The SAN is dead! Long live the storage fabric!

Current Job Listings

Prior to my days as an analyst I spent many years toiling away in corporate IT. The area within IT that I had the most interest and passion in was the network. It’s a little like when Spock said to Captain Kirk that “commanding a starship was his first, best destiny.” Sure, Kirk might go and try being a diplomat, but at the end of the day he finds his way back to the bridge.

Similarly, I might go dabble with servers, applications or other areas, but I always looked at things through the network lens. Sometimes that’s helpful, like when yanking out that old, legacy PBX in favor of a VoIP system is a no-brainer. However, sometimes the networking thought process could make you look for problems that aren't there.

One such example has been the evolution of the storage area network (SAN). The idea of a totally parallel network running a dedicated, single-purpose protocol like FibreChannel confounds network professionals and is one of the reasons why the networking industry has been beating the drum to put a bullet in the protocol and migrate to something more mainstream like FibreChannel over Ethernet (FCoE). Why? Well, because Ethernet is perceived to be better. It’s simple, it’s ubiquitous, it costs less, and it’s the way everything else has gone.

However, juxtapose this thought process with the conservative nature of the storage buyer. Storage is critical to the success of a company. It’s the company data, the crown jewels if you will, and any kind of problem with storage could mean lost data, fines, lawsuits, and the termination of the employment of said storage manager who was responsible for the integrity of the data. Given that, I now understand the plight of the storage manager and the conservative tendencies. Corporate data is just too important to tinker with, and as the saying goes, “if it ain’t broke, don’t fix it.”

I do believe that the day has come to leverage the best of both worlds and evolve the legacy SAN to a storage fabric. The following are some points to consider when deploying a storage fabric.

  • Choice of protocol. When it comes to FCoE versus FibreChannel, there’s really no right answer. FibreChannel is a tried-and-true technology that is really the Ferrari of storage networking. However, if cost is an issue or the company has made the decision to standardize on Ethernet, then FCoE is a viable option today. A company could actually leverage both technologies and use FibreChannel for mission-critical data and then leverage Ethernet for other data. For large organizations, I would recommend sticking with FibreChannel and keeping the most important data on the best network.
  • Fabric characteristics. Regardless of technology choice, it’s important to deploy the network as a fabric. This means any-to-any-port connectivity, multi-pathing, low-latency transport, continuous uptime and load balancing. FibreChannel has had these characteristics for years, but Ethernet Fabrics have recently caught up. For Ethernet-based environments, a key to success here is to dump spanning tree in favor of TRILL or Shortest Path Bridging to bring in fabric-like characteristics. A fabric also allows for data center interconnectivity, so down the road organizations can create a single, multi-location storage fabric.
  • Diagnostics and analytics. This is an area in which Ethernet has historically outpaced FibreChannel, but Brocade’s Gen5 release of FibreChannel brought with it a number of diagnostic tools and other advancements that have closed the gap with Ethernet. Some of these have become standard in this current generation of FibreChannel and I would expect the next release to evolve even further when it’s made available in about a year.
  • Keep it on a dedicated network. This is a crucial point no matter what technology is used. While we love to converge things in networking, the storage fabric should remain a dedicated network. There’s simply no reason to converge the storage fabric with the data network. I know there’s a cost argument to be made here, but I’m a believer that the downside risk of a converged network outweighs the benefits and assurance of a dedicated fabric.

The era of the SAN is over and now it’s time to start thinking along the lines of a fabric for storage connectivity. As I stated above, the choice of protocol isn’t nearly as important as ensuring the principals of a fabric are followed.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Now read: Getting grounded in IoT