The Energy Scoop on Unified Fabrics
It may seem overly simple but generally speaking consolidation can be a great way to reduce total energy use in your data center over a given period. The implications of implementing unified fabrics from our Cisco Nexus line is no exception. However, the savings you achieve might not be as simple as you think.The biggest power savings from unified fabrics come from your servers, not from your switches. This is mainly due to the ability to consolidate I/O and in turn reduce your NICs and opt for HBAs. There is some great analysis on this that has been done by one of our solutions architects, Frank D’Agostino, pictured in the video above.Frank and I began discussing the best way to approach a credible efficiency analysis for unified fabrics some time ago. From my perspective there were just too many assumptions to take for a broad-brush approach. Luckily Frank knows a lot about facilities, energy, networking, server, storage and data centers in general. Frank was able to apply some field experience of actual deployments to support the White Paper Energy Efficient Unified Fabrics, a great read.From my perspective, the way Frank has approached this analysis is highly illustrative of taking an aligned perspective towards efficiency analysis. When I say aligned I mean that if you decrease energy use in one infrastructure domain/segment you could very well increase it somewhere else if you don’t consider the larger context. This is where silo’d thinking will in many cases negate efforts to be more energy efficient.Another point that becomes apparent when you look at Frank’s work is the effect the network can have in a data center when deployed correctly. The Network maintains a comparatively small energy footprint in the data center but it can have broad reaching, cross-domain impacts. In this case you’re effectively taking one class of infrastructure (networking) and through the right architecture, decreasing the energy footprint of another class (servers). You can estimate the savings specific to your operation through the Nexus Unified Fabrics TCO calculator found in the planning tools section of the Efficiency Assurance Program. There is also some good reference design guidance in the same section along with the video featured on this post.In our marketing we often refer to this result as the “network effect”. When I am asked to explain this concept to facilities oriented professionals I opt for a simpler example than unified fabrics. Think of it in terms of printers. Back when Green was a noun and not a verb we had a peer to peer printing relationship. What happened when we put an IP address on a printer? 200 printers went to 1 printer overnight. How Green was that?The same printer example applies across a range of network-centric approaches to optimizing energy. Isn’t this to a large extent what the current approaches to virtualization are delivering? Virtual appliance services for example functionally accomplish the same thing as you are reducing the actual hardware you need to deliver services, print or otherwise.One word of caution, if you are looking to take an aligned approach (meaning you are looking at the energy implications of a “solution”) don’t stop your analysis at the IT level. You need to make sure you are also looking at the supporting facilities where there could be as much as 2-3 times the energy overhead. In addition, if a solution leads to consolidation it will change your power support model so it will need to be reassessed anyway.Lastly, if you are implementing the types of virtualization (VMware for example) that can introduce significant dynamic compute loads, dynamic thermal and electric loads will follow. Dynamic power loads can be a real threat to availability if not planned for so make sure your facilities teams are aware of any efforts in this space early and often.Thanks for reading!