The other night I had the opportunity to meet up for a quick dinner while traveling with some folks I used to work with when I wore a vendor hat. Directions to the venue were pretty straightforward as I knew where I was going even with out gps or “ever lost” or other electronic guidance system. During dinner, (I had the crab cakes which were great), not surprising given the backgrounds and current vocations of the people I was dining with, the technology tour and trends around the industry discussion came to that of what is and what is not a director vs. a switch and should new ultra large scale directors like those recently announced by both Brocade and Cisco be referred to as directors or something different.
I’m sure the switching folks from the various camps all have their different views, and if you follow marketing fundamentals, as a vendor you want to position yourself as something new and different to own a market or category. Hence when the discuss turned to why these new platforms were or were not directors and how they were something more, it reminded me of where we were in the industry about seven years ago when the director vs. switch debates raged. So with that in mind, when I was asked as to what I thought a director was vs. a switch, my answer is this:
Switches can be highly available with redundant power supplies, hot swap components, non-disruptive code load and activation, multiple physical interfaces (e.g. Fibre Channel and Ethernet) with multiple protocols (e.g. FCP, FICON, FCIP, iSCSI, etc) and even intelligence and so forth. Switches can even be modular, such as those that enable blocks or groups of ports to be turned on or added via mezzanine or daughter cards or even stackable switches. Switches generally speaking and with all due respect tend to be smaller, however if you go by the standard of what some consider a director vs. a switch, the ultra large scale Qlogic switches make for an interesting quandary if you set size aside. Likewise if you use the test that a director supports FICON for the mainframe then the Brocade/McData switches make for another quandary as some of those support mixed mode.
IMHO a director is this, a switching device on a large scale with modularity to leverage different types of blades supporting different types of physical interfaces (Ethernet, Fibre Channel, InfiniBand, WAN) and protocols (FICON, FCP, iSCSI, FCIP, iFCP, etc) with scalable performance, interoperability, no single point of failure (outside of the chassis, backplane or firmware) and generally much larger than a traditional switch or collection of stackable switches. Put another way, think “big iron” like a mainframe vs. small iron like a small server, director = big iron, switch being smaller volume oriented like volume servers of which there are places for both.
So, to avoid confusing the market place, and in particular those who buy these technologies, if you want to think of a large switch as a big switch, fine, if you want to think of a brand new ultra large platform for converged networks and unified data center fabrics as a director, be my guest as you are not alone, if you want to call it a backbone switch great, after all, IMHO its not as much as what they are called, its being able to speak the language of your audience and address what it is they are looking for instead of teaching them a new language and vocabulary, its the functionality and how the different solutions address your various needs and plug into your environments.