered in PUE and CO2 emissions. Ignore all the hype and focus on one basic statistic: the power bill. Servers and network gear use electricity. The more servers running all the time, the hotter the data center and that leads to running the cooling systems more, running up the power bill higher. If you have a lot of older equipment installed — 3 years or more — that’s more electricity on your power bill which translates to higher operational expense per month.
Fewer servers, less heat and power, cooler data center, less overall power, lower operational expense. What’s not to like?
With the exception of “Big Data” shops and major Internet players, everyone from General Motors (GM) to the U.S. government has too many servers on line. Many Fortune 500 companies are in the process of migrating data center operations from a number of distributed sites and all too many “silos” of different applications and operating systems to a consolidation of resources, with fewer physical servers, fewer data centers, and fewer people to maintain everything.
Step one is to buy the newest, fastest equipment possible. Intel (which, of course, has a vested interest in seeing new equipment being sold), likes to boast that replacing an older server with the latest technology using 30 percent less energy will save up to $480 on a four year life cycle.
On top of the newest hardware goes virtualization, so fewer servers are needed to do the same work. However, fewer servers also translates to more processes per server and the same number of people/processes needs to access them which leads us back to faster optical connections. You can do a lot of clever things with load-balancing hardware and network mapping to distribute load, perhaps segmenting network links to prevent bottlenecks.
Fewer servers means there are fewer ways to be clever with routing and cable and all the other tricks of the trade. Add in virtualization and you need the highest speed optical connections available, because there are fewer physical network connections. The good news is once you start consolidating network connections with faster optical gear, out goes a lot of the (typically older) copper-based gear and all of its space and power needs, leading to a ripple effect where less cooling is required.
Finally, server consolidation leads to less physical footprint — real estate — and to fewer data centers. More concentration of resources means both faster LAN and WAN links, as a combination of local virtualization within the data center and distributed, on-demand “cloud” services within the enterprise spread across multiple geographic locations requires more optical Ethernet to keep everything flowing efficiently and smoothly between the data center and thousands of end users.
General Motors work in consolidating its data centers and IT organization is leading edge, all the more remarkable due to the company’s bankruptcy and recovery. The company is moving away from 23 data centers distributed around the globe to two (2) based in Michigan. Payback on its brand-new, U.S. LEED Gold-certified $150 million Warren Enterprise Data Center complex is expected to occur within three years due to power savings.
But it’s not just power that’s being revamped. GM has outsourced its IT infrastructure since 1996, leasing space and contracting employees. Consolidation into two data centers means GM no longer has to worry about WAN connectivity between 23 distributed data centers, but can focus on inter-data-center-connectivity between its two mega sites – and secure, high-speed access to these two mega sites from all its locations. GM currently has 2,500 virtual servers running in its Warren data center, with a pair of high-speed “backbone” rooms supporting all connectivity to the rest of the world, including the company’s VoIP phone network, telepresence, and connections to GM’s partners.
Federal Express (FedEx) is another company hopping onto the virtualization express. At VMWorld 2013, Chief Engineer of Architecture Chris Greer said the company has about 22,000 virtual machines in operation. The company is pushing to virtualize as much as it can, but the move has trigged a purchase in more storage which, “puts more pressure on the network to move faster,” Greer said.
It should be no surprise that FedEx’s Colorado Springs data center, opened in 2011, is LEED gold certified, with around 70 percent of the servers in the facility being virtualized. Along the way to a more efficient IT operation, FedEx got rid of 200 different applications to manage addresses and numerous other legacy applications that didn’t “fit” into its consolidation plan. Individual departments may not have liked being pushed towards consolidation, but the company will reap ongoing savings for years to come as it upgrades its in-house server architecture and examines how to leverage a hybrid-cloud infrastructure (which will, of course, require matching speedy optical connections) with Amazon.