As convergence and SDE (software-defined-everything) make their way into the mainstream and add real value, organizations both large and small battle with the question of, “What should we do about our storage networks?” Stick with Fibre-Channel based approach or, as depreciation cycles end and/or new data-center locations come online, refresh to an Ethernet only solution?
This is not as simple as deciding what technology to go with. There are many factors involved in making this decision which include:
- Acquisition cost
- On-going cost
- Pool of talented technicians
- The internal politics between the LAN/WAN and SAN teams
As vendors and consultants continue to battle it out (“Fibre Channel rocks Ethernet” and “Ethernet Rules the World,”) end-users are stuck in the middle trying to make a very difficult and complex decision with the end result being one that will have a significant impact to the performance and availability of there applications (which at the end of the day is all that really matters.)
There are many commonalities between the requirements for Fibre Channel and Ethernet based storage networks. An enormous one that is often overlooked is that the cable plant required for each one is basically identical.
Generally speaking, OM4 -50um fibre is used for both 8/16 Bb Fibre Channel and 10/40/100 Gb Ethernet in short haul and/or campus based environments. The expense of these cable plants can be quite extensive but when considering Fibre Channel or Ethernet, the cable plant cost for each will be basically a wash.
With this harmony come a few disparities which need to be taken into account when making this decision. Some of the glaring ones include:
Speeds and Feeds
When looking at speeds and feeds Fibre Channel has 4/8/16 Gb (32 Gb standard no gear yet – Targeted 2015 Deployment) and Ethernet has 10/40/100 Gb although 40/100 Gb are usually dedicated to internal switch-to-switch interconnects. With the pervasiveness of Flash/DRAM based storage and alike attaining these speeds in the real world are not as far out as they may have once seemed.
When designing a Fibre Channel or Ethernet storage network it is important to take the same methodical approach to both. The design concepts around physical switch separation, dual HBAs-TOEs-CNAs, zoning, and vlans should be applied in either scenario. This will ensure some of the unfounded reliability claims of Ethernet from years past are just that past. This also helps when constructing the financial model debunking the myth that Fibre Channel is so much more expensive. (Like for Like)
With this in mind, making the determination which technology to choose should be boiled down to an apples-to-apples comparison. As I have traveled down this path in my day job, I made every attempt to take emotions and politics out of the equation (A very difficult task indeed) and focus on the cost, resources and technological benefits of both solutions.
The simplest way to begin is to construct a matrix of features and a TCO (Total Cost of Ownership) / ROI (Return on Investment). This helps greatly reduce the amount of politics in the equation and allows you make the right decision for your specific environment based on your business and application requirements.
In summary, I’d say don’t come down in either camp (Why fight the battle?) Use the matrix and financial models described in this post to take the emotional and political in-fighting out of the equation and rely on the metrics and like for like comparison.